# API keys Source: https://www.wirebase.com/docs/admin-guide/models/api-keys Connect AI providers for your organization. Every member chats with the models your keys unlock. ## Overview **Admin → API Keys** has three tabs: | Tab | Holds | | - | - | | **Provider Keys** | Keys for AI model providers: OpenAI, Anthropic, Google, xAI, Groq, Mistral, DeepSeek, Fireworks, Cohere, Amazon Bedrock, OpenRouter, Ollama and others. | | **Service Keys** | Keys for supporting services, such as the web search provider. | | **Custom Providers** | OpenAI-compatible endpoints you add yourself. See [Custom providers](/docs/admin-guide/models/custom-providers). | ## Add a provider key 1. On **Provider Keys**, find the provider and click **Add**. 2. Paste the key and save. Wirebase validates it with the provider first. 3. The provider's default models are enabled automatically. Fine-tune them in [Models](/docs/admin-guide/models/models). **Quick start** offers two shortcuts: **Connect** OpenRouter (authorize Wirebase to create a key, no copy and paste), and **Get a free key**, which lists providers with free tiers. ## Security * Keys are encrypted at rest and never displayed again after saving. To change one, click **Edit key** and enter a new value. * Keys are scoped to your organization. Other organizations can't see or use them. * Keys in server environment variables are ignored; keys are only ever configured here. * Every create, update and delete is recorded in the [audit log](/docs/admin-guide/usage/audit-logs). ## Remove a key Click **Delete key**. All of that provider's enabled models are removed and members can no longer select them. ## Which keys does each feature need? | Feature | Needs | | - | - | | Chat | Any provider key with enabled chat models. | | Knowledge bases and project files | An **OpenAI** or **OpenRouter** key, for embeddings. | | Image generation | A provider with image models, such as OpenAI or Google. | | Web search | A search key under **Service Keys** (Exa). See [Chat features](/docs/admin-guide/models/chat-features). | # Chat features Source: https://www.wirebase.com/docs/admin-guide/models/chat-features Turn chat capabilities on or off for your organization, and set up web search. ## Overview **Admin → Settings → Chat Features** controls which chat capabilities members can use. | Feature | What it controls | | - | - | | **Voice Chat** | Voice input in chat. | | **File Upload** | Attaching files to chat messages. | | **Image Generation** | Generating images in chat. | | **Web Search** | The built-in web search tool. Needs a search provider and key (below). | | **Code Execution** | Running JavaScript and Python in chat. Use with caution. | | **Temporary Chat** | Starting temporary chats that aren't saved. | Changes apply to every member of the organization. ## Set up web search 1. Turn on **Web Search**. 2. Under **Search Provider**, select **Exa**. 3. Click **Add API key** and paste your Exa key. Web search keys are stored per organization and encrypted like provider keys. There is no server-wide fallback key. ## Related settings * To turn connectors off for the whole organization, see [Integrations](/docs/admin-guide/settings/integrations). * To control which models members can use, see [Models](/docs/admin-guide/models/models). # Custom providers Source: https://www.wirebase.com/docs/admin-guide/models/custom-providers Use any OpenAI-compatible endpoint, such as vLLM, LM Studio or an internal gateway, as a model provider. ## Overview A custom provider connects Wirebase to any endpoint that implements the OpenAI Chat Completions API, for example a self-hosted vLLM server, an internal AI gateway or a provider Wirebase doesn't list. ## Add a custom provider Go to **Admin → API Keys → Custom Providers** and click **Add Custom Provider**. | Field | Description | | - | - | | **Name** | A label for the provider, such as "Internal vLLM". | | **Endpoint URL** | The base URL, for example `https://llm.internal.example.com/v1`. A trailing `/chat/completions` path is normalized automatically. | | **API Key** | Sent as `Authorization: Bearer …`. Leave blank if the endpoint doesn't require a key. | | **Custom Headers** | Optional extra headers sent with every request. | | **Models** | Model IDs to expose (for example `llama-3.1-70b`), each with an optional display label. Models are entered manually. | Save, then enable the models in [Models](/docs/admin-guide/models/models). The endpoint must be reachable from the public internet over HTTPS. For guidance on what the endpoint needs to support, see [OpenAI-compatible providers](/docs/developer-guide/openai-compatible-providers). ## Edit or remove Use **Edit** to change the URL, key, headers or model list, and **Delete** to remove the provider and its models. # Models Source: https://www.wirebase.com/docs/admin-guide/models/models Choose which models members can use, see their capabilities, and set pricing for spend reporting. ## Overview **Admin → Models** lists every model available from the providers you've added keys for. Model lists, context lengths, capabilities and pricing are maintained by Wirebase and appear once a provider is configured. ## Enable and disable models 1. Select a provider. 2. Switch models on or off. **Enabled** models appear in every member's model selector; **Available** ones don't. 3. Save your changes. Use the search box and filters (**Enabled only**) to find models quickly. Click **Refresh** to pull the latest models from every configured provider. ## Model details Click a row to see a model's details: * **Identity**: model ID, provider, modality, release date and knowledge cutoff. * **Limits**: context length and maximum output. * **Capabilities**: tool calling, vision (image input), file upload, JSON mode, reasoning, prompt caching, web search, image generation. * **Pricing per 1M tokens**: input, output, cache read and cache write. * **Status**: enabled, deprecated, whether it's still in the provider's catalog, and when it was last synced. Members can only use @mentions, agents, workflows and connectors with models that support **tool calling**. Keep at least one such model enabled. ## Override pricing Spend in [reporting](/docs/admin-guide/usage/reporting) and [usage limits](/docs/admin-guide/usage/usage-limits) is calculated from each model's price per token. If your organization has negotiated rates, or a model has no list price (for example a self-hosted model), set your own input and output price per 1M tokens. **Reset to provider default** removes your override. Wirebase doesn't bill you for model usage. Providers bill the key owner directly. Spend figures in Wirebase are estimates for monitoring and budgeting. # Admin overview Source: https://www.wirebase.com/docs/admin-guide/overview What organization admins can manage in Wirebase, and where to find it. ## Open the admin dashboard Click your avatar in the sidebar and choose **Admin**. The option only appears for organization **admins**. The **Dashboard** shows total users, who's signed in now, messages in the last 7 days, storage, recent activity and quick links to common tasks. ## What you can manage Connect AI providers for your whole organization. Choose which models members can use, and set pricing. Invite people, change roles, suspend or remove members. Group members for sharing projects. Set monthly AI budgets for the org, roles or people. Spend, adoption, reliability and tool usage. Every administrative and security-relevant action. Single sign-on and automatic user provisioning. Serve Wirebase on your own domain, white-labelled. Plans, seats and invoices. ## Organizations Everything in Wirebase belongs to an organization: API keys, models, agents, workflows, connectors, knowledge and chats. Data never crosses between organizations, and admin settings apply only to your own organization. # Members & roles Source: https://www.wirebase.com/docs/admin-guide/people/members-and-roles Invite people to your organization, assign roles, and suspend or remove members. ## Invite members Go to **Admin → Users** and click **Add Member**. Enter the person's **Email Address**, choose a **Role**, and optionally a **team** they'll join when they accept. They receive an email with a link to join. Pending invitations are listed alongside members, where you can resend or **Cancel Invitation**. Admins and editors can invite members. The Free plan includes up to 3 members. On the Team plan you pay per seat; if you're out of seats, the invite dialog offers to add more. See [Billing](/docs/admin-guide/settings/billing). ## Roles Each member has one role in the organization. | Capability | Member | Editor | Admin | | - | :-: | :-: | :-: | | Chat, temporary chats, projects, memory | ✓ | ✓ | ✓ | | Create and share agents | ✓ | ✓ | ✓ | | Use shared workflows and connectors | ✓ | ✓ | ✓ | | Create and edit workflows | | ✓ | ✓ | | Add and manage connectors | | ✓ | ✓ | | Invite members | | ✓ | ✓ | | Open the admin dashboard: keys, models, members, teams, limits, reporting, settings | | | ✓ | New invitees default to **Member**. Whoever creates an organization is its first **Admin**. ## Manage members In **Admin → Users**, search by name or email, then use a member's menu to: * **Change Role**. * **Suspend Member**: they lose access to this organization for 7 days. Their other organizations aren't affected. **Lift Suspension** restores access early. * **Remove Member**: removes them from the organization. This can't be undone. Click a member to see their details and AI usage over time. Every role change, suspension and removal is recorded in the [audit log](/docs/admin-guide/usage/audit-logs). ## Provision automatically With an identity provider, you can create and remove members automatically with [SCIM](/docs/admin-guide/security/scim), and have them sign in with [SSO](/docs/admin-guide/security/sso). # Teams Source: https://www.wirebase.com/docs/admin-guide/people/teams Group members within your organization, for example Engineering or Sales. ## Overview Teams are groups of members inside your organization. Use them to share a [project](/docs/user-guide/chat/projects) with just one group instead of everyone. ## Create a team 1. Go to **Admin → Teams** and click **Create Team**. 2. Enter a name, for example "Engineering". 