> ## Documentation Index > Fetch the complete documentation index at: https://basedash.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # MCP connectors > Connect external MCP servers to Basedash so AI can use their tools MCP connectors let you connect external [Model Context Protocol (MCP)](https://modelcontextprotocol.io/) servers to Basedash. Once connected, Basedash can sync the tools exposed by that server and make them available to AI in your workspace. > [!NOTE] > > This page is about connecting external MCP servers to Basedash. If you want to > connect an MCP client to Basedash itself, see [MCP > server](/docs/features/mcp-server). ## What MCP connectors are for MCP connectors are useful when you want Basedash to use tools outside of your database, such as internal APIs, support systems, product tools, or other services that already expose an MCP server. This lets you extend Basedash beyond SQL queries. For example, you can connect a remote MCP server so AI can look up information, take actions, or combine tool output with analysis from your existing data sources. ## How it works When you add an MCP connector, Basedash: 1. Connects to the remote MCP server over streamable HTTP or SSE 2. Syncs the tools exposed by that server 3. Uses OAuth if the server requires authentication 4. Lets you manage who can use the connector and how each tool can be used After setup, the connector appears alongside your other data sources in the command menu. ![A connected Linear MCP connector with its synced tools, where read tools like list issues are always allowed, creating and updating issues needs approval, and deleting issues is blocked](https://mintcdn.com/basedash/aH3kb-Bl5OLwdg2n/images/features/mcp-connectors/hero.webp?fit=max\&auto=format\&n=aH3kb-Bl5OLwdg2n\&q=85\&s=ae8c6f6ed9dd814c7691d493bed048ee) ## Add an MCP connector To add an MCP connector: 1. Open the command menu 2. Go to **Data sources** 3. Choose **Add MCP server** 4. Enter a name and the remote MCP server URL 5. Optionally add any required headers 6. Save the connector and complete OAuth if prompted Basedash will then sync the available tools from that server. > [!NOTE] > > Remote MCP servers over streamable HTTP or SSE are supported. ## OAuth and redirect URIs Some MCP servers require OAuth before Basedash can use their tools. In that case, Basedash starts an OAuth flow in your browser when you finish adding the connector or when you reconnect it later. If your MCP provider requires a whitelisted redirect URI, use: `https://charts.basedash.com/api/mcp-connectors/oauth/callback` If you use a self-hosted deployment, the redirect URI uses your deployment's `BASE_URL`: `/api/mcp-connectors/oauth/callback` For example: * `https://basedash.your-company.com/api/mcp-connectors/oauth/callback` * `https://analytics.acme.io/api/mcp-connectors/oauth/callback` For self-hosted deployments, make sure [`BASE_URL`](/docs/self-hosting/configure) is set correctly so OAuth callbacks can return to your instance. ## Access controls MCP connectors support access control at two levels. ### Connector access You can control who can access the connector: * Everyone in the organization * Specific groups * Specific members This is useful when a connector should only be available to certain teams or operators. ### Tool access Each synced tool can also have its own access mode: * **Always allow**: AI can use the tool without asking * **Needs approval**: AI must request approval before using the tool * **Blocked**: AI cannot use the tool This gives you fine-grained control over which MCP tools can be used automatically and which ones should require human review. ## Updating and reconnecting If the remote MCP server changes, you can sync tools again from the connector page. If the connector's authentication expires or breaks, Basedash will prompt you to reconnect it. Updating the server URL or auth configuration may require a fresh tool sync and reauthorization. ## Best practices * Give the connector a clear name so people know what system it connects to * Use connector-level access controls if the MCP server should only be available to part of your organization * Start new tools in **Needs approval** until you are comfortable allowing them automatically * Keep provider redirect URI allowlists updated if the MCP server uses OAuth ## FAQ > **Can I connect a local or private MCP server?** > > No. Basedash only connects to remote MCP servers over streamable HTTP or > SSE, at an HTTPS URL it can reach on the public internet. Servers that run > over stdio on your own machine, or only inside a private network, can't be > added. > **Who can add an MCP connector?** > > Only admins can add connectors and change who can use them. A connector with > no access rules is available to everyone in the organization. > **Does each person sign in to the MCP server with their own account?** > > No. Each connector has one connection, and every allowed person's tool calls > run through it with that account's permissions. Limit [connector > access](#connector-access) to the people who should act as that account. > **Which tools need approval by default?** > > Tools the server marks as read-only start as **Always allow**. Every other > tool starts as **Needs approval**. Syncing tools again keeps any modes you > changed. > **Can automations and Slack use connector tools?** > > In Slack, the AI can use **Always allow** and **Needs approval** tools, and > approval requests appear in the thread. [Automations](/docs/features/automations) > run without anyone watching, so they only use **Always allow** tools. > [Insights](/docs/features/insights) don't use connector tools. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.