Skip to main content
The hosted Genseo MCP server exposes project-bound SEO tools through the Model Context Protocol. It uses Streamable HTTP and the same allowlisted operations as the Genseo Public API.

Endpoint

An MCP connection can only access the project attached to its authorization. It cannot enumerate a workspace, create projects, read billing data, or access internal database operations.

Before you connect

You need:
  1. A Genseo account and project.
  2. A project-bound API key from Project → Brand → API / Developer.
  3. An MCP client that supports Streamable HTTP or a local stdio process.
Keep the key in environment or encrypted secret storage. Never place credentials in the MCP URL.

Claude Code

Create .mcp.json in your plugin or project configuration:
Set the key before starting Claude Code:
Open /mcp and verify the genseo connection.

Codex

Use an environment-backed Bearer token:
Then set the key before starting Codex:

Other MCP clients

Configure the same Streamable HTTP endpoint and send this request header from the client’s secure credential storage:
Clients that do not support remote HTTP MCP can use the local CLI server described below.

Local stdio MCP

Install the Genseo CLI globally:
Configure the client to launch genseo-mcp:
Or run without a global installation:

Verify the connection

Ask the agent:
A successful connection calls genseo_me and returns the authorized project. If the returned project is unexpected, stop and replace the credential before using write tools.

Available tools

Every tool includes a human-readable title and safety annotations that identify read-only and potentially destructive operations.

Required agent behavior

  1. Call genseo_me before another Genseo tool.
  2. Use only the returned project.
  3. Read a post before updating it.
  4. Prefer drafts before generation or publishing.
  5. Check integrations before publishing.
  6. Ask for confirmation before genseo_posts_publish unless autonomous publishing was explicitly requested.
Treat authorization and project-boundary errors as final. Back off after rate limits and treat 202 Accepted as successful queueing.

Plugin package

For an installable package containing the MCP configuration and shared Agent Skills, use the public Genseo MCP repository or follow the AI Plugins guide.