3. Click **Manage members** and add organization members. You can also put new people on a team when you invite them. They join the team when they accept. ## Manage teams From a team's menu you can **Rename**, **Manage members** (add, or **Remove from team**) or **Delete** it. Deleting a team doesn't remove its members from the organization. Only admins can create and manage teams. Members can see the teams they belong to. # Custom domain Source: https://www.wirebase.com/docs/admin-guide/security/custom-domain Serve your Wirebase workspace on your own domain, with your branding and access rules. ## Overview Map a domain you own, such as `ai.yourcompany.com`, to your organization. Members use Wirebase at that address, with a sign-in page in your brand, and you control who can sign in there. Custom domain and white-label are part of the **Enterprise** plan. Configure them in **Admin → Settings → Custom Domain**. ## Connect your domain Enter the domain or subdomain you want to use, for example `ai.yourcompany.com`. At your DNS provider, add the **CNAME** record shown on the page. It routes traffic to Wirebase, and an SSL certificate is issued automatically. Add the **TXT** record shown on the page at `_wirebase-verify.`. It proves you own the domain. Once both records are in place, click **Verify**. DNS changes can take up to 48 hours to propagate. When verification succeeds, the status changes from **Pending Verification** to **Active**. If your domain has a CAA DNS record, it must allow the certificate authority used for custom domains (`ssl.com`), or the SSL certificate can't be issued. If you have no CAA record, there's nothing to do. ## Access rules The **Access** tab controls sign-in on your domain only. Sign-in on [www.wirebase.com](http://www.wirebase.com) is unaffected, and every rule is enforced server-side. | Rule | Effect | | - | - | | **Members of this organization only** | Anyone signed in with an account outside your organization is blocked on this domain. | | **Email and password** | Turn off to require single sign-on. This also disables password reset on the domain. | | **Single sign-on (SAML / OIDC)** | Sign in through your configured [identity provider](/docs/admin-guide/security/sso). | | **Open sign-up** | Off by default: new accounts can only be created by invitation or through your IdP. | | **Allowed sign-up email domains** | Restrict sign-ups to specific email domains, such as `yourcompany.com`. | Google, GitHub and Microsoft sign-in aren't available on a custom domain. Use SSO or invitations instead. ## White-label branding The **White-label** tab customizes the sign-in page on your domain: **Brand Color**, **Favicon URL**, **Login Page Title**, **Login Page Copy** and **Custom CSS**, with a live **Login Page Preview**. Your organization's name and logo come from [Branding](/docs/admin-guide/settings/branding). ## Remove the domain Click **Remove Domain**. The domain stops routing to your organization immediately, its SSL certificate is revoked, and its white-label settings are deleted. # SCIM provisioning Source: https://www.wirebase.com/docs/admin-guide/security/scim Create, update and deactivate Wirebase users automatically from your identity provider. ## Overview SCIM keeps your Wirebase organization in sync with your identity provider (IdP). When you assign someone to the Wirebase app in your IdP, they're created in Wirebase; when you unassign them, they're deprovisioned. SCIM is available on the **Enterprise** plan. Configure it in **Admin → Settings → SCIM**. ## Set up SCIM Click **Generate SCIM Token**, enter a **Provider ID** (for example `okta-scim`), and copy the token. It's shown only once. In your IdP's SCIM settings: | Setting | Value | | - | - | | **SCIM Base URL** | `https://www.wirebase.com/api/auth/scim/v2/` (also shown on the page) | | **Authentication Method** | HTTP header / Bearer token | | **Token** | The token from step 1 | Use your IdP's test action to verify it works, then assign users or groups to the app. Active connections are listed under **Active Connections**. Generating tokens and deleting connections is recorded in the [audit log](/docs/admin-guide/usage/audit-logs). Combine SCIM with [SSO](/docs/admin-guide/security/sso) so provisioned users sign in through your IdP. # Single sign-on (SSO) Source: https://www.wirebase.com/docs/admin-guide/security/sso Let members sign in through your identity provider with OIDC or SAML 2.0. ## Overview With SSO, members sign in to Wirebase through your identity provider (IdP), such as Okta, Microsoft Entra ID or Google Workspace. Users whose email domain matches an SSO provider are signed in through it. | Protocol | Plan | | - | - | | **OIDC** (OpenID Connect) | Team and Enterprise | | **SAML 2.0** | Enterprise | Configure SSO in **Admin → Settings → SSO**. ## Add an OIDC provider Create an OIDC web application. Set its redirect (callback) URL to: ```text theme={null} https://www.wirebase.com/api/auth/sso/callback/ ``` Replace `` with the ID you'll choose in the next step. Click **Add OIDC** and fill in: | Field | Example | | - | - | | **Provider ID** | `okta-prod`: a unique slug for this provider | | **Issuer URL** | `https://accounts.google.com` | | **Client ID** / **Client Secret** | From your IdP app | | **Email Domain** | `yourcompany.com`: users with this email domain use this provider | ## Add a SAML provider Click **Add SAML** and enter a **Provider ID**. The dialog then shows the values to configure in your IdP: | IdP setting | Value | | - | - | | **SP Entity ID / Audience URI** | Shown in the dialog | | **ACS / Reply URL / Callback URL** | `https://www.wirebase.com/api/auth/sso/saml2/callback/` | | **SP Metadata URL** | `https://www.wirebase.com/api/auth/sso/saml2/sp/metadata?providerId=` | Use the values above, then download your IdP's metadata XML. Paste the IdP metadata, set the **Email Domain**, and save. If you serve Wirebase on a [custom domain](/docs/admin-guide/security/custom-domain), use the URLs exactly as the dialog shows them; they reflect the domain you're configuring from. ## How members sign in On the sign-in page, members click **Sign in with SSO** and enter their work email. Wirebase sends them to your IdP and, after they authenticate, signs them in to your organization. New users from your domain are added to your organization automatically. ## Remove a provider Delete it from the providers list. Creating and deleting SSO providers is recorded in the [audit log](/docs/admin-guide/usage/audit-logs). ## Related * [SCIM](/docs/admin-guide/security/scim): create and deactivate users automatically. * [Custom domain](/docs/admin-guide/security/custom-domain): require SSO on your own domain. # Plans & billing Source: https://www.wirebase.com/docs/admin-guide/settings/billing Wirebase plans, seats, trials and managing your subscription. ## Plans Wirebase charges for seats and features, not for AI usage. You bring your own provider keys and pay providers directly, at their prices. | | Free | Team | Enterprise | | - | - | - | - | | Price | \$0 | $10 per member/month ($8 billed yearly) | Custom | | Members | Up to 3 | Unlimited (pay per seat) | Unlimited | | All AI models, agents & workflows | ✓ | ✓ | ✓ | | Website chat widget | | ✓ | ✓ | | SSO (OIDC) | | ✓ | ✓ | | SAML 2.0 SSO | | | ✓ | | SCIM directory sync | | | ✓ | | Custom domain & white-label | | | ✓ | | Automations | 2, daily at most | 20, hourly | 100, hourly | | Audit log retention | 30 days | 90 days | Unlimited | | File storage | 1 GB | 10 GB | Unlimited | | Support | Community | Priority | Dedicated success manager | The Team plan starts with a **14-day free trial**. See [wirebase.com/pricing](https://www.wirebase.com/pricing) for current prices. ## Manage your subscription Go to **Admin → Settings → Billing** to see your **Current Plan** and **Seats**, switch between **Monthly** and **Yearly** (save 20%), and click **Manage Subscription** to update your payment method, change your seat count, view invoices or cancel. On Enterprise, contact your account manager to adjust seats or pricing. ## Seats On the Team plan each member uses a seat. If you invite more people than you have seats, you'll be asked to purchase additional seats. ## Payment problems If a payment fails, the billing page shows **Payment failed** and asks you to update your payment method. If it isn't resolved, access can be suspended until it is. # Branding & theme Source: https://www.wirebase.com/docs/admin-guide/settings/branding Customize your organization's name, logo and the look of the chat interface. ## Branding **Admin → Settings → Branding** sets how your organization appears: | Field | Where it's used | | - | - | | **Organization Name** | The sidebar and throughout the app. | | **Logo** | Upload an image or enter a URL. Shown in the sidebar. | | **Tagline** | The main screen and the page title. | | **Description** | Social media link previews and search results. | ## Chat theme **Admin → Settings → Chat Theme** changes the look of the chat interface for every member. 1. Pick a **Base Theme**. 2. Optionally override individual colors under **Color Overrides**: core colors (background, foreground, primary), secondary and accent, surfaces (card, popover, border, input), the sidebar and chart colors. 3. Use **Reset to theme default** on a color, or **Reset All**. ## Your own domain To brand the sign-in page and serve Wirebase on your own domain, see [Custom domain](/docs/admin-guide/security/custom-domain). # Delete an organization Source: https://www.wirebase.com/docs/admin-guide/settings/delete-organization Permanently delete your organization and all of its data. Deleting an organization is permanent and can't be undone. ## What gets deleted From **Admin → Settings → Danger Zone**, **Delete Organization** permanently: * Deletes all chat threads and messages. * Deletes all agents, workflows and connectors. * Deletes all uploaded files from storage. * Revokes all API keys. * Removes the custom domain configuration. * Deletes the accounts of members who have no other organizations. ## Delete your organization 1. Go to **Admin → Settings → Danger Zone**. 2. Click **Delete Organization**. 