Embedder builds AI coding agents for firmware — 100 paying users and 500K lines of code changed in 3 weeks
Sep 10, 2025 with Bob Wei
Key Points
- Embedder has built an AI coding agent for firmware development that solves a gap general-purpose tools like Claude Code cannot: it provides chip-level documentation and runs code directly on hardware via GDB debugging.
- Three weeks after launch, Embedder has over 100 paying users and 500,000 lines of code changed, with an enterprise deal in progress.
- The company targets embedded systems across manufacturers including STM32, ESP, TI, and NXP, betting that firmware development tooling lags web development by years.
Summary
Embedder is building AI coding agents for firmware on microcontrollers and embedded Linux devices from manufacturers including STM32, ESP, TI, and NXP. The company's CTO Bob Wei came from Tesla's robotaxi program and identifies a gap that general-purpose coding agents like Claude Code and Cursor cannot fill. Those tools lack chip-level documentation context and cannot test code directly on hardware. Embedder solves both problems with an information and hardware interaction layer that pulls device outputs and runs a GDB debugger against the board.
Three weeks after launch, Embedder has over 100 paying users and 500,000 lines of code changed. The company has a first enterprise deal in progress. Firmware development at research labs and trillion-dollar manufacturers runs far behind the tooling available to web developers. Closing that gap is Embedder's wedge.