Craft, not conjuring.
Open-source tools that make AI agent workflows auditable, bounded, and trustworthy.
AI agents act at machine speed. The discipline to govern them must be built in, not bolted on. heurema builds infrastructure for teams that delegate aggressively and verify everything.
- Discovery before automation. Find better work, not just faster work.
- Accountability is not delegatable. Agents execute. Humans own outcomes.
- Context is a product. Bad context produces confident wrong output.
- Tests are contracts. If a machine cannot enforce it, it is a suggestion.
- Silent autonomy is a defect. No audit trail means no work.
- Failure is never silent. It surfaces immediately and names an owner.
- Open source is a constraint. If it cannot be reviewed, it does not ship.
| Tool | What it does |
|---|---|
| signum | Risk-adaptive development pipeline with adversarial consensus code review |
| herald | Daily curated news digest — zero API keys, fully local |
| proofpack | Proof-carrying CI gate for AI agent changes |
| arbiter | Multi-AI orchestrator — Codex CLI + Gemini CLI for review, ask, implement, panel |
| teams-field-guide | Comprehensive guide to Claude Code multi-agent teams |
| emporium | Plugin marketplace — install any tool with one command |
claude plugin marketplace add heurema/emporium
claude plugin install signum@emporiumThe measure of a heurema tool: when something goes wrong, you know exactly where, and exactly whose call it was.