3. Type the organization's name to confirm. Organizations on a paid plan can't be deleted from the dashboard. [Contact support](https://www.wirebase.com/contact) and we'll help you. ## Export first Before deleting, members can download conversations they want to keep as Markdown or JSON. See [Share & export](/docs/user-guide/chat/share-and-export). # Emails Source: https://www.wirebase.com/docs/admin-guide/settings/emails Send invitations, password resets and notifications from your own mail server, and check deliverability. ## Overview **Admin → Settings → Emails** configures outgoing email for your organization: invitations, password resets and notifications. ## Deliverability The top of the page shows how recent emails fared: **Sent**, **Delivered**, **Delayed**, **Failed**, **Bounced** and **Complained**, plus a list of recent failures. ## Use your own SMTP server Under **SMTP Settings**, enter: | Field | Example | | - | - | | **Host** | `smtp.example.com` | | **Port** | `587` | | **Username** / **Password** | Your SMTP credentials. Leave the password blank to keep the existing one. | | **From Email** | `noreply@example.com` | | **From Name** | The sender name shown on emails. | Save, then send a test email. Test emails use your saved settings, so save first to test changes. ## Notifications Turn **Email Notifications** on or off to control whether members of your organization receive notification emails. # Integrations Source: https://www.wirebase.com/docs/admin-guide/settings/integrations Govern which integrations and connectors are available in your organization. ## Overview **Admin → Settings → Integrations** controls access to tools from outside services. ## Connectors Turn **Enable Connectors** off to stop members from adding and using [connectors](/docs/user-guide/workspace/connectors) (MCP servers) in chat. Turn it back on to restore access; existing connectors are kept. ## Connected Apps When available on your deployment, **Connected Apps** lets members connect accounts for popular apps and use their tools in chat. Use **Enable Integrations** to turn it on or off, and **App Access** to allow all apps or restrict members to selected ones. # Audit logs Source: https://www.wirebase.com/docs/admin-guide/usage/audit-logs A record of administrative and security-relevant actions in your organization. ## Overview **Admin → Reporting → Audit Logs** lists actions taken in your organization, with the timestamp, action, status (success or failure), resource, user, IP address and details. ## Filter and export * Filter by action, resource type and status, or click a user to see only their actions. * Click **CSV** or **JSON** to export the current view. ## What's recorded | Area | Events | | - | - | | Users & sign-in | User created or updated, sign-ins, platform role changes, bans | | Members | Role changes, suspensions and reinstatements, removals, invitations cancelled or resent | | Organization | Settings updates, organization deletion | | Teams | Team created, renamed or deleted; members added or removed | | Content | Agents, workflows, prompts, automations and projects created, updated, deleted or shared; chat imports | | Website widget | Enabled, disabled, updated, key regenerated | | Connectors & providers | Connectors added or removed; custom providers and API keys created, updated or deleted | | Security | Custom domain added, verified, removed or policy changed; SSO providers and SCIM connections created or deleted | | Usage | Quotas created, deleted or exceeded; pricing overrides; storage quota exceeded; reporting exports | ## Retention | Plan | Audit logs kept for | | - | - | | Free | 30 days | | Team | 90 days | | Enterprise | Unlimited | # Reporting Source: https://www.wirebase.com/docs/admin-guide/usage/reporting Spend, adoption, reliability and tool usage across your organization. ## Overview **Admin → Reporting** shows how your organization uses AI. Pick a date range at the top of each page. | Page | What it shows | | - | - | | **Overview** | Spend, AI requests, active users, cost per active user, error rate and p95 latency, plus top models, members and tools. | | **Usage & Cost** | Spend and token usage by member and by model. | | **Activity** | Engagement and token usage: messages, active conversations, usage over time, top models and top users. | | **Tools** | Which tools and connectors are used, how often, and how often they fail. | | **Audit Logs** | Every administrative and security-relevant action. See [Audit logs](/docs/admin-guide/usage/audit-logs). | Click a member to see their individual usage. ## Spend figures Spend is estimated from each model's price per token. If a model has no price, its spend is under-reported, and the page shows **Spend is under-reported** with a link to **Set model pricing**. See [Models](/docs/admin-guide/models/models). ## Export Use **Export** to download **Spend by member**, **Spend by model**, **Tool usage** or **Daily trend** for the selected range as CSV. Tool calls are recorded from new conversations onward, so older history doesn't appear on the Tools page. # Usage limits Source: https://www.wirebase.com/docs/admin-guide/usage/usage-limits Set monthly AI budgets for the whole organization, for roles, or for specific people. ## Overview Usage limits help you keep AI spend predictable. Each limit is a monthly budget in US dollars, calculated from model token prices (see [Models](/docs/admin-guide/models/models) to override prices). Wirebase never caps your usage on its own. Limits are budgets **you** choose to set; with none configured, usage is unlimited and billed by your providers as usual. ## Add a limit Go to **Admin → Access & Limits → Usage Limits** and click **Add Quota**. | Setting | Options | | - | - | | **Scope** | **Org-wide**, **By Role** (Admin, Editor or User), or **Specific Users**. | | **Monthly Budget** | An amount in USD, for example `50.00`. Leave blank for no limit. | | **Enforcement Mode** | How the limit is applied when it's reached (below). | ### Enforcement modes | Mode | When the budget is reached | | - | - | | **Hard Block** | Members can't send new messages until the next billing period. | | **Soft Cap** | Usage continues, with warnings. | | **Warn Only** | Alerts are shown; nothing is blocked. | Members see a notice in chat when they're approaching their budget. When a quota is exceeded, it's recorded in the [audit log](/docs/admin-guide/usage/audit-logs). # Code node Source: https://www.wirebase.com/docs/developer-guide/code-node Transform data in a workflow with sandboxed JavaScript. ## Overview The Code node runs your JavaScript on the server in an isolated sandbox. Use it for deterministic transformations that would be awkward or unreliable with an LLM: parsing, filtering, math, reshaping JSON. ## How it works Your code is the **body of a function**: * Inputs arrive on a `params` object. * `return` a value to produce the node's output. It must be JSON-serializable. ### Params Under **Params**, add keys and set each value to a literal or a `/` reference to an earlier node's output. Inside your code, read them as `params.`. ### Result schema Optionally describe the returned value's shape under **Result Schema** so later nodes can reference its individual fields. Leave it as-is to keep the output an untyped object. ## Example With params `orders` (referencing an HTTP node's response) and `minTotal` set to `100`: ```javascript theme={null} const big = params.orders.filter((o) => o.total >= Number(params.minTotal)); return { count: big.length, revenue: big.reduce((sum, o) => sum + o.total, 0), ids: big.map((o) => o.id), }; ``` ## Limits | Limit | Value | | - | - | | Runtime | Synchronous JavaScript only: no `await`, timers or promises. | | Network and files | None. Use an **HTTP** node to fetch data, then pass it in as a param. | | Timeout | 10 seconds by default, up to 30 seconds. | | Memory | 64 MB. | | Code size | 64,000 characters. | | Console output | `console.log` output is captured (up to 100 entries) to help you debug. | ## Debugging tips * Click **Run** in the workflow editor, enter test inputs, and open each node's result to inspect what it received and returned. * Throw an `Error` with a clear message when input isn't what you expect; the workflow stops and shows it. # Build an MCP connector Source: https://www.wirebase.com/docs/developer-guide/mcp-connectors Expose your own API to Wirebase as tools by running a remote MCP server. ## Overview Wirebase is an MCP **client**. Any remote server that speaks the [Model Context Protocol](https://modelcontextprotocol.io) can be added as a [connector](/docs/user-guide/workspace/connectors), and its tools become available in chat, agents and workflows. ## Requirements | Requirement | Details | | - | - | | **Transport** | Streamable HTTP (recommended) or SSE. Wirebase tries Streamable HTTP first and falls back to SSE. | | **Reachability** | A public `https://` URL Wirebase can reach. Local (stdio) servers aren't supported. | | **Capabilities** | Tools. Each tool needs a name, a description and a JSON Schema for its input. | ## Authentication Choose one: Accept a token in a header, typically `Authorization: Bearer `. The person adding the connector enters it under **Headers**. Implement the MCP authorization spec (OAuth 2.1 with PKCE). Wirebase discovers your authorization server from the MCP server's metadata and registers itself with dynamic client registration: | Client metadata | Value | | - | - | | `client_name` | `wirebase-` | | `redirect_uris` | `https://www.wirebase.com/api/mcp/oauth/callback` | | `grant_types` | `authorization_code`, `refresh_token` | | `token_endpoint_auth_method` | `none` (public client, PKCE) | | `scope` | `mcp:tools` | Users click **Authorize** and complete sign-in in a popup. Tokens are encrypted at rest and refreshed automatically. By default each user authorizes separately; the connector's owner can turn on **Shared Authentication** instead. ## Tool naming Tools appear in Wirebase as `_`, for example `github_create_issue`. Names are sanitized to letters, numbers, underscores and hyphens and capped at 124 characters. Keep tool names short and descriptive. ## Write tools that models use well * **Describe when to use the tool**, not just what it does. The description is what the model reads when deciding. * **Keep input schemas small and typed.** Use enums for fixed choices and mark required fields. * **Return compact results.** Tool output is truncated at about 40,000 characters before it reaches the model. Return the fields that matter, and paginate large lists. * **Return clear errors.