hwcontract

Your firmware is correct on paper and wrong on the wire.

Coding agents write WS2812 drivers, ESC bitstreams, and boot logs that pass review and then fail the moment the signal hits a real chip. hwcontract closes that loop. It captures what the hardware actually did and returns a verdict you can act on:

• pass: within spec
• marginal: in spec but too close to a rail. Works on your bench, dies on a cold board in the field. It fails the verdict: the judge will not ship it.
• fail: out of spec, with the measured value and how far off it is

Two things a green verdict gives you beyond the table:

• Every pulse is judged, not just the median. Captures carry the full pulse distribution; a glitchy tail that a median hides comes back as marginal or fail, with the violating-pulse count in the hint.
• Evidence on every verdict: contract hash, capture hash, capture parameters, tool version, timestamp. A green build in CI traces back to the exact bytes that produced it.

No hardware in your hand? The demo below runs the whole thing on a real recorded signal, so you can see exactly what you get before wiring anything up.

hwcontract judging a real WS2812B capture, a DMA-broken SPI trace, and a serial boot log

See it work in 30 seconds

pip install hwcontract
python3 -m hwcontract.judge --demo

That judges a real 24-LED NeoPixel capture against two contracts. Same signal, two verdicts:

measured on the real WS2812B signal (300000 samples @24MHz):
  T0H 333 ns   T1H 833 ns   T1L 417 ns   T0L 917 ns   RESET 992250 ns

=== generic WS2812 contract -> FAIL ===
  T0H     350   333  PASS
  T0L     800   917  MARGINAL  only 33ns from max; nudge toward typ 800
  T1H     700   833  MARGINAL  only 17ns from max; nudge toward typ 700
  T1L     600   417  FAIL      183ns short (typ 600)
  RESET 50000    -  PASS

=== matching WS2812B contract -> PASS ===
  (all five edges PASS)

Same hardware, two contracts: the generic one fails, the chip-specific one passes. A WS2812B isn't a WS2812. Measure the real signal, hold it to a spec, and match the contract to the actual chip.

See the temporal engine catch a DMA bug

python3 demo/spi_dma_temporal.py

100 synthesized SPI frames as raw CS/SCK/MOSI waveforms at 100MHz, reduced to pin edges and judged against the bundled spi-frame contract. Frame 77 has the Zephyr LPSPI DMA fault: chip-select asserts after the clock starts. Frame 42 settles MOSI 10ns before the sampling edge. Both come back with exact timestamps, and the same broken edges are re-imported as sigrok-style B/E jsontrace annotations:

cs-precedes-first-clock  800  1  FAIL  trigger at 1540310ns: no gpio.cs.falling
                                          in [1530310ns, 1540310ns] (first of 1)
mosi-setup               800  1  FAIL  forbidden spi.mosi.* at 843300ns is 10ns
                                          before spi.sck.rising at 843310ns

The data is perfect in all 100 frames; a loopback test passes. The ordering is broken in two, and only a cross-signal assertion notices.

What you get

• 28 bundled contracts for the parts people actually use: WS2812/WS2813/ SK6812 NeoPixels, DShot ESCs (150/300/600/1200), servos, I2C, NEC IR remotes, DS18B20, DHT11/DHT22, HC-SR04, A4988/DRV8825 stepper drivers, PWM fans, plus serial boot logs for ESP32, ESP8266, Zephyr, MicroPython, Raspberry Pi, U-Boot, and STM32 bootloaders. Each one has the datasheet's real min/typ/max numbers.
• Temporal assertions between decoded events. SVA-style cross-signal checks (ordering, setup windows, forbidden states) on sigrok jsontrace output, judged for every occurrence with latency percentiles and first-failure timestamps.
• Add a protocol by dropping in one YAML file. No code change.
• An MCP server your agent can call, or plain CLI commands you can run by hand.
• Evidence on every verdict: contract hash, capture hash, capture parameters, tool version, timestamp. A green build traces back to the exact bytes.
• Reasonable by default: timing edges are all measured in nanoseconds, serial contracts are Python regex, verdicts come back with the measured value and the delta so an agent knows exactly what to fix.

Install

pip install hwcontract              # judge + logic-analyzer adapter
pip install "hwcontract[serial]"    # + live serial capture (pyserial)
pip install "hwcontract[untrusted]" # + google-re2 (ReDoS-immune, for untrusted contracts)
pip install "hwcontract[all]"       # everything

Live logic-analyzer capture (check_ws2812 / check_dshot) also needs sigrok-cli on PATH. Judge-only tools (judge_contract, judge_serial) need nothing extra.

Wire it into an agent

One stanza per client, add it once. After install, the hwcontract command is on your PATH.

Claude Code

claude mcp add hwcontract -- hwcontract

Codex CLI: ~/.codex/config.toml

[mcp_servers.hwcontract]
command = "hwcontract"

opencode / Cursor / Gemini / any stdio MCP client

{ "mcpServers": { "hwcontract": { "command": "hwcontract" } } }

Transport is stdio by default (local, no auth surface). For remote-only clients (e.g. ChatGPT connectors), run hwcontract --http 8791 and expose it via a tunnel with HWCONTRACT_TOKEN set for bearer auth.

