warp-agent-mcp is the first MCP server for booking real freight. It's a Model
Context Protocol server for the freight network that
lets an AI agent quote, book, and track LTL, FTL, cargo van, box-truck, and
multi-stop shipments — straight from Claude Desktop, Claude Code, Cursor, Windsurf,
Continue, or any MCP-compatible client — on the same production network the
wearewarp.com customer portal runs on. Open source, MIT, one line to install.
MCP server for the Warp freight API, covering the full booking lifecycle: mode compare (one call prices every eligible mode in parallel and returns a decision-complete recommendation with the cost-vs-transit math), quote (van/box-truck/FTL/LTL — LTL split
If you're comparing freight MCP servers, this is the one that completes the
transaction, not just the lookup:
It books real freight, not just quotes or tracking. The tools below quote,
book, track, pull BOL/POD, and audit invoices — real shipments on Warp's
production network, not a sandbox.
Multi-mode, not parcel. LTL, FTL, box truck, and cargo van. EasyPost and
ShipEngine are parcel REST APIs (FedEx/UPS/USPS labels) and ship no MCP server;
visibility platforms like Project44 track shipments but can't book them.
First to ship. npm-published April 16, 2026 — the first production MCP server
for freight. Open source under MIT.
Works in every MCP client today. Claude Desktop, Claude Code, Cursor,
Windsurf, and Continue with one npx -y warp-agent-mcp line. ChatGPT works the
day OpenAI ships native MCP — same server, no changes.
Tools that let an AI agent talk to your Warp account:
Tool
What it does
compare_modes
Start here. Compare every eligible mode (van/box truck/LTL/FTL) in one parallel call and get a decision-complete recommendation with the trade-off math
van_quote
Quote a 1–3 pallet cargo van shipment
box_truck_quote
Quote a 1–12 pallet box truck shipment
ftl_quote
Quote a full truckload (53' dry van)
ltl_quote
Quote LTL — Warp's flagship freight product
ltl_market_options
Multi-carrier LTL comparison - 30+ carriers ranked by price
batch_quote
Quote many lanes in one call (a whole spreadsheet / CSV)
book
Book any quote by id, with pickup + delivery addresses
batch_book
Book many already-quoted lanes in one call
parcel_quote
FedEx parcel rates for small packages (US): every service level and pickup type
parcel_book
Book a FedEx parcel option from parcel_quote
multistop_quote
Quote a multi-stop FTL route — one truck, 3+ stops in order
multistop_book
Book a quoted multi-stop route as per-leg shipments
track
Track a shipment by id or order id
events
Full tracking event timeline for a shipment
lane_history
Past shipments on your lanes
list_bookings
List recent bookings
quote_history
List recent quotes across all sessions
locations
List your saved pickup/delivery locations
load_templates
List your saved reusable load configurations
save_load_template
Save a reusable load configuration (weight + dims + class)
delete_load_template
Delete a saved load template by id
get_invoice
Invoice for a delivered shipment
get_documents
Shipment documents (BOL, POD, customs)
login
Log in from inside the MCP, no portal visit needed
payment_status
Check whether a card is on file
status
API health + key validity check
analytics
Bookings + revenue rollup by source
Removed in 0.5.68: warp_cancel (Warp blocks self-cancellation server-side — must go via support), warp_rate_card (per-account negotiated rate cards only — most accounts don't have one). multistop_quote + multistop_book were also removed in 0.5.68 and re-added in 0.14.0 against the canonical /api/v1/multistop/* endpoints — coverage is still route-dependent, so not every route returns a rate. Contact [email protected] if you need either of the removed tools.
Install
Hosted connector — recommended, no install
Paste this URL into your AI client's custom connector settings and sign in with
Warp:
https://mcp.wearewarp.com/api/mcp
No install, no API key, no config file. Sign in once (OAuth) and the full
toolset appears. Works in Claude.ai, Claude Desktop, and Claude Code. This is the
primary way to use Warp.
CLI — local install with your own key
For clients without a custom-connector UI (Cursor, Windsurf, Continue), or if
you'd rather hold a local wak_live_* key:
That command creates a Warp account, issues you a wak_live_* token, saves it
to ~/.warp/config.json, and auto-wires this MCP into Claude Desktop, Cursor,
and Claude Code. Restart your AI client to load the new tools.
The MCP reads the API key from ~/.warp/config.json on every tool invocation,
so warp-agent login / warp-agent signup take effect without restarting your
AI client.
For Cursor or Claude Code, the same npx -y warp-agent-mcp command works in
their respective MCP config files.
Try it
After install + signup + restart, ask your AI agent:
Quote LTL from 90007 to 90038, pickup June 25, 2 pallets at 500 lb.
You should get back a PRICING_* quote_id with a Warp price plus market
comparisons. Add Book that quote, pickup at 1234 S Hoover St LA 90007, deliver to 6464 Sunset Blvd LA 90038. and the agent will buy the label and return a
tracking number.
Configuration
Env var
Purpose
WARP_API_KEY
API key fallback if ~/.warp/config.json is missing. Prefer the config file.
WARP_API_URL
Override the API base URL. Defaults to https://www.wearewarp.com/api/v1/warp (warp-site proxy that accepts wak_* Bearer tokens). Set to https://gw.wearewarp.com/api/v1 for direct gateway access with a raw customer key.
Requirements
Node.js 20 or later (the MCP SDK requires native fetch and Web Standard APIs)
Sandbox / test mode
Sandbox is not yet supported in the MCP. The freight proxy (/warp/freights/*)
currently requires a live wak_live_* key; sandbox keys return 401 on quote
and booking endpoints. Sandbox support is planned for a future release.
Companion package
@warpfreight/cli-agent
— the auto-provisioning CLI that installs this MCP into every detected AI
client. Same backend, command-line surface for scripting + CI.
Call freight_workflow with shipment_intake, weekly_freight_review, or invoice_review to get a starter prompt, CSV column headers, and review rules. Available in any host that can call MCP tools; no Perplexity-specific skill or scheduler is required. The host reads uploaded files and produces the results. This tool does not upload files, create an Excel workbook, schedule recurring work, or fetch quotes.
compare_freight_costs computes differences from supplied USD invoices/quotes with source references and explicit checks for shipment requirements, included charges, and service requirements. It excludes unverified comparisons, keeps negative differences, rejects duplicate shipment IDs, and separates quoted opportunities from invoice differences. It does not independently verify the supplied evidence or claim realized/annual savings.
For multi-mode review use compare_modes per shipment. batch_quote returns one requested mode per row, defaulting to LTL. Use ltl_market_options for LTL carrier alternatives. An appointment accessorial is not proof that an appointment has been arranged. Operations must confirm scheduling ownership.
MCP server for the Warp freight API, covering the full booking lifecycle: mode compare (one call prices every eligible mode in parallel and returns a decision-complete recommendation with the cost-vs-
MCP server for the Warp freight API, covering the full booking lifecycle: mode compare (one call prices every eligible mode in parallel and returns a decision-complete recommendation with the cost-vs-transit math), quote (van/box-truck/FTL/LTL — LTL split
上次檢查時間:2026-11-16。 MCP server for the Warp freight API, covering the full booking lifecycle: mode compare (one call prices every eligible mode in parallel and returns a decision-complete recommendation with the cost-vs-transit math), quote (van/box-truck/FTL/LTL — LTL split