** A short, specific error message lets the model correct itself and retry. Admins can add per-tool and per-connector [custom instructions](/docs/user-guide/workspace/connectors#customize-how-tools-are-used) to steer usage without changing your server. ## Test your server Add it under **Workspace → Connectors → Add custom connector**, open it, and use **Tools Test** to call each tool with sample input before using it in chat. # OpenAI-compatible providers Source: https://www.wirebase.com/docs/developer-guide/openai-compatible-providers Connect self-hosted or internal models through an OpenAI-compatible Chat Completions endpoint. ## Overview Admins can add any OpenAI-compatible endpoint as a [custom provider](/docs/admin-guide/models/custom-providers): self-hosted servers such as vLLM, an internal AI gateway, or a provider Wirebase doesn't list. ## Endpoint requirements * Implements `POST /chat/completions` in the OpenAI format, including streaming (`stream: true`) with server-sent events. * Reachable from the public internet over HTTPS. * Accepts the API key as `Authorization: Bearer `, or needs no key. Add any other headers your gateway needs under **Custom Headers**. * Supports function calling (`tools`) if you want members to use tools, agents, workflows and connectors with these models. ## Configure Enter the base URL, for example `https://llm.internal.example.com/v1`. Wirebase appends `/chat/completions`; if you paste the full path, it's normalized. Then list the model IDs your endpoint serves, exactly as the endpoint expects them, and enable them in **Admin → Models**. ## Pricing Custom models have no list price, so their spend shows as under-reported in [reporting](/docs/admin-guide/usage/reporting). Set input and output prices per 1M tokens in **Admin → Models** to include them in spend and [usage limits](/docs/admin-guide/usage/usage-limits). # Developer overview Source: https://www.wirebase.com/docs/developer-guide/overview Embed Wirebase agents on your site, connect your own services, and extend workflows with code. ## What you can build Put an agent on any website as a chat bubble with one script tag. Expose your own APIs to Wirebase as tools by running a remote MCP server. Plug in self-hosted or internal models through an OpenAI-compatible endpoint. Reference for every node: LLM, Condition, Tool, HTTP, Template, Code and more. ## For AI assistants and crawlers These docs are available in machine-readable form: * [`/docs/llms.txt`](https://www.wirebase.com/docs/llms.txt): an index of every page. * [`/docs/llms-full.txt`](https://www.wirebase.com/docs/llms-full.txt): the full text of the docs. * Append `.md` to any page URL for its Markdown source. Use the menu at the top of any page to copy it as Markdown or open it in ChatGPT or Claude. # Website widget Source: https://www.wirebase.com/docs/developer-guide/website-widget Embed a Wirebase agent on your website as a chat bubble. ## Overview The website widget puts one of your [agents](/docs/user-guide/workspace/agents) on your site as a chat bubble in the bottom-right corner. Visitors chat without signing in. The agent answers from its knowledge bases using your organization's API keys. The website widget is available on the **Team** and **Enterprise** plans. ## Set up the widget Go to **Workspace → Agents**, open the agent and save it if it's new. Scroll to **Publish → Website chat widget**. Turn on **Turn on website widget**, then set: | Setting | Description | | - | - | | **Model** | The model visitors chat with. Usage counts against your organization's quota. | | **Websites** | The sites the widget may appear on, such as `www.example.com`. If people visit with and without `www`, add both. Up to 20; wildcards aren't supported. | | **Welcome message** | Optional first message, up to 500 characters. | | **Accent color** | The launcher and highlight color. | Click **Save widget settings**. Copy the **Embed code** and paste it just before the closing `` tag on every page that should show the widget. ## Embed code ```html theme={null} ``` | Attribute | Required | Description | | - | - | - | | `data-agent-key` | Yes | The agent's public embed key (`wbe_…`). Copy it from the widget settings. | | `data-accent` | No | A 6-digit hex color, such as `#2563eb`. Defaults to blue. | The script adds a launcher button and, when a visitor opens it, a chat panel. It's self-contained: it runs in a closed shadow root, adds no globals, loads no other dependencies, and doesn't leak styles into your page or pick up yours. The chat runs in a sandboxed iframe that can't navigate your page. ### Single-page apps Load the script once, for example in your root layout. It mounts a single widget per page, even if the script tag is rendered again. ## What visitors' chats can do For safety, widget chats are more limited than chats inside Wirebase: * The agent can search **its own knowledge bases** only. Connectors, other tools and web search are off. * Messages are limited to 4,000 characters, and the conversation sent to the model is capped at the last 20 messages. * Visitors sending messages too quickly are rate limited. * Each chat shows **Powered by Wirebase**. ## Security * **Allowed websites**: the widget only renders on the sites you list. On any other site, the chat page refuses to be framed. * **Key rotation**: click **Regenerate key** to issue a new key. The old key stops working immediately, so update the snippet everywhere it's used. * **Audit trail**: enabling, disabling, updating and regenerating the key are recorded in the [audit log](/docs/admin-guide/usage/audit-logs). ## Troubleshooting Check the browser console for `[wirebase] embed: missing or invalid data-agent-key`. Confirm the widget is turned on and saved, and that the exact host of the page (including or excluding `www`) is listed under **Websites**. Pages opened from a file (`file://`) can't show the widget. The widget was turned off, the key was regenerated, or the agent or its model is no longer available. Check the agent's widget settings. # Workflow node reference Source: https://www.wirebase.com/docs/developer-guide/workflow-nodes Every workflow node type and its settings. ## Passing data between nodes Every node can use the outputs of nodes that run before it. In text fields, type `/` to insert a reference to an earlier node's output, or to one field of it when the output has a schema. References are resolved when the workflow runs. Workflows run from the **Input** node through the graph in dependency order. Branches that a **Condition** doesn't take are skipped. Cycles aren't allowed. ## Input The entry point. Define the workflow's parameters as **Input Fields**: each has a variable name, type, description and optional default value. When the assistant calls the workflow as a tool, it provides values matching this schema, so describe each field clearly. ## Output The exit point. Choose which values from earlier nodes to return. The result is what the assistant receives from the tool call. ## LLM Generates text with a model. | Setting | Description | | - | - | | **Model** | Any enabled model. | | **Messages** | The prompt. Use `/` to reference earlier outputs. | | **Structured Output** | Return a JSON object matching a schema instead of text. Edit the schema in **Output Schema Editor**, or **Generate Schema with AI** from example JSON. | Structured output is useful for extraction, classification and reshaping data, not just writing. ## Condition Branches the workflow with **if**, **else if** and **else**. Each condition compares an earlier value using an operator for its type: | Type | Operators | | - | - | | Text | equals, not equals, contains, not contains, starts with, ends with, is empty, is not empty | | Number | equals, not equals, greater than, less than, greater than or equal, less than or equal | | Boolean | is true, is false | ## Tool Runs a tool from a [connector](/docs/user-guide/workspace/connectors). Pick the tool and write instructions in the message; a model fills in the tool's parameters from your instructions and the referenced data. ## HTTP Calls an external API or webhook. | Setting | Description | | - | - | | **URL** | Literal, or a reference to an earlier output. | | **Method** | `GET`, `POST`, `PUT`, `PATCH`, `DELETE` or `HEAD`. | | **Headers** | Key/value pairs; values can be references. | | **Query** | Key/value query parameters; values can be references. | | **Body** | Request body; can be a reference. | | **Timeout** | Default 30 seconds. | For security, requests to private or internal network addresses are refused. ## Template Builds text, such as an email or report, by mixing literal text with `/` references to earlier outputs. ## Code Runs JavaScript in a sandbox to transform data. See [Code node](/docs/developer-guide/code-node). ## Note A comment on the canvas for your team. It doesn't run. # Settings Source: https://www.wirebase.com/docs/user-guide/account/settings Your profile, chat preferences, memory, shared links and data. Open settings from your avatar in the sidebar → **Settings**. ## Profile Update your name and avatar, change your password, and see your account details. ## Statistics See your own AI usage: which models you used and how much, over time. ## User Instructions Tell the assistant about yourself once, and it applies to every chat: * **What should we call you?** * **What best describes your work?** For example, software engineer or product manager. * **What personal preferences should be taken into account in responses?