Let your agent work in every app you use

Any app that publishes a remote MCP server over streamable HTTP or SSE. Many SaaS tools offer one, including Linear, HubSpot, Notion, GitHub, Stripe, and Intercom, and you add each by pasting its server URL. Basedash syncs the tools the server exposes, so what the agent can do depends on that server: reading issues, updating CRM records, creating pages, or whatever else it offers. Local servers that only run over stdio on your own machine can't be added.

Yes. If your team runs its own MCP server in front of internal APIs, a support system, or a product backend, add its URL the same way you would a vendor's. It has to be reachable from Basedash over streamable HTTP or SSE. Add an API key or other static headers if the server expects them, or let Basedash run the OAuth flow if the server requires it. On a self-hosted deployment, OAuth callbacks use your own base URL, so Basedash and the server can both stay inside your network.

When the agent wants a tool set to Needs approval, it pauses and asks before anything runs. In web chat you approve or deny the call in the conversation, and in Slack the request shows up in the thread with Approve and Deny buttons. If it's approved, the tool runs and the agent carries on with the result; if it's denied, the call never reaches the other app. Automations and API chats can't wait for a person, so they only use tools set to Always allow.

It depends on what the server declares. When Basedash syncs a connector, tools the server marks read-only with MCP's readOnlyHint annotation start as Always allow, and every other tool starts as Needs approval, so anything that might write waits for a person until you decide otherwise. You can change any tool's mode on the connector page, including Blocked, which keeps the agent from using it at all. Tools added by a later re-sync get the same defaults.

You choose for each connector: everyone in the organization, specific groups, or specific members. People without access don't get its tools in their chats, so a connector to a finance or HR system can stay with the team that owns it. Tool modes apply on top of that, so even people with access only reach the tools set to Always allow or Needs approval, never the blocked ones.

Static headers you add, like an API key, are encrypted before they're saved. When a server uses OAuth, Basedash runs the flow with PKCE, stores the access and refresh tokens encrypted at rest, and refreshes them automatically. If authorization expires or is revoked, Basedash asks you to reconnect the connector instead of failing quietly. Reading a connector through the public API never returns its configured headers or OAuth tokens.

Yes, on the Enterprise plan. Audit logs record every connector tool call as an event with the person it ran for, the connector and tool, whether it succeeded, and how long it took. Creating, editing, or deleting a connector and changing its tool access are logged as well. Admins can review the events in Basedash or pull them into a SIEM with an administrator API key through the audit log endpoint.