Flexport opens freight booking to AI agents: assistants can now quote, track and book ocean and air cargo

The details

• Flexport launched the Model Context Protocol server on Tuesday, Sept. 29, letting connected agents track shipments, surface exceptions and customs holds, search rates on any lane, or book outright — submitting cargo details against negotiated rates and returning a booking ID and transit time, according to FreightWaves’ report of the launch.
• The server exposes six tools (track, browse shipments, get a rate, browse quotes, create a booking, account lookup), covers ocean FCL/LCL and air freight across contract and spot markets, and authorizes each assistant under one person’s Flexport login, reaching only what that person can access. An administrator enables the connection org-wide and can revoke any user or switch the whole integration off, per The Conveyor’s review of the terms.
• Requests landing through MCP are processed by Flexport’s own internal AI agents, which the company says handle 21 million tasks a year, including a quoting agent that turns rate requests into custom quotes. Flexport says every agent runs behind guardrails set through a custom evaluation platform, with exceptions routed to human experts.
• Supported front ends include Claude, ChatGPT and Codex, Microsoft Copilot, Meta’s Muse and internal tools; the server runs independently from Flexport’s public API with its own rate limits, the company says. More than 13,000 companies use the platform, per FreightWaves.
• The same release extended Flexport Atlas — the free global-trade visibility map introduced in February — beyond ocean to air, rail and barge routing between any two locations, and added AI-driven HTS code classification for customs.
• Flexport’s software terms, last revised in late August before the launch, cover operations performed “by an AI system on behalf of the User, or as a result of AI system behavior,” including hallucinations, The Conveyor reported; the company “strongly recommends” setting approval requirements for write operations such as creating bookings.
• “The future of global logistics is autonomous,” founder and CEO Ryan Petersen said in the release.

What buyers should do

• Decide, in writing, which actions an AI agent may take unsupervised: tracking and rate inquiry are low-risk; bookings commit spend and should default to human approval under most delegation policies.
• Update your freight RFP and supplier checklists: ask every forwarder and TMS vendor now whether they expose an MCP or agent interface, and what audit trail and revocation controls come with it.
• Check your Flexport (or any carrier portal) account permissions before connecting assistants — each agent inherits a person’s full access, so scope logins before scaling them.
• Put AI-initiated bookings inside your normal purchase controls: spend thresholds, routing-guide compliance checks and invoice-to-booking reconciliation.
• Brief category managers that HTS classification done by AI still needs a customs-compliance owner; misclassification liability does not transfer to the software.

What to watch next

• Adoption data: whether Flexport publishes booking volumes placed through agents, the first real measure of autonomous freight purchasing.
• Copycat launches from competing forwarders and digital freight platforms — agents will need the same governance questions answered before year-end peak bookings.
• Whether U.S.–China port-fee decisions and peak-season capacity strain push more spot booking to automated quoting in Q4.

The bigger picture

This is the execution layer catching up with AI sourcing tools: agents that quote, negotiate and now book freight build on the ROI picture we tracked in logistics AI quoting automation and sit alongside autonomous trucking on the Laredo lane. For procurement, the through-line is that agent interfaces turn rate shopping from a periodic RFQ into continuous pricing — which favors buyers with clean data and explicit approval guardrails.