AXME CODE
Give Claude Code a memory that persists across every session
Claude Code forgets your project every session. Developers spend 10–20 minutes re-explaining context every morning, re-uploading files, re-summarizing what was decided last week.
AXME Code stores decisions, file maps, and active context so Claude Code starts each session with full project knowledge — not from a blank slate.
Claude Code forgets your project every session. Developers spend 10–20 minutes re-explaining context every morning.
Memory that survives the chat window
Coding agents start each session from zero — stack choices, folder conventions, and open tasks get re-explained in long pastes. That tax compounds across a team and across weeks of a project.
AXME Code stores structured knowledge under .axme-code/: oracle (stack, layout, patterns, glossary), decisions with enforcement levels, pattern memories from past debugging, and safety rules — loaded via axme_context at session start.
CAPABILITIES
How it works.
Decisions
Architecture choices persisted.
File maps
What matters in the repo.
Active context
Open tasks and conventions.
DEEP DIVE
Production patterns.
Search & retrieve
Query past decisions.
Team-shared
Context for the whole squad.
Size management
Summarize and prune smartly.
Session startup time
Without AXME Code
# 15 min: re-explain repo, paste files, recap decisions
With AXME Code
# axme_context(path="/repo") # oracle + decisions + handoff loaded
Seven modules in .axme-code/
Oracle answers where things live and how the repo is built. Decisions record ADR-style choices so debates do not restart. Memories capture lessons from incidents and wins. Safety holds git, shell, and path policy. Backlog tracks cross-session tasks. Handoff documents where work stopped. Worklog records what already happened.
In Cursor, connect via .cursor/mcp.json; in Claude Code, use the marketplace plugin or MCP config. All clients share the same storage with atomic writes.
Common questions
- Is .axme-code/ committed to git?
- Typically gitignored — it is local operational memory. Export decisions you want in-repo via your own ADR process if needed.
- How is this different from CLAUDE.md?
- Static markdown is one file without enforcement, search, or handoff rituals. AXME is categorized, queryable, and updated through MCP tools.
- Does it work in Cursor?
- Yes — see /integrations/cursor/ for MCP setup and session-start rules.
Related reading
Deeper dives from the AXME blog.
How I Fixed Claude Code's Amnesia (Actually)
Claude Code forgets your project every session. CLAUDE.md helps, but it doesn't scale. I built an MCP server plugin that gives Claude Code persistent memory, decisions, and safety guardrails.
Read post →The 4 Types of Context Your AI Coding Agent Needs
Most agents treat all context as 'stuff in the context window'. But facts, patterns, project state, and references have different lifecycles and different retrieval rules. Mixing them is why agents feel dumb.
Read post →You've Had This Conversation With Claude Code 47 Times
Every session starts the same way: explaining your stack, your deploy flow, your conventions. You've said it forty times. The agent keeps nodding. Here's why re-explaining every session is the wrong fix.
Read post →
Related Code features
Related links
Ship your first durable agent — in under 10 minutes.
Free tier. No credit card. Self-host or hosted — your choice.