MCP Quick Start
Connect an AI agent to your workspace in a few minutes.
What is the MCP server?
The JotReview MCP server exposes your workspace to AI agents as a set of tools they can call. Any client that speaks the Model Context Protocol — Claude Desktop, Claude Code, Cursor, or your own app — can connect and work with your feedback in natural language.
Agents can:
- •Triage feedback — read requests, set status, tag, approve, or reject
- •Manage the roadmap — move items between columns
- •Write changelogs — draft, edit, and publish entries
- •Post comments and manage boards and tags
The server is remote (hosted by JotReview) and speaks Streamable HTTP, so there is nothing to install or run yourself. You only need an access token.
Create an access token
Tokens are created from your dashboard. MCP access is available on the Pro and Business plans.
Open Settings > MCP Server, click Create token, give it a name (for example Claude Desktop), and choose an access level. Copy the token immediately — for security it is shown only once and stored hashed. If you lose it, revoke it and create a new one.
Two access levels are available:
- •Read only — the agent can list and read, but never change anything
- •Read & write — full access to create, update, moderate, and delete
Every token is scoped to a single workspace and starts with jr_mcp_.
Connect your agent
Add the server to your agent's MCP configuration. Because it uses bearer-token auth over Streamable HTTP, connect through mcp-remote, which bridges local MCP clients to remote servers.
For Claude Desktop, edit claude_desktop_config.json (Settings > Developer > Edit Config):
Cursor uses the same shape in .cursor/mcp.json. After saving, restart the client so it picks up the new server.
Make your first call
Once connected, your agent has the full JotReview tool set. Try a prompt like:
- •"What can you do in my JotReview workspace?" — the agent calls
get_workspace_infoand lists its capabilities - •"Show me the top-voted pending requests."
- •"Move that request to In Progress and tag it as a bug."
If a call is refused, the agent reports a clear error — usually a missing scope (read-only token) or insufficient role. See Authentication & Scopes for details.