Annealage
AI harness for agentic hardware an-NEEL-ij

Heat. Fold. Repeat.

Annealage brings agent-driven tools across hardware development, from schematic design and review to firmware, embedded interfaces, hardware-in- the-loop testing and printable parts. Some agents run locally, others are hosted, and MCP connects agents to the work where it happens.

[ ok ] Source-available. Pod open source under AGPL-3.0.

agent@annealage MCP · 1 SURFACE
01 canvas reload # build_ui re-run <10ms passed
02 canvas tap btn_send passed
03 pod dut flash dut fw.bin # swd, verified passed
04 pod dut gdb dut # arm-none-eabi-gdb over wifi passed
05 pod bench la dut --pins 16-19 # → cap.vcd passed
06 agent: refine, re-run
loop write → flash → test → debug → fix
FIG. 1: the loop

One loop. Every turn driven by the agent.

Canvas and Pod expose their actions over a single MCP surface, so an agent runs the embedded iteration loop end to end. It edits code, flashes the target, exercises it, reads back what happened and refines. No human in the inner loop.

Process / iteration AGENT-DRIVEN
  1. 01 Write code edit
  2. 02 Flash program
  3. 03 Test verify
  4. 04 Debug inspect
  5. 05 Fix refine
Write code. Flash. Test. Debug. Fix. The agent drives every turn.
The surface

One MCP surface across the toolkit.

The tools integrate with agents in different ways. Pod and Mesh are available now. Canvas is in private use. Agents can flash targets, exercise hardware and pin feedback to a part.

Where the toolchain fits the hand.

  • MCP One surface; Canvas alone exposes 8 core tools plus a 23-action catalog.
  • SRC Source-available: Canvas and Mesh free for non-commercial use, Pod open source under AGPL-3.0.
  • LOOP Write → flash → test → debug → fix, with no human in the inner loop.