Agents & MCP
Connect an agent
Point Claude Code, Cursor, or any MCP client at your workspace.
Feedgent ships an MCP server so your coding agent can read feedback, triage duplicates, update the roadmap, and publish changelogs — the same loop your team runs, over the same API.
What you need
- An API key — create one in Dashboard → Agents (details)
- Your workspace slug (the
acmeinacme.yourdomain)
Claude Code
claude mcp add feedgent \
--env FEEDGENT_URL=https://api.yourdomain.com \
--env FEEDGENT_API_KEY=your-key \
--env FEEDGENT_WORKSPACE=acme \
-- npx -y @feedgent/mcpOr in .mcp.json:
{
"mcpServers": {
"feedgent": {
"command": "npx",
"args": ["-y", "@feedgent/mcp"],
"env": {
"FEEDGENT_URL": "https://api.yourdomain.com",
"FEEDGENT_API_KEY": "your-key",
"FEEDGENT_WORKSPACE": "acme"
}
}
}
}Cursor and other MCP clients use the same command/env shape in their own config files.
Environment variables
| Variable | Required | What it is |
|---|---|---|
FEEDGENT_URL | no (defaults to local dev) | Base URL of the API |
FEEDGENT_API_KEY | yes | Key from the dashboard Agents page |
FEEDGENT_WORKSPACE | yes | Workspace slug the agent operates on |
No MCP client?
Everything the MCP server does is plain REST underneath — the OpenAPI spec
at /openapi.json is the source of truth, and @feedgent/sdk is the typed
client generated from it.