OSS Series · #3

The Hottest OSS CAN & Bus Tooling for the Agent Loop

Part three of our running list of the open source that matters when an AI agent — not just a human — is working on the car. First the debuggers, then the SDV stack. This time we go down to the wire: the CAN and diagnostics projects an agent can call directly to read, decode and drive the bus. Nine that are live on GitHub and worth knowing.

Every layer above eventually resolves to frames on a bus. Reading them, decoding them against a database, running a diagnostic session, replaying a capture — that's the unglamorous plumbing where an AI agent either has a clean programmatic seam to work through, or it doesn't. The good news: the open-source CAN toolchain is almost entirely scriptable — command-line output an agent can parse, Python objects it can construct, packet layers it can build — which is exactly the surface an agent needs to close the loop from "read the bus" to "act on it."

This list is the bus-level counterpart to our Open Source Register. Same rules: these are independent projects at varying maturity, listed for their role in an agent-native workflow, not ranked by stars. Inclusion is not endorsement — verify each against its own repository before adopting.

Part of a series

This is #3 in our running list of the hottest open source for agent-native automotive engineering. Catch up on #1 — MCP servers wiring hardware debuggers (TRACE32, GDB) into AI agents — and #2, the SDV stack an agent can drive. Next up: #4, the embedded runtimes & CI simulation agents lean on.

Building or maintaining automotive open source?

The New Automotive is a curated register and marketplace. Want your project listed, corrected, or a warm intro to teams adopting it? Reach out — no cost, no spam.

Request an Intro