Secure Gmail access for your AI agents
Add FGAC as an MCP server, sign in with Google, and control exactly what your agents can read, send, and delete.
Add the MCP server
Every tool uses the same URL: https://fgac.ai/api/mcp
Claude Code
claude mcp add \
--transport http \
fgac https://fgac.ai/api/mcpRun this in any project directory. The server persists across sessions.
Cursor
Open Settings → MCP Servers. Click Add new HTTP server. Paste the URL above.
Windsurf
Open Settings → MCP. Click Add Server. Choose Streamable HTTP and paste the URL above.
Claude Desktop
Open Settings → Developer → MCP Servers. Add a new server with type "url" and paste the URL above.
FGAC uses MCP Streamable HTTP with OAuth 2.1 (DCR + PKCE). Your client handles authentication automatically.
Sign in & approve the connection
Your agent triggers an OAuth flow. Then approve it in your dashboard.
- AYour agent tries to use a tool — a browser window opens asking you to sign in with Google. This links your agent to your FGAC account. No API keys needed.
- BApprove in your FGAC Dashboard — new connections appear as ⏳ Pending. Click Approve and assign a Key Profile to grant access.
Set access rules
Control exactly what each agent can read, send, and delete.
Block sensitive emails — 2FA codes, password resets, financial alerts. Uses regex pattern matching on subjects and senders.
Only allow sending to approved domains or specific email addresses. Everything else is rejected.
Create separate profiles with different permissions. Assign each agent connection to the right one.
Multiple email accounts
One agent connection can access multiple Gmail inboxes. Other users delegate access from their own FGAC dashboard, then your agent uses the account parameter to specify which inbox.
Alternative: Direct Proxy Key
For agents that don't support MCP, create a proxy API key and use our downloadable skill files.