iLEAP MCP

A Model Context Protocol (MCP) server for iLEAP — bring trusted logistics emissions data into any AI assistant.

ileap-mcp is an open, standalone MCP server that exposes iLEAP and PACT emissions data to any MCP client — Claude, ChatGPT connectors, Kiro, Cursor, and the rest. Point it at any iLEAP-conformant endpoint and your AI assistant can list product footprints, inspect transport activity data, and submit events using natural language, without anyone writing a bespoke API integration.

It ships preconfigured against the public SINE Foundation iLEAP demo endpoint, so you can try it end to end in minutes — no credentials of your own required.

Why this exists

Logistics is a large, hard-to-see slice of most products' carbon footprint. The iLEAP standard — free to implement, open source, and software-friendly — lets trustworthy emissions data flow across transport modes and organizations, aligned with ISO 14083 and the GLEC Framework v3.1, and interoperable with PACT.

AI assistants are quickly becoming how people query and reason over operational data. ileap-mcp is the bridge: it turns a conformant iLEAP HTTP API into a set of MCP tools an assistant can call directly. The goal is to lower the barrier to adoption for the whole iLEAP community — implementers, software providers, shippers, and carriers — so that "ask your assistant about your logistics emissions" becomes a one-command setup.

Related MCP server: js-translation-helps-proxy

What it does

The server wraps the endpoints an iLEAP-conformant host serves and exposes each as an MCP tool:

Authentication to the upstream endpoint uses OpenID discovery + OAuth2 client-credentials; pagination follows the RFC 8288 Link cursor convention required by the specs. The server is transport-agnostic on the client side: it speaks Streamable-HTTP for hosted deployments and stdio for local use.

Quick start

Try it in 60 seconds (against the SINE demo)

Build and run the container locally, talking to the live demo endpoint:

$ docker build -t ileap-mcp .
$ docker run --rm -p 8080:8080 \
    -e ILEAP_BASE_URL=https://api.ileap.sine.dev \
    -e ILEAP_CLIENT_ID=hello \
    -e ILEAP_CLIENT_SECRET=pathfinder \
    ileap-mcp

$ curl http://127.0.0.1:8080/healthz
{"status":"ok","upstream":"https://api.ileap.sine.dev"}

The MCP endpoint is then reachable at http://127.0.0.1:8080/mcp.

Add it to your MCP client

Point any MCP client at the running server. For a local stdio setup:

{
  "mcpServers": {
    "ileap": {
      "command": "python",
      "args": ["-m", "ileap_mcp.server"],
      "env": {
        "ILEAP_MCP_STDIO": "1",
        "ILEAP_BASE_URL": "https://api.ileap.sine.dev",
        "ILEAP_CLIENT_ID": "hello",
        "ILEAP_CLIENT_SECRET": "pathfinder"
      }
    }
  }
}

For a hosted server over HTTPS:

{
  "mcpServers": {
    "ileap": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://your-host.example.com/mcp"]
    }
  }
}

Run from source

$ git clone https://github.com/bernard357/ileap-mcp.git
$ cd ileap-mcp
$ cp .env.example .env          # defaults target the SINE demo
$ make setup                    # create venv, install deps
$ make run                      # serve Streamable-HTTP on :8080

Point it at your own endpoint

ileap-mcp is a swappable client — the same tools work against any conformant iLEAP host. Change the environment and you are done:

Deployment

The server is a plain container and runs anywhere containers run — locally via docker run, or on any container platform. See DEPLOYMENT.md for building the image, configuration, and running it in production. The repository is a clean, self-contained server that anyone can build and run on the platform of their choice.

Standards and conformance

• iLEAP Technical Specifications v1.1.0 — https://specs.ileap.global/

• PACT (Partnership for Carbon Transparency, WBCSD) — https://www.carbon-transparency.org/

• ISO 14083:2023 and the GLEC Framework v3.1 — the methodological basis

• Automated Conformance Testing (ACT) — https://act.sine.dev/

This project is a client of the iLEAP protocol; it does not redefine the standard. Where behavior is ambiguous, the specifications are authoritative.

Community and contributing

ileap-mcp is built to be adopted and extended by the iLEAP community. Whether you implement iLEAP, provide logistics software, or just want your assistant to speak emissions data, contributions are welcome.

• Discuss and propose — open a GitHub Discussion or an issue.

• Contribute code — read CONTRIBUTING.md for the dev setup, coding conventions, test expectations, and the pull-request flow. Good first areas: additional iLEAP data types, richer filtering, client-side auth strategies, and conformance test coverage.

• Be kind — participation is governed by our Code of Conduct.

We review pull requests openly and aim to keep the project a neutral, useful piece of shared infrastructure for the whole iLEAP ecosystem.

Origin and stewardship

ileap-mcp grew out of conversations within the iLEAP community, where the idea of a concrete, shared contribution — a standard way to bring logistics emissions data into AI assistants — resonated with other members. This project is that contribution: a standalone, vendor-neutral MCP server that anyone can run, fork, and build on.

It is maintained by Bernard Paques and is kept deliberately independent of any single platform, so it stays useful shared infrastructure for the whole iLEAP ecosystem.

Questions or anything else: bernard.paques@gmail.com.

License

Licensed under the MIT License. You are free to use, modify, and distribute this software, including commercially, subject to the terms of the license.

Acknowledgements

Thanks to the SINE Foundation and the Smart Freight Centre for developing and openly sharing the iLEAP standard and the public demo endpoint this project targets, and to WBCSD PACT for the underlying data-exchange protocol.