** For example, "keep explanations brief" or "ask clarifying questions before giving detailed answers". You can also rename the assistant with **Assistant Name**. Shortcut: ⌘ Shift L. ## Memory Turn memory on or off and manage saved memories. See [Memory](/docs/user-guide/chat/memory). ## Connector Instructions Your own instructions for how connectors should be used in your chats. ## Chat display Choose how much of each tool call is shown by default: **Compact**, **Expanded** or **Debug**. See [Tools](/docs/user-guide/chat/tools). ## Shared Links Every public link you've created, with **Unshare** to revoke it. See [Share & export](/docs/user-guide/chat/share-and-export). ## Archived Chats Chats you've archived. Open one, or **Unarchive Chat** to return it to the sidebar. ## Import chats Import conversations from ChatGPT or Claude. See [Import chats](/docs/user-guide/chat/import-chats). ## Data Management **Delete All Chats** permanently deletes all your conversations. This can't be undone. ## Language and theme Switch the interface language and light or dark theme from the menu under your avatar. Wirebase is available in English, Spanish, French, Japanese, Korean, Norwegian and Chinese. # Compare models Source: https://www.wirebase.com/docs/user-guide/chat/compare Send one prompt to two or three models, watch them answer side by side, and keep the best answer. ## Overview Compare mode runs the same prompt on several models at once so you can judge quality, speed and cost before you commit to one. ## Run a comparison Click the compare icon next to the model selector, or go to `/compare`. Choose two or three models with **Add model**. Type your prompt and click **Compare**. The answers stream in side by side. Click **Use this answer** on the one you like. It's saved as a new chat with that model, and you can keep going from there. Each answer shows timing and size details: time to first token, total time and input/output tokens. ## Things to know * Compare is **text only**: tools, connectors and attachments are turned off. * Answers you don't keep are not saved. * Only the final answer is saved. A model's reasoning isn't kept. * You need at least two enabled models. If only one is available, ask an admin to enable more models or add another provider key. # Files & images Source: https://www.wirebase.com/docs/user-guide/chat/files-and-images Attach documents and images to a message, generate images, and dictate with your voice. ## Attach files Click **+** next to the message box and choose **Upload File**, or drag files onto the chat. You can attach images, PDFs, text and Markdown, CSV and JSON, Word, Excel and PowerPoint files, archives, audio and video. Each file can be up to 50 MB. Whether the model can read a file depends on the model. Most current models read images; many read PDFs directly. If a model can't read an attachment, switch to one that lists **Image Input** or **File Upload** in its capabilities. For documents you'll ask about repeatedly, add them to a [project](/docs/user-guide/chat/projects) or a [knowledge base](/docs/user-guide/workspace/knowledge) instead of attaching them to every chat. Your admin can turn file uploads off for the organization. Organizations on the Free plan have 1 GB of file storage. ## Generate images Click **+ → Generate Image** and describe the image you want. If more than one image model is enabled, pick one from the menu. The image appears in the conversation, and you can ask for changes or regenerate it. Image generation uses image models from your organization's providers, such as OpenAI and Google. If you see **No image models enabled**, ask an admin to add a key for a provider with image models. Admins can turn image generation off in **Admin → Settings → Chat Features**. ## Dictate a message Click the microphone icon and speak. Your words appear in the message box as you talk; click again to stop. Review the text, then send it. Dictation uses your browser's built-in speech recognition, so it works in browsers that support it, such as Chrome, Edge and Safari. # Import chats Source: https://www.wirebase.com/docs/user-guide/chat/import-chats Bring your conversation history from ChatGPT or Claude into Wirebase. ## Overview Imported chats appear in your history with their original dates, and you can keep chatting in them with any model. ## Step 1: Export from your old tool In ChatGPT, go to **Settings → Data controls → Export data**. OpenAI emails you a `.zip` download link. In Claude, go to **Settings → Privacy → Export data**. Anthropic emails you a `.zip` download link. ## Step 2: Import into Wirebase Go to **Settings → Import chats**, or click **Coming from ChatGPT or Claude? Import your chats** in the sidebar. Drop the `.zip` file you were emailed, or the `conversations.json` inside it. Files up to 200 MB are supported. Keep the page open until the import completes. You'll see how many conversations were imported. ## What gets imported * Conversation text is imported, with original titles and dates. * Images, files, tool output and account data are **not** imported. * Your export is read in your browser. Only conversation text is uploaded. * Importing the same export again skips chats that were already imported. * Conversations with no messages are left out. # Keyboard shortcuts Source: https://www.wirebase.com/docs/user-guide/chat/keyboard-shortcuts Every keyboard shortcut in Wirebase. Press ⌘ / (Ctrl /) in the app to see this list at any time. On Windows and Linux, use Ctrl wherever you see ⌘. | Action | Shortcut | | - | - | | New chat | ⌘ Shift O | | Search chats | ⌘ K | | Toggle temporary chat | ⌘ Shift E | | Toggle sidebar | ⌘ Shift S | | Change tool mode | ⌘ P | | Copy last message | ⌘ Shift C | | Open chat preferences | ⌘ Shift L | | Delete current chat | Shift Backspace | | Show shortcuts | ⌘ / | # Memory Source: https://www.wirebase.com/docs/user-guide/chat/memory Let Wirebase remember lasting facts about you and use them in future chats. ## Overview With memory on, Wirebase can remember facts about you, like your name, role and preferences, and use them in later chats so you don't repeat yourself. * Memories are private to you, and separate for each organization you belong to. * Memory never stores secrets. Anything that looks like a password, key or card number is refused. * You can keep up to 100 memories per organization. ## How memories are saved * **In chat**: ask the assistant to remember something ("Remember that I prefer TypeScript examples"). It can also save a fact you mention when it's clearly useful later. You'll see **Memory updated** in the conversation, with **Undo**. * **By hand**: go to **Settings → Memory** and use **Add a memory**. To make the assistant forget something, ask it in chat, or delete the memory in settings. ## Manage memories Open **Settings → Memory** to: * See every saved memory and where it came from (**Saved in chat** or **Added by you**). * Delete individual memories, or **Clear all**. * Turn **Use and save memories** off. When off, chats neither read nor save memories; existing memories are kept until you delete them. Memory is available once you are part of an organization. # Mentions Source: https://www.wirebase.com/docs/user-guide/chat/mentions Type @ to call an agent, workflow, connector, tool or knowledge base by name. ## Overview Type `@` in the message box to open the mention menu. Mentions tell the assistant exactly what to use for this message, instead of leaving it to decide. | Mention | What happens | | - | - | | **Agents** | The agent answers, with its own instructions and tools. | | **Workflows** | The assistant runs the published workflow as a tool, filling in its inputs from your message. | | **Connector Tools** | The assistant uses a connector (all its tools) or one specific tool. | | **Knowledge Bases** | The assistant searches that knowledge base and cites what it finds. | | Built-in tools | Web search, charts, code execution and the other built-in tools. | ## How to use mentions 1. Type `@` and start typing a name to filter the list. 2. Use the arrow keys and Enter, or click, to pick an item. It appears as a chip in your message. 3. Add as many mentions as you need, then write your request and send. ```text Example theme={null} @research-agent compare the pricing pages of our top three competitors and put the results in a table ``` Mentions work in every tool mode, even **None**. Only the mentioned tools are sent to the model, which keeps answers focused and saves tokens. ## What you can mention You can mention anything you own plus anything shared with your organization: org agents and workflows set to **Public** or **Read Only**, connectors shared with the organization, and shared knowledge bases. Workflows must be **published** to appear. # Models Source: https://www.wirebase.com/docs/user-guide/chat/models Choose which AI model answers, and understand where models come from. ## Choose a model Click the model name next to the message box to open the model selector. Models are grouped by provider. The model you pick answers your next message, and you can switch at any point in a conversation. Earlier messages stay in the thread, so the new model sees the full history. The list only shows models your organization has enabled. If a model you need is missing, ask an admin to add the provider's key or enable the model in **Admin → Models**. ## Where models come from Wirebase connects to AI providers with API keys your organization adds. Each key unlocks that provider's models, and admins choose which of them members can use. See [Add an API key](/docs/user-guide/get-started/add-api-key). Supported providers include OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, DeepSeek, Fireworks, Cohere, Amazon Bedrock, OpenRouter, Ollama and any OpenAI-compatible endpoint. ## Model capabilities Models differ in what they can do. In **Admin → Models**, each model lists its capabilities, such as: * **Tool calling**: can use tools, agents, workflows and connectors. * **Vision / Image input**: can read images you attach. * **File upload**: can read documents such as PDFs directly. * **Reasoning**: thinks step by step before answering. Reasoning appears in a collapsible section above the answer. * **Web search**: has the provider's own web search. If you attach a file a model can't read, or mention a tool with a model that can't call tools, pick a different model. ## Compare models Not sure which model is best for a prompt? Send it to two or three models at once. See [Compare models](/docs/user-guide/chat/compare). ## Usage limits Your organization may set monthly AI budgets. If you're close to your limit, you'll see a notice above the message box. When a hard limit is reached you can't send new messages until the next billing period. Contact your admin if you need more. See [Usage limits](/docs/admin-guide/usage/usage-limits). # Chat basics Source: https://www.wirebase.com/docs/user-guide/chat/overview Start conversations, switch models, organize threads and find past chats. ## Start a chat Click **New Chat** in the sidebar, or press ⌘ Shift O (Ctrl Shift O on Windows). Type your message in the box at the bottom and press Enter. The message box understands more than