Freigent exposes a Model Context Protocol (MCP) server at:
https://app.freigent.ai/api/mcpPoint any MCP client at it and your AI assistant can work Freigent for you, read the queue, look at loads and drivers, draft updates, with the same org scoping the app enforces.
What It Can Do
Read tools cover the daily questions: workflow queue state (exact tier order), load and driver detail, broker info, invoice and payment status, compliance items.
Write tools cover the safe operational moves, update load status, assign a driver, create load notes, send driver messages, process rate confirmations, invoice/expense/bookkeeping operations, and workflow task approvals.
What It Deliberately Cannot Do
The same human-held rules as the app apply, plus chat-specific guards:
- Money-moving actions are reject/dismiss only from MCP, invoices are approved in the authenticated portal, never rubber-stamped from a chat.
- Consequential Safety task types cannot be approved or rejected generically: drug tests, accident investigations, FMCSA items and their kin must be resolved in the carrier portal through their dedicated enact paths.
Connecting a Client
MCP clients authenticate via OAuth (the /api/mcp/oauth endpoints); most clients
handle the flow when you add the server URL above. Example for Claude Code:
claude mcp add --transport http freigent https://app.freigent.ai/api/mcpYou will sign in with your Freigent account and approve the org scope. Everything the client can see and do is bounded by your own membership, the server enforces org isolation server-side, the same as the web app.