Back to feed
Dev.to
Dev.to
7/9/2026
basou: A local-first harness for steering AI coding agents with replayable decision records

basou: A local-first harness for steering AI coding agents with replayable decision records

Original: When AI Agents Write the Code, What's Missing Are the Reins — Introducing basou

Short summary

basou is an open-source, local-first harness for steering AI coding agents, tackling three pain points: context loss between sessions, re-explanation tax, and lack of auditability. It captures decisions, handoffs, and event logs as plain JSONL files with hash-chain integrity, integrating with Claude Code and Codex without running its own LLM. The tool keeps a single source of truth and regenerates human-facing Markdown from it, making agent workflows verifiable and offline-capable.

  • basou provides a local-first harness for managing AI coding agent sessions with decision records, handoffs, and replayable event logs
  • Works with Claude Code and Codex by wrapping or importing their logs; never runs its own LLM, keeping records verifiable and offline
  • Uses JSONL with hash-chain integrity as the single source of truth, regenerating Markdown handoff and decision docs for human review

Generated with AI, which can make mistakes.

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more