Speaks MCP 2026-07-28, the stateless revision: per-request _meta, server/discover, no handshake. Clients that still open with initialize get the old shape back. Each request picks its own era, so nothing to configure.

If the client can't find hwcontract (PATH issues)

GUI apps and some agents don't inherit your shell PATH, so a bare hwcontract can fail with "command not found". Two robust fixes:

• Use the absolute path: which hwcontract → put that full path in command.
• Or invoke via Python (no PATH lookup for the script): command: "python3", args: ["-m", "hwcontract.server"]. Works from any directory once installed.

Contract paths: pass an absolute contract_path, or set HWCONTRACT_ROOT to your contracts folder. Relative paths resolve against it, defaulting to the process's working directory, which the client controls and may not be your project. Paths outside the root are rejected. Bundled examples install with the package under hwcontract/examples/.

The tools

Event contracts are the SVA-style layer: relationships between decoded events, checked for every occurrence, with latency distributions and first-failure timestamps. Feed them sigrok-cli --protocol-decoder-jsontrace output and judge from the CLI:

python3 -m hwcontract.temporal spi-frame.contract.yaml trace.json

The bundled spi-frame.contract.yaml catches the Zephyr LPSPI class of bug (CS asserting after SCK starts, MOSI setup collapse) that loopback tests cannot see.

Prefer plain pytest over MCP? pytest-hwcontract is a plugin that turns verdicts into tests: a FAIL, MARGINAL or MISSING edge fails the test with the verdict table in the message, JUnit included.

Gate CI on it

The repo ships a GitHub Action, so captures checked into the repo get judged on every PR:

- uses: MohibShaikh/hwcontract@action-v0
  with:
    timing: "ws2812b=captures/strip.csv"     # contract=capture-glob, bundled names work
    serial: "boot=logs/boot.log"
    samplerate: 24000000                     # for CSV captures (0/1 per line)
    junit: hwcontract-junit.xml              # shows in the tests tab

A FAIL, MARGINAL or MISSING edge fails the step, annotates the failing line, and writes JUnit. The action self-tests on every push to this repo with one clean and one deliberately broken capture.

How it fits together

  observers (capture)                  judge (this repo)
  ─────────────────────                ─────────────────
  logic analyzer  ─ pulse widths ─┐
  serial port     ─ log text ─────┼─►  contract × observation  ─►  pass/marginal/fail
  sigrok jsontrace ─ events ──────┘         (judge.py / temporal.py)

• judge.py. The pure judge for timing and serial, plus contract validation. No hardware, no framework, cached.
• temporal.py. Cross-event temporal assertions: selectors, signed windows, latency distributions, first-failure timestamps.
• jsontrace.py. Imports sigrok-cli's Google Trace Event JSON into normalized events.
• sigrok_adapter.py. Turns a logic-analyzer capture into pulse-width distributions for WS2812 and DShot.
• serial_adapter.py. Captures a serial log, or replays a saved one.
• server.py. The MCP server, stdio and HTTP JSON-RPC, stdlib only.
• *.contract.yaml. What "correct" looks like. Human-editable, and they double as regression tests.

The contract format

Timing (ws2812.contract.yaml, dshot.contract.yaml): pulse widths in ns

contract: ws2812
headroom_pct: 20         # in-spec but within 20% of a rail => "marginal"
edges:
  - {name: T0H, min: 200, typ: 350, max: 500}   # '0' bit high time

Serial (boot.contract.yaml) uses Python regex:

contract: boot
kind: serial
expect: ["IMU init OK", "boot v\\d+"]
forbid: ["panic", "Guru Meditation", "\\bnan\\b"]

Events (spi-frame.contract.yaml) assert relationships between decoded events — SVA-style temporal checks on raw pin edges or sigrok annotations:

contract: spi-frame
kind: events
assertions:
  - {name: cs-precedes-first-clock, when: spi.sck.rising,
     require: gpio.cs.falling, within: [-10us, 0ns]}
  - {name: mosi-setup, when: spi.sck.rising,
     forbid: spi.mosi.*, before: 20ns}

Add a protocol = drop a new YAML. No code change for another timing signal.

Kill switch

Instantly disable every hardware-touching tool (captures) while leaving the pure judge tools working:

export HWCONTRACT_SAFE=1          # env, or:
touch /home/tsd/projects/hardware/KILLSWITCH   # file next to server.py

Security

Every tool argument is treated as hostile, since the caller is an LLM that can be prompt-injected. Contract paths are confined to the server dir, HWCONTRACT_ROOT overrides that. driver/channel/port are charset-validated, samples/seconds/samplerate are clamped, sigrok-cli runs with a timeout, YAML is safe_load. Do not expose this server over the network without adding authentication.

Self-tests: no hardware, run from anywhere

hwcontract --selftest                       # full MCP round-trip
python3 -m hwcontract.judge --demo
python3 -m hwcontract.sigrok_adapter --demo
python3 -m hwcontract.serial_adapter --demo
pytest                                      # the tests/ suite; pip install -e .[dev]

CI runs the suite on every push and PR, and the PyPI publish job waits for it.