Achromat MOM: a shared memory layer that turns agent tool-call failures into reusable prevention lessons
ChannelLivid · reddit · 2026-10-30
A developer launched Achromat Machine Operational Memory (MOM), a free shared memory layer that records what goes wrong when agents call tools — plus the verified fix — rather than conversational context. Before risky calls (payments, credit decisions, shell, SQL, deploys), an agent asks MOM for a verified compliant version; on a hit it gets a compliant-arguments blueprint with [VERIFIEDSTATE] placeholders, a steering directive, and the statutory anchor. On a miss, live policy returns a BLOCK with a fix directive; the corrected BLOCK→ALLOW pair becomes a reviewable lesson shared with every other agent. Lessons exist only when a signed BLOCK and later ALLOW are bound together (Ed25519 per-agent identity), and matching is on argument shape, never values. Ships with an MCP server (npx -y @achromat-ai/mcp-server), HTTP API, and skill.md, starting with fintech, robotics, and DevSecOps.
More from coding & agent
- AI Engineer Code Summit 2026 opens applications, gathering 1,000 top AI engineers — threepointone · 2026-10-31
- Todoist's self-healing agent auto-fixes 1 in 5 Sentry issues and opens PRs — zeeg · 2026-10-31
- Dev's Multica 1.0 build log: 432M tokens, 89 agent runs, just $183 — jiayuan_jy · 2026-10-31
- Stardock's QA now runs agents that play the game and file reports — draginol · 2026-10-31
- Limma MCP structures unstructured data end-to-end so agents never miss a detail — FlimsyIngenuity7248 · 2026-10-31
- Agent audit metric: can you name which tests ran and files changed without reading its summary? — ronbodkin · 2026-10-31