plain text: | Do this | To | | - | - | | Type `@` | Call an agent, workflow, connector, tool or knowledge base. See [Mentions](/docs/user-guide/chat/mentions). | | Click the model name | Switch models for the next message. See [Models](/docs/user-guide/chat/models). | | Click **+** | Attach files or generate an image. See [Files & images](/docs/user-guide/chat/files-and-images). | | Click the microphone | Dictate your message instead of typing it. | | Open the tool menu | Choose which tools the assistant may use, and how. See [Tools](/docs/user-guide/chat/tools). | ## Work with a thread Every conversation is saved as a thread in the sidebar, grouped by **Today**, **Yesterday**, **Last 7 days** and **Older**. Open a thread's menu to: * **Rename** it. * **Share & Export**: create a public link, or download as Markdown or JSON, or print. See [Share & export](/docs/user-guide/chat/share-and-export). * **Move to project**, so it picks up the project's instructions, files and knowledge. See [Projects](/docs/user-guide/chat/projects). * **Archive Chat** to hide it from the sidebar without deleting it. Archived chats live in **Settings → Archived Chats**. * **Delete Chat**. You can also edit a message you sent and resend it, or copy the assistant's last answer with ⌘ Shift C. ## Search your chats Press ⌘ K (Ctrl K) to search your chat history. Results are grouped by date, and you can preview a conversation before opening it. ## Answers you can check When the assistant uses a knowledge base, the answer includes numbered citations. Click one to see the passage it came from. When it uses tools such as web search or a connector, each call appears in the conversation. Choose how much of a tool call is shown in **Settings → Chat display**. ## Personalize responses Open **Settings → User Instructions** (⌘ Shift L) to tell the assistant what to call you, what you do, and how you like answers written. These apply to every chat. See [Settings](/docs/user-guide/account/settings). ## Related Every shortcut in one list. Chats that aren't saved. # Projects Source: https://www.wirebase.com/docs/user-guide/chat/projects Group related chats with shared instructions, files and knowledge. ## Overview A project is a home for related work, such as a product launch, a client or a research topic. Every chat inside a project automatically uses the project's: * **Instructions**: how the assistant should respond in every chat in the project. * **Files**: documents, PDFs or notes any chat in the project can search. * **Knowledge bases**: knowledge bases from your workspace, attached to the project. ## Create a project In the sidebar under **Projects**, click **New project**. Give it a name, an optional description and a color. Describe the tone, context and formats the assistant should follow. These are added to every chat in the project. Drag files into **Files** (up to 100 MB each), and use **Add knowledge** to attach existing knowledge bases. Type in the project's message box to start a chat inside it. Project files are searched with embeddings, which need an OpenAI or OpenRouter key in your organization. Until an admin adds one, files can't be searched. ## Move chats in and out From any chat's menu, choose **Move to project** and pick a project. The chat starts using the project's instructions, files and knowledge from the next message. Choose **Remove from project** to take it out again. ## Share a project Click **Share** on the project page and choose who has access: | Who has access | Meaning | | - | - | | **Only you** | Private to you. | | **Everyone in your organization** | Every member can open the project. | | **A team** | Members of one [team](/docs/admin-guide/people/teams). | Then choose what they can do: **Chat in the project**, or **Chat and edit instructions, files and knowledge**. Sharing a project shares its setup, not your conversations. Everyone's chats in a shared project stay private to them. ## Delete a project Deleting a project keeps its chats and moves them out of the project. Attached knowledge bases are not deleted. # Share & export Source: https://www.wirebase.com/docs/user-guide/chat/share-and-export Share a conversation with a public link, or download it as Markdown, JSON or PDF. ## Share a public link In the chat's menu, choose **Share & Export**. Click **Create Link**. The link is copied to your clipboard. Anyone with the link can view the conversation as it was when you shared it. Messages you add afterwards stay private. Shared links are public. Don't share conversations containing confidential information. ### Revoke a link Go to **Settings → Shared Links** to see every link you've created. Click **Unshare** to revoke one. People with the link can no longer view the conversation. ## Download a conversation From **Share & Export**, under **Download**: | Format | Best for | | - | - | | **Markdown** | Pasting into docs, wikis or notes. | | **JSON** | Keeping a structured, machine-readable copy. | | **Print** | Printing or saving as PDF from your browser's print dialog. | # Temporary chat Source: https://www.wirebase.com/docs/user-guide/chat/temporary-chat Ask something without saving the conversation. ## Overview A temporary chat is a scratch conversation that isn't saved to your history. Use it for quick, throwaway questions. ## Start a temporary chat Press ⌘ Shift E (Ctrl Shift E) or click the temporary chat toggle in the header. The panel says **This chat won't be saved.** * Click **Reset Chat** to clear it and start over. * Open **Temporary Chat Instructions** to set a system prompt for the temporary chat only. When you close the panel, the conversation is gone. Admins can turn temporary chats off in **Admin → Settings → Chat Features**. # Tools Source: https://www.wirebase.com/docs/user-guide/chat/tools Let the assistant search the web, run code, draw charts, call APIs and use your connectors. ## Tool modes The tool mode decides how freely the assistant uses tools. Switch it from the tool menu next to the message box, or press ⌘ P (Ctrl P). | Mode | Behavior | | - | - | | **Auto** | Decides when to use tools without asking you. | | **Manual** | Asks your permission before using any tool. You approve or reject each call. | | **None** | Doesn't use tools. @mentions still work. | ## Choose which tools are available Open **Tools Setup** from the tool menu to pick which tools the assistant can use. Save combinations you use often with **Save As Preset**, and switch between presets later. ## Built-in tools Searches the web and reads pages, so answers can include current information with links. Your admin must turn on **Web Search** and add a search provider key. See [Chat features](/docs/admin-guide/models/chat-features). Use **Search Options** to pass options like the number of results. Draws bar, line and pie charts, and interactive tables, directly in the conversation. Runs JavaScript or Python to calculate, transform data or plot. Python runs in your browser with Pyodide; it can fetch files over HTTP when the server allows it (CORS). Your admin can turn code execution off. Calls a public URL or REST API and reads the response. Searches a knowledge base you mention and cites passages. See [Knowledge](/docs/user-guide/workspace/knowledge). Creates and edits images. See [Files & images](/docs/user-guide/chat/files-and-images). ## Connector tools Connectors add tools from outside services, such as creating a GitHub issue or querying a Notion database. Connectors you add, and connectors shared with your organization, appear in the tool menu and the `@` menu. See [Connectors](/docs/user-guide/workspace/connectors). ## See what a tool did Each tool call appears in the conversation with its request and response. Choose the default detail level in **Settings → Chat display**: * **Compact**: header only. Click a call to see its request and response. * **Expanded**: request and response as a readable tree. * **Debug**: the raw payload, exactly what the tool sent and returned. # Add an API key Source: https://www.wirebase.com/docs/user-guide/get-started/add-api-key Connect an AI provider to your organization so members can chat with its models. ## How keys work in Wirebase Provider API keys belong to your **organization**, not to individual users. Once an admin adds a key, every member can chat with the models enabled for that provider. Usage is billed by the provider directly to the key's owner. * Keys are encrypted at rest and are never shown again after you save them. * A key is validated with the provider before it is saved. * Keys set as server environment variables are ignored. Keys are always configured in the app. Only organization **admins** can add, change or remove keys. ## Add a key Go to **Admin → API Keys** and stay on the **Provider Keys** tab. Find the provider and click **Add**. Wirebase validates the key, saves it encrypted, and switches on that provider's default models. Go to **Admin → Models** to turn individual models on or off. See [Models](/docs/admin-guide/models/models). ## Connect OpenRouter without a key Under **Quick start**, click **Connect** next to OpenRouter. You'll authorize Wirebase on OpenRouter and a key is created for your organization automatically. OpenRouter gives access to models from many providers through a single key. ## Supported providers Wirebase supports chat models from OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, DeepSeek, Fireworks, Cohere, Amazon Bedrock, OpenRouter and Ollama (local server or Ollama Cloud), plus any OpenAI-compatible endpoint you add as a [custom provider](/docs/admin-guide/models/custom-providers). Enter an AWS Access Key ID, Secret Access Key and Region (default `us-east-1`). Use the region where your Bedrock models are available. For **Ollama (Local)**, enter the base URL of your Ollama server, for example `http://localhost:11434/api`. The server must be reachable from Wirebase. Wirebase checks connectivity and lists the models it finds. For **Ollama Cloud**, paste an API key from ollama.com/settings/keys. ## Service keys The **Service Keys** tab holds keys for services that aren't chat models, such as the web search provider. See [Chat features](/docs/admin-guide/models/chat-features). ## Troubleshooting Your organization has no provider keys yet, or every model is switched off. Ask an admin to add a key or enable models. The provider