Active Source of Truth for Your Coding Agents
Tracks which coding-agent decisions are current and feeds only those back
MLA watches your Claude Code or Codex sessions, captures decisions and constraints as they're made, and, before an agent acts on a new prompt, prepends only the decisions still in force — flagging when a newer decision has superseded an older one instead of handing the agent both. Nothing an agent proposes becomes accepted project truth until a person approves it in a review queue, and decisions it can't reconcile are kicked back to a human too. The client is open source and free during beta, and the founder publishes 7 benchmarks against plain CLAUDE.md and RAG, including cases where MLA came up short.
What holds up
- +Open-source client, free during beta
- +Human approval required before anything becomes project truth
- +Publishes benchmarks including its own losses, not just wins
- +Works across both Claude Code and Codex
Mind the limits
- −No stated pricing beyond the free beta
- −Adds a review/approval step rather than fully automating context
Featured here? Take the badge
Put it on your site — it links back to this review. Free for every listed product, always.
<a href="https://stillworks.dev/products/p/active-source-of-truth-for-your-coding-agents/"><img src="https://stillworks.dev/badge/active-source-of-truth-for-your-coding-agents.svg" alt="Picked by StillWorks" width="250" height="54"></a>