Skip to main content

Integrations

The Integrations group is where EnClaw connects to the outside world: chat channels, SaaS tools, MCP servers and LLM providers. Every item in this group requires the org_admin role or above — below that, the whole group disappears from the sidebar (sidebar items are role-filtered, and a group with no visible children is dropped entirely).

Org Integrations​

/integrations holds channel and tool integrations shared across the org — bot tokens, OAuth apps and webhook targets. Per-user OAuth lives under My Connectors instead.

What you see:

  • One card per integration with a type badge, name, description and an Active/Inactive status badge, plus Test, Edit and Delete buttons.
  • Per-type summary fields on each card — e.g. Slack shows workspace, app ID and allowed channels; Teams shows app ID, tenant and allowed teams; Telegram shows the bot and allowed groups; JIRA shows the instance and email.
  • Marketplace catalogue sources are listed inline with the same card chrome, badged Default (preferred in the marketplace picker) and Instance (available to all tenants) where applicable.

Key actions:

  • + Add integration opens a single modal for every type. The Core list covers Slack, Microsoft Teams, Telegram, JIRA, Productive.io, Reddit, Google AI, Google (Gmail / Drive OAuth), HubSpot (CRM), Lusha (B2B contact enrichment) and GitHub. The Marketplace list adds catalogue sources, custom REST connectors and any connectors published by a marketplace source.
  • Each channel integration can nominate a Default Agent to handle incoming messages; leave it empty and the Orchestrator auto-routes to the best available agent.
  • Test runs a live connection check. For types with granular permissions it returns per-action permission probes (ok / warning / error with detail).
  • When editing, secret and token fields can be left blank to keep the existing value — masked values are never written back.

See the full setup walkthroughs in Integrations.

My Connectors​

/my-connectors manages personal credentials that agents use to act on your behalf — your inbox, your JIRA tickets, your repositories. It sits in the admin group in the sidebar, and each user manages only their own connectors.

  • + Add Connector supports Outlook / Microsoft 365, Slack (identity linking for reliable message routing), JIRA (personal API token), Gmail and GitHub.
  • Each connector card shows a type badge, Active/Inactive status and a token-health badge (Connected, Session Expired, Disconnected), with Test, Reconnect / Authorise and Disconnect / Remove actions.
  • Outlook connectors expose a per-action permission list (read email, read/write calendar, read/write To-Do, OneNote, OneDrive, send email) colour-coded by risk level; Manage Permissions edits the allowed set. Gmail connectors show their allowed actions the same way.
  • Cards show the linked account (Microsoft email, Google email, Slack user and workspace ID, GitHub username) and when the connector was last used.

MCP Servers​

/mcp-servers brings third-party tools into EnClaw via the Model Context Protocol. Each registered server's tools become available to agents, namespaced as mcp__<slug>__<tool>, subject to RBAC scope.

  • Add MCP server captures name, description, URL, transport (streamable HTTP or SSE), auth (none, bearer token or custom headers) and scope — personal, team or org (new servers default to org so their tools reach every agent). An optional allowed-tools list restricts which of the server's tools are exposed.
  • Each server card shows a sync-status badge (pending / ok / error), scope badge, slug, transport, auth type, tool count and last-sync time, with any last sync error surfaced inline.
  • Test checks connectivity and reports the server name, version and protocol version. Sync pulls the server's tool list. Tools expands an inline table of tool name, namespaced name and description.
  • Edit and Delete round out the card actions; deleting a server also drops all of its discovered tools.

LLM Providers​

/llm-providers configures the AI model providers your agents run on.

  • Add Provider supports Anthropic, OpenAI, Azure OpenAI, Azure AI Foundry, Google AI, DeepSeek, Ollama (local), LM Studio (local), AWS Bedrock and Custom / OpenAI-compatible endpoints.
  • The form adapts per type: API key and base URL for most providers, deployment name and API version for Azure OpenAI, region and model ID for Bedrock. Local providers (Ollama, LM Studio) skip the API key.
  • The Default Model / Deployment Name picker combines a static default catalogue with a Refresh button that fetches the live model list from the provider using your key.
  • A provider can be marked as the org default, and a platform operator can additionally publish a provider as an instance-wide default available to all tenants.
  • Editing keeps the saved API key unless you enter a new one; Test verifies connectivity and model health, including a sample prompt/response.

See LLM Providers for provider-specific setup detail.