rejected the key. Check that you copied it completely, that it hasn't been revoked, and that billing is set up on the provider's side. # Best practices Source: https://www.wirebase.com/docs/user-guide/get-started/best-practices Practical ways to get better, more consistent answers from Wirebase. ## Tell the assistant about yourself once Fill in **Settings → User Instructions**: what to call you, what you do, and how you like answers. These preferences apply to every chat, so you don't repeat them. See [Settings](/docs/user-guide/account/settings). ## Pick the right model for the task Fast, inexpensive models are fine for drafting and quick questions. Use a reasoning model for complex analysis, math or code. When you're unsure, [compare](/docs/user-guide/chat/compare) two or three models on the same prompt and keep the best answer. ## Use projects for ongoing work If you keep giving the same background in new chats, create a [project](/docs/user-guide/chat/projects). Its instructions, files and knowledge bases apply to every chat inside it. ## Turn repeated prompts into tools * Prompts you reuse word-for-word → save them as [prompts](/docs/user-guide/workspace/prompts) with `{{variables}}`. * A persona or role with specific tools → build an [agent](/docs/user-guide/workspace/agents). * A fixed multi-step process → build a [workflow](/docs/user-guide/workspace/workflows). * Something that should happen on a schedule → create an [automation](/docs/user-guide/workspace/automations). ## Ground answers in your documents For questions about your own material, add it to a [knowledge base](/docs/user-guide/workspace/knowledge) and mention it with `@`. Answers then cite the passages they came from, so you can check them. ## Keep tool use under control Use **@mentions** to tell the assistant exactly which agent, workflow or connector to use. If you want to approve each tool call before it runs, switch the tool mode to **Manual**. See [Tools](/docs/user-guide/chat/tools). ## Write clear instructions When writing instructions for an agent or project: * Say who the audience is and what a good answer looks like. * Give formats explicitly ("answer in a table with columns X, Y, Z"). * Say what to do when information is missing ("ask a clarifying question" or "say you don't know"). * Keep them short. Long instructions are harder for models to follow consistently. # Welcome to Wirebase Source: https://www.wirebase.com/docs/user-guide/get-started/introduction One AI workspace for your whole team: every major model, your own API keys, and the agents, workflows and connectors that put them to work.
Wirebase Wirebase
## Overview Wirebase is an AI chat platform for teams. Your organization brings its own API keys for OpenAI, Anthropic, Google, xAI, Groq, OpenRouter and others, and everyone chats with whichever enabled model fits the job, from one interface. On top of chat, you can build **agents** with their own instructions and tools, chain steps into **workflows**, connect outside services through **connectors** (MCP), and let the assistant search your documents through **knowledge bases**. Admins control which models are on, set usage limits and review audit logs. ## Why Wirebase? Switch between models from different providers mid-conversation, or compare two or three answers side by side. Provider keys belong to your organization and are encrypted at rest. You pay providers directly, at their prices. Build reusable agents and visual multi-step workflows, then call them in chat with an @mention. Connect remote MCP servers like GitHub, Notion, Linear or Stripe so the assistant can act on your tools. Upload documents or add web pages, and get answers grounded in them, with citations. Roles, teams, SSO, SCIM, usage limits, audit logs and custom domains. ## Next steps Create your organization, add a provider key and send your first message in a few minutes. Threads, models, tools, files and everything else in the chat window. Give an assistant a role, instructions and tools, and share it with your team. Manage keys, models, members, limits and security for your organization. Put an agent on your website as a chat bubble. # Quickstart Source: https://www.wirebase.com/docs/user-guide/get-started/quickstart Create your organization, connect an AI provider and send your first message. ## Overview The first time you sign in, Wirebase walks you through a short setup. Here is what it covers: Sign up with your email address and a password, or continue with Google. Connect at least one AI provider so your organization has models to chat with. Optional: invite teammates by email. You can do this at any time. Pick a model and start chatting. **Estimated time:** about 5 minutes, most of it spent getting an API key from your provider. ## Step 1: Create your account Go to [wirebase.com](https://www.wirebase.com) and click **Start free**. When you sign up, Wirebase creates an organization for you. Your organization owns everything you set up: API keys, enabled models, agents, workflows, connectors and knowledge bases. If a teammate has already invited you, open the link in your invitation email instead. You'll join their organization rather than creating a new one. ## Step 2: Add an API key Wirebase does not resell model access. Your organization adds its own provider keys, and every member chats with the models those keys unlock. In the setup dialog, under **Add your first API key**: 1. Choose a provider: **OpenAI**, **Anthropic**, **Google**, **Groq** or **OpenRouter**. 2. Paste your key. Wirebase checks it with the provider before saving it. 3. The provider's default models are switched on for your organization. Don't have a key? Click **Get a free one**. Some providers, such as Google and Groq, offer free tiers. With OpenRouter you can click **Connect** and authorize Wirebase to create a key for you, without copying and pasting. You can add more providers later in **Admin → API Keys**. See [Add an API key](/docs/user-guide/get-started/add-api-key). ## Step 3: Invite your team Enter teammates' email addresses separated by commas. Each person gets an email invitation and joins as a **member**. Admins can change roles later in **Admin → Users**. See [Members and roles](/docs/admin-guide/people/members-and-roles). ## Step 4: Send your first message Pick a starter prompt, or open a new chat and type anything. Use the model selector next to the message box to switch models at any time. You're set up. From here, try typing `@` in the message box to see the agents, workflows and tools you can call. ## Coming from ChatGPT or Claude? You can bring your conversation history with you. See [Import chats](/docs/user-guide/chat/import-chats). ## Next steps Everything you can do in a conversation. Get better answers, faster. # Agents Source: https://www.wirebase.com/docs/user-guide/workspace/agents Purpose-built assistants with their own instructions and tools, invoked with an @mention. ## Overview An agent is an assistant set up for one job: a research analyst, a support writer, a SQL helper. It has its own instructions and a fixed set of tools, so it behaves the same way every time anyone uses it. Find agents under **Workspace → Agents**. **My Agents** lists yours; **Org Agents** lists agents teammates have shared. ## Create an agent Click **Create Agent**. You can fill in the form yourself, or click **Generate with AI**, describe what you want in a sentence or two, and let Wirebase draft it for you. | Field | What to enter | | - | - | | **Name and icon** | A short name. This is what people type after `@`. | | **Description** | What the agent does. Shown in lists and the mention menu. | | **This agent is an expert in…** | The agent's role, for example "stock analysis". | | **Instructions** | The agent's role, personality, guidelines and knowledge. This is its system prompt. | | **Tools** | Tools the agent can use: built-in tools, connector tools, workflows and knowledge bases. | Click **Save**. Write instructions the way you'd brief a new colleague: who the audience is, what a good answer looks like, and what to do when information is missing. See [Best practices](/docs/user-guide/get-started/best-practices). ## Use an agent In any chat, type `@` and pick the agent, then write your request. The agent answers with its own instructions and tools. You can bookmark agents you use often from the agents list. ## Share an agent Set **Visibility** on the agent: | Visibility | Who can use it | Who can edit it | | - | - | - | | **Private** | Only you | Only you | | **Read Only** | Everyone in your organization | Only you | | **Public** | Everyone in your organization | Everyone in your organization | Shared agents appear under **Org Agents** for your teammates. ## Put an agent on your website On the Team plan and above, any saved agent can be published to your website as a chat bubble. Visitors chat without signing in, and the agent answers from its knowledge bases. See [Website widget](/docs/developer-guide/website-widget). # Automations Source: https://www.wirebase.com/docs/user-guide/workspace/automations Run a prompt or workflow on a schedule and collect the results. ## Overview An automation runs a prompt or a workflow on a schedule, for example a daily news digest, a weekly metrics summary or an hourly check on a connector. Every run's result is saved so you can read it later or continue it in chat. Find automations under **Workspace → Automations**. ## Create an automation Click **New Automation**, then set: For example, "Daily news digest". Choose **Once**, **Hourly**, **Daily**, **Weekly** or **Monthly**, then the time (or minute, for hourly), the day where relevant, and your **Timezone**. Choose what runs: * **Prompt**: write what the AI should do on each run, and pick a **Model**. Optionally pick an **Agent** to run it with that agent's instructions and tools. * **Workflow**: pick a published workflow and fill in its **Workflow inputs** (or **Edit as JSON**). Click **Run Now** to try it immediately. A run can take a few minutes. ## Results Open an automation and go to **Results** to see each run, with the tools it called and any structured output. From a result you can **Copy** it, **Run again**, or **Continue in chat**. **Activity** shows every run across all your automations. ## Pause, resume and failures * **Pause** stops an automation without deleting