assurance

Your AI agent says it's done. Assurance tells you what it didn't check.

uvx assurance audit          # in a project where you've used Claude Code
uvx assurance audit --demo   # or see a report on a bundled sample session

Real output on the bundled sample. The agent was asked to fix a rounding bug "and make sure the tests pass", and ended with "All done — the totals are correct now."

Claude Code session demo-8f2 — 13 min in /home/you/my-app
10 tool calls, 3 failed — Bash 6, Edit 2, Grep 1, Read 1

  Looped: 3 rounds of Bash `pytest -q tests/test_invoice.py` failing the same way, with nothing new read
  After the last edit (14:09): no test or check command ran
  Not classified: 2 shell commands (make lint-fix, python script), so whether they read, wrote or tested anything is unknown.
  Also in the transcript: 1 assistant turn, 1 user turn, 1 bookkeeping record.
  Not read: 0 lines.

Run it after every session. Add this to ~/.claude/settings.json. When Claude finishes without testing its last edit, you're told, and with --nudge Claude is sent back to run the tests:

{ "hooks": { "Stop": [ { "hooks": [ { "type": "command", "command": "uvx assurance audit --hook --nudge" } ] } ] } }

Every command

Nothing here consults a model, opens the network, or reports a check it could not run as a pass.

This package contains no code of its own. It installs assurance-cli, assurance-deps, assurance-budget and assurance-authority, each of which also installs on its own. The MCP server is separate: pip install assurance-mcp.

Source, issues and documentation: https://github.com/i-ops-hq/assurance · Apache-2.0.