Documentation
MCP

MCP: How your AI agent talks to Forge

Forge exposes every content operation as an MCP tool. Your AI agent connects once and can create, update, publish, and query content — within the same role system your human editors use.

MCP (Model Context Protocol) is an open protocol by Anthropic for structured communication between AI agents and external systems. Forge implements MCP natively — it is not a plugin or an add-on.

What your agent can do

Once connected, your AI agent has access to a set of typed tools — one per content operation. Each tool enforces the same validation and role rules as the HTTP API.

ToolRole RequiredDescription
create_postAuthor+Draft new content entries.
publish_postEditor+Promote drafts to live status.
list_postsEditor+Query and filter all entries.
delete_postEditor+Remove content permanently.

Connecting to Forge MCP

Forge MCP runs as a stdio server. Add it to your Claude Desktop config or any MCP-compatible client.

claude_desktop_config.json
{
  "mcpServers": {
    "forge": {
      "command": "forge-mcp",
      "args": ["--url", "https://your-forge-site.com/mcp"]
    }
  }
}

Authentication

All MCP tool calls require a bearer token. Tokens are scoped to a role — Author, Editor, or Admin. Create and revoke tokens via the token management tools or the CLI.

Token security

Store bearer tokens in environment variables. Never commit them to version control.

Minimum role

Assign the minimum required role. An Author token can draft but cannot publish.