Most teams already use an AI assistant beside their tracker. The friction is always the same: the assistant cannot see your real tasks, permissions, or project context without brittle copy-paste or custom glue code.
Gealo now ships a per-tenant MCP server so tools like Claude Desktop, Cursor, and VS Code can read (and optionally write) your workspace through the same permission checks as the web app.
What MCP gives you here
- Remote server URL:
https://<your-workspace>.gealo.app/mcp— one endpoint per tenant. - OAuth + consent: agents sign in through the browser; owners can revoke access anytime.
- 66+ tools: tasks, sprints, releases, comments, agent-readable notifications, and more — all scoped to the connected workspace.
- Governance: master switch, client allow/block lists, and optional
mcp:writescope behind tenant policy.
Read vs write
By default agents get read-only access (mcp:write is off). When an owner enables write access, agents must re-authorize to pick up the new scope — the same pattern as escalating API permissions elsewhere.
Where to configure it
Workspace admins open Tenant → AI Agents for the MCP URL, connected clients, consents, and monthly tool-call usage against plan limits. The public overview lives at /mcp-server; the full tool directory lists every name, parameter, and scope.
When MCP beats copy-paste
If you want an agent to list open tasks, summarize a sprint, or draft a release note from real data, MCP removes the manual export step. If you only need a one-off answer, chat inside Gealo may still be faster.
For the bigger product picture, see the features overview and the July changelog. Comparing stacks? The Gealo vs alternatives pages are honest about when a point tool still wins.