This is the FINAL PyPI release of
gen-mcp-server. GEN no longer ships a downloadable MCP: the official GEN MCP is the hosted server athttps://mcp.gen.pro. No further versions will be published to PyPI.The package still works so existing integrations have a migration window, but it will not receive new tools, fixes, or API updates — the hosted server is already ahead of it and is the only surface kept in sync with the GEN API. Installing it emits a
DeprecationWarning; the stdio CLI prints a notice to stderr on start.
No install. Point your MCP client at the hosted endpoint and authenticate with your GEN Personal Access Token (from https://gen.pro — log in, pick an agent, open the API page in the sidebar, click Create API Key):
{
"name": "gen",
"url": "https://mcp.gen.pro",
"headers": { "Authorization": "Bearer gen_pat_…" }
}(Existing configs using https://mcp.gen.pro/mcp keep working.)
Claude Code:
claude mcp add --transport http gen https://mcp.gen.pro \
--header "Authorization: Bearer gen_pat_…"Manus: Settings → Integrations → Custom MCP Servers → Add Server
- Server URL:
https://mcp.gen.pro - Authentication: Bearer token → your
gen_…PAT
claude.ai connectors / ChatGPT: add a custom MCP connector with the same URL + bearer token.
The hosted server is multi-tenant — each caller acts as their own GEN identity; your PAT scopes everything you can see and do.
MCP (Model Context Protocol) server for the GEN platform — lets Claude and any MCP-compatible AI drive the GEN Auto Content Engine + Agent Core: the full 5-step journey from onboarding an agent to publishing a video.
Step 1 Step 2 Step 3 Step 4 Step 5
────── ────── ────── ────── ──────
Set Up → Generate → Idea to → Edit & → Export &
Agent Ideas Vidsheet Generate Publish
Read the gen://api-reference MCP resource for the full teaching document, and
api.gen.pro for the API reference. Tool descriptions
start with the step name (e.g. "Step 4 (Edit & Generate): …") so AI tooling can
route quickly.
The stdio mode still functions for the migration window:
uvx gen-mcp-server # or: pip install gen-mcp-server && gen-mcp-server
export GEN_API_KEY=your-api-keyIt prints a deprecation notice to stderr on start. Do not build new
integrations on it; use https://mcp.gen.pro.
Optional base-URL overrides (legacy stdio only):
export GEN_API_BASE_URL=https://api.gen.pro/v1
export GEN_AGENT_API_URL=https://agent.gen.pro/v1
export GEN_AGENT_CORE_API_URL=https://agent-core.gen.pro/v1This repository remains the source of the hosted mcp.gen.pro deployment; only
the PyPI distribution is retired.
git clone https://github.com/poweredbyGEN/gen-mcp-server.git
cd gen-mcp-server
uv venv && . .venv/bin/activate
uv pip install -e .
GEN_MCP_PORT=8080 gen-mcp-server --http # hosted streamable-http mode, serves `/` and `/mcp`- API Docs: api.gen.pro — full API reference and guides
- TypeScript SDK: github.com/poweredbyGEN/gen-typescript-sdk
- The downloadable package is deprecated; the official GEN MCP is the hosted
server at
https://mcp.gen.pro(GEN-4785) DeprecationWarningon import when installed from PyPI; stderr notice on stdio CLI start. Hosted (--http) mode is unaffected- Includes everything on main at cut time, including the Python/FastMCP rewrite (GEN-3788) and the 2026-07-30 tool additions (billing, social, schedule, assets, generation, compose — 31 tools) never previously published
- Restructured all 81 tools around the 5-step user journey (GEN-2879)
- Rewrote the embedded
gen://api-referenceresource as a step-by-step teaching document with chained curl examples - Tool descriptions now start with the phase (e.g. "Step 1 (Agent Setup): …") so AI tooling can route quickly
- Added Agent Core flat endpoints (GEN-2755)
- Stripped vendor branding from tool descriptions
- Added
gen_run_research, expanded API_REFERENCE with agentic chat docs - Content monitoring + publishing tools
MIT