AXME CODE
Log every architecture decision Claude Code makes — and retrieve them later
AI coding agents make dozens of decisions per session — naming conventions, library choices, architecture patterns — and none of it is recorded. Three weeks later, no one knows why.
AXME Code creates a structured decision log from your coding sessions: what was decided, why, what was rejected, and what the trade-offs were.
AI coding agents make dozens of decisions per session — naming, libraries, patterns — and none of it is recorded. Three weeks later, no one knows why.
Decisions your agent can enforce
Architecture choices buried in chat are invisible to the next developer and the next agent session. Three weeks later nobody knows why you picked Postgres over Dynamo or why deploys must go through CI.
AXME Code records decisions with title, rationale, alternatives, trade-offs, and enforcement level (required vs advisory). Required decisions surface in axme_context and can block violations on hosts that support hooks.
CAPABILITIES
How it works.
Decision
What was chosen.
Rationale
Why it was chosen.
Alternatives
What was rejected.
Trade-offs
Costs and benefits captured.
Invisible → queryable
Chat scrollback
# buried in conversation history
Decision log
axme decisions list --tag=auth-refactor
From scrollback to queryable records
Save decisions with axme_save_decision during work — not only at session end. List and search via axme_decisions or the context bundle. Link decisions to incidents or PRs in the body so future readers understand provenance.
Pair with safety rules when a decision must be machine-enforced — e.g. no direct push to main — rather than hoping the model remembers.
Common questions
- What is a required vs advisory decision?
- Required means agents and hooks should treat it as mandatory policy. Advisory guides without automatic blocking on all clients.
- Can decisions conflict?
- Newer decisions supersede older conflicting ones in context; audit KB with axme-code audit-kb for duplicates.
- How do decisions relate to memories?
- Decisions are commitments with enforcement; memories are experiential lessons — both load in axme_context.
Related reading
Deeper dives from the AXME blog.
Decisions Are Not Memories. Most Agent Memory Systems Confuse Them.
'I learned we use Postgres' and 'we decided Postgres over MySQL because of JSON support' are different data. One is a fact. The other is a contract. Mixing them is why agent memory drifts.
Read post →Stop Writing CLAUDE.md Files. Write Decisions Instead.
CLAUDE.md is a passive artifact. Decisions are active rules with enforcement levels, supersede chains, and audit history. Here's why I stopped maintaining CLAUDE.md and what I switched to.
Read post →
Related
Related links
Ship your first durable agent — in under 10 minutes.
Free tier. No credit card. Self-host or hosted — your choice.