← all work.

MCP servers / in production

Freight Data MCPs

Two MCP servers that give Claude direct access to the freight industry's two standard data sources: SMC3 RateWare for LTL tariff rates and PC*MILER for truck miles. They're the backbone of the AI Rating Engine.

role.

built and own it

status.

in production

type.

MCP servers

stack.

mcp, python, smc3 rateware xl, trimble pc*miler, rest apis

Mockup of an AI chat calling the RateWare and PC*MILER MCP servers to check a shipment rate
illustrative mockup with sample data. no client data.
2servers: SMC3 RateWare (CzarLite) and PC*MILER
427 / 427rated shipments matched on an independent recheck in one audit
4xhow much LTL prices can swing between tariff vintages, which is why every rate is checked
1,000sof shipments rated in one call through a file, with caching

why they matter.

Most LTL contracts say something like "70% off CzarLite, dated 2016, with a $142 minimum". To know what a shipment should have cost, you need the tariff price on that exact vintage, then the discount and the minimum. Truckload contracts and fuel surcharges are often priced per mile and the miles in a payment file can be blank or wrong.

smc3 rateware.

Rates one shipment, a batch or thousands at once by reading and writing a file, so big audits never push data through the chat. It can return a full rate matrix for a lane and list which tariffs are loaded.

It checks that SMC3 actually used the tariff date the contract names before trusting a price and it uses the tariff's own minimum and deficit weight logic instead of a rate times weight shortcut. Canada and Mexico are handled with full postal codes.

pc*miler.

Practical route truck miles for one lane or a hundred at a time, plus hours, tolls and geocoding. It can rate on the PC*MILER version a contract names and results are cached by lane so repeat audits are fast.

how the audit uses them.

Before rating, the contract's tariff is validated. Unique shipments are rated once and cached, missing miles are filled in and an independent verifier rechecks every LTL row from the raw SMC3 output.

An MCP server over Azure AI Search, so auditors' AI tools can search contracts and past claims, is in progress.

how it works.

  1. validate the contract's tariff date
  2. rate unique shipments once and cache them
  3. fill in missing miles with PC*MILER
  4. recheck every rated row before it's used