it. **Resume** puts it back on schedule. * An automation that fails three times in a row is turned off. Resuming it clears the failure count. * If the agent or workflow it runs is deleted, choose a new target to start it again. * A **Once** automation finishes after its run. Pick a new date to schedule it again. ## Plan limits | Plan | Automations | Most frequent schedule | | - | - | - | | Free | 2 | Daily | | Team | 20 | Hourly | | Enterprise | 100 | Hourly | # Connectors Source: https://www.wirebase.com/docs/user-guide/workspace/connectors Connect remote MCP servers so the assistant can read from and act on your tools. ## Overview Connectors link Wirebase to outside services using the [Model Context Protocol](https://modelcontextprotocol.io) (MCP). Each connector adds tools, such as "create a GitHub issue" or "search Notion", that the assistant can use in chat, in agents and in workflows. Find connectors under **Workspace → Connectors**. Adding connectors requires the **editor** or **admin** role. Every member can use connectors shared with them. Admins can turn connectors off for the organization. ## Add a connector from the library The **Discover** tab lists popular remote MCP servers, including GitHub, Notion, Linear, Atlassian, Asana, Stripe, PayPal, Canva and Neon. Click **Connect** on one. If the service uses OAuth, you'll be sent to sign in and approve access, then returned to Wirebase. ## Add a custom connector Click **Add custom connector** and fill in: | Field | Description | | - | - | | **Name** | Letters, numbers and hyphens only. It becomes the tool prefix in chat, so a `github` connector's tools are named like `github_create_issue`. | | **Server URL** | The remote MCP endpoint (Streamable HTTP or SSE), for example `https://mcp.example.com/mcp`. | | **Headers** | Optional. Add an `Authorization` header if the server requires a token. | You can switch to **Edit as JSON** to paste a configuration with a `url` and optional `headers`. Wirebase connects to **remote** MCP servers over HTTPS. Local servers that run as a command on your computer (stdio) aren't supported. To build your own server, see [Build an MCP connector](/docs/developer-guide/mcp-connectors). ## Authorize a connector If a connector shows **Authorization needed**, click **Authorize** and complete the service's sign-in. By default, each user authorizes separately and the connector acts with their own account. Turn on **Shared Authentication** to have everyone share one OAuth session. With shared authentication on, anyone who uses the connector acts with the account that authorized it. Only enable it if you trust every user with that access. ## Test tools Open a connector to see its tools. Select a tool to view its input schema, then use **Tools Test** to call it with sample input. **Create Input with AI** drafts example input from a short prompt. ## Customize how tools are used Add **custom instructions** to guide the model: * **Connector Customization** applies whenever the connector is available. Example: "Always search the Engineering space first." * **Tool Customization Instructions** apply to one tool. Example: "Always enter the email in the format [example@example.com](mailto:example@example.com)." You can also set your own instructions for connectors in **Settings → Connector Instructions**. ## Share a connector | Visibility | Meaning | | - | - | | **Private** | Only you can use it. | | **Read-only** | Others can view it but not modify it. | | **Organization** | Available to all members of your organization. | ## Use connectors in chat Pick connector tools in the tool menu, or `@mention` a connector or a single tool. See [Mentions](/docs/user-guide/chat/mentions). # Explore Source: https://www.wirebase.com/docs/user-guide/workspace/explore Discover community-built agents, workflows and prompt templates. ## Overview [Explore](https://www.wirebase.com/explore) is a public gallery of agents, workflows and prompt templates. You can browse it without an account. ## Browse * Switch between **Agents**, **Workflows** and **Prompts**. * Search by keyword, filter by category, and sort by **Popular** or **Newest**. * Open an item to see what it does and how it's set up, plus related items. ## Use an item Each item has a button to try it in your own workspace: | Button | What happens | | - | - | | **Use this Agent** | Opens a new chat with the agent already mentioned. | | **Use this Workflow** | Opens a new chat with the workflow already mentioned. | | **Use this Prompt** | Opens a new chat with the prompt text in the message box, ready to edit. | If you're not signed in, you'll be asked to sign in first, then taken straight to the chat. # Knowledge Source: https://www.wirebase.com/docs/user-guide/workspace/knowledge Upload documents or web pages, then get answers grounded in them, with citations. ## Overview A knowledge base is a searchable collection of your documents. When you mention one in chat, or attach it to an agent or project, the assistant searches it for relevant passages and cites them in its answer. Find knowledge bases under **Workspace → Knowledge**. Knowledge bases use embeddings, which need an **OpenAI** or **OpenRouter** key in your organization. Ask an admin to add one if you can't create a knowledge base. ## Create a knowledge base Click **New Knowledge Base** and give it a name and description. Click **Upload Documents**, or **Add from URL** to fetch a public web page. Each document moves from **Pending** to **Processing** to **Ready**. Once ready, it's searchable. Use **Test Search** to check what comes back for a question, with a match percentage for each passage. ### Supported files PDF, DOCX, XLSX, PPTX, TXT, Markdown, HTML, CSV, JSON and XML, up to 100 MB per file. ## Use a knowledge base * **In chat**: type `@` and pick the knowledge base under **Knowledge Bases**. * **In an agent**: add it in the agent's tools. The agent searches it whenever relevant. * **In a project**: attach it with **Add knowledge**. Every chat in the project can search it. Answers include numbered citations. Click one to see the source passage. ## Manage documents From a document's menu you can **Preview**, **Download**, **Reprocess** (re-index it) or **Delete** it. ## Share a knowledge base Set it to **Private** (only you) or **Shared** (all members of your organization can use it). # Prompts Source: https://www.wirebase.com/docs/user-guide/workspace/prompts Save reusable prompt templates and share them with your organization. ## Overview Save prompts you use often, like "Write a blog post about…" or "Review this contract for…", as templates in one place, and share the good ones with your team. Find prompts under **Workspace → Prompts**. **My Prompts** lists yours; **Org Prompts** lists prompts teammates have shared. ## Create a prompt Click **New Prompt** and fill in: | Field | Description | | - | - | | **Title** | For example, "Blog Post Writer". | | **Description** | What the prompt does. Shown in the prompt list. | | **Prompt** | The template text. | ### Variables Mark the parts you change each time with double curly braces, like `{{topic}}` or `{{audience}}`. The editor lists the variables it detects under the prompt, so it's clear what to fill in before you send it. ```text Example theme={null} Write a 600-word blog post about {{topic}} for {{audience}}. Use a friendly, practical tone and end with three key takeaways. ``` ## Share a prompt Turn on **Share with organization** to make the prompt available to every member under **Org Prompts**. ## Find more prompts Browse community prompt templates on [Explore](/docs/user-guide/workspace/explore). Click **Use this Prompt** on any of them to open a new chat with the prompt already in the message box. # Workflows Source: https://www.wirebase.com/docs/user-guide/workspace/workflows Build multi-step AI pipelines visually, then call them from chat with an @mention. ## Overview A workflow is a fixed sequence of steps, drawn as connected nodes on a canvas. Data flows from an **Input** node, through steps like LLM calls, conditions, HTTP requests and connector tools, to an **Output** node. Once published, a workflow becomes a tool: mention it in chat with `@` and the assistant fills in its inputs from your message and runs it. Find workflows under **Workspace → Workflows**. Creating and editing workflows requires the **editor** or **admin** role. Every member can run workflows shared with them. ## Build a workflow Click **Create Workflow**. Give it a name and description. The assistant sees the name as the tool name and the description as the tool description, so make both clear. Select the **Input** node and add input fields. These are the parameters the assistant will provide when it calls the workflow. Use **Add Next Node** to add nodes. Inside a node, type `/` to reference the output of an earlier node. Connect the last steps to the **Output** node and choose which values to return. Click **Run**, enter test values for the inputs, and check each node's result. Click **Publish**. Published workflows are available in chat but can't be edited. Switch back to **Draft** to edit; while in draft, the workflow is unavailable in chat. The editor auto-saves every 10 seconds. Use **Auto Layout** to tidy the canvas. ## Node types | Node | What it does | | - | - | | **Input** | Defines the parameters the workflow accepts. | | **LLM** | Generates text with a model. Turn on **Structured Output** to return JSON matching a schema. | | **Condition** | Branches with if / else-if / else logic on earlier values. | | **Tool** | Runs a connector tool. You write instructions; an LLM fills in the tool's parameters. | | **HTTP** | Calls an external API or webhook. | | **Template** | Builds text, such as an email or report, from earlier values. | | **Code** | Transforms data with sandboxed JavaScript. | | **Note** | A comment on the canvas. Doesn't run. | | **Output** | Collects and returns the final result. | See the [node reference](/docs/developer-guide/workflow-nodes) for every option. ## Share a workflow Set visibility to **Private** (only you), **Read Only** (everyone can use it, only you edit) or **Public** (everyone can use and edit). Shared workflows appear under **Org Workflows**. ## Run on a schedule To run a workflow automatically, create an [automation](/docs/user-guide/workspace/automations) that targets it.