{
  "markdown": "# AXME Code — Claude Code Plugin\n\n> **Alpha** — This plugin is in active development. Expect breaking changes between versions. [Report issues](https://github.com/AxmeAI/axme-code/issues).\n\nPersistent memory, architectural decisions, and safety guardrails for Claude Code. Your agent starts every session with full project context — stack, decisions, patterns, safety rules, and a handoff from the previous session.\n\n[![Alpha](https://img.shields.io/badge/status-alpha-orange)]()\n[![Version](https://img.shields.io/badge/version-0.6.3-blue)]()\n[![License: MIT](https://img.shields.io/badge/license-MIT-blue)](LICENSE)\n\n**[Main Repository](https://github.com/AxmeAI/axme-code)** · **[Website](https://code.axme.ai)** · **[Issues](https://github.com/AxmeAI/axme-code/issues)**\n\n---\n\n## What It Does\n\nClaude Code forgets your project every session. AXME Code fixes that.\n\n- **Persistent Knowledge Base** — oracle (stack, structure, patterns, glossary), architectural decisions with enforcement levels, accumulated memories from past sessions\n- **Safety Guardrails** — hooks intercept dangerous commands *before* execution (not prompts — hard enforcement at the harness level). Blocks `git push --force`, `rm -rf /`, `npm publish`, direct commits to `main`, and more\n- **Session Continuity** — handoff system tells the next session exactly where work stopped, what's in progress, and what to do next\n- **Automatic Learning** — background auditor extracts memories, decisions, and safety rules from session transcripts when you close the window\n- **19 MCP Tools** — `axme_context`, `axme_save_decision`, `axme_save_memory`, `axme_safety`, `axme_backlog`, and more\n\nYou keep using Claude Code exactly as before. AXME Code works transparently in the background.\n\n## Install\n\nFrom terminal:\n```bash\nclaude plugin install axme-code\n```\n\nFrom Claude Code interactive CLI:\n```\n/plugin install axme-code\n```\n\n## First Use\n\nAfter installing the plugin, open any project in Claude Code. On the first session:\n\n1. The plugin creates a `CLAUDE.md` section that instructs the agent to call `axme_context`\n2. The agent will say the project is not initialized and run `axme-code setup --plugin`\n3. Setup scans your project (uses an LLM to analyze your stack, extract decisions, detect safety patterns) and creates the `.axme-code/` knowledge base\n4. From the next session onward, the agent automatically loads the full knowledge base\n\n**Setup requires a Claude subscription** (the LLM scan uses Claude Sonnet). If the scan fails, a deterministic fallback creates a minimal knowledge base from file system inspection.\n\n## How It Works\n\n### Hooks (automatic, transparent)\n\n| Hook | When | What |\n|------|------|------|\n| **SessionStart** | Session opens | Installs SDK dependency, creates CLAUDE.md if missing |\n| **PreToolUse** | Before every tool call | Checks safety rules, blocks violations |\n| **PostToolUse** | After Edit/Write | Tracks which files changed |\n| **SessionEnd** | Session closes | Spawns background auditor to extract knowledge |\n\n### MCP Server\n\nProvides 19 tools for reading and writing the knowledge base. The agent calls these tools naturally during work — no manual intervention needed.\n\n### Storage\n\nAll data lives in `.axme-code/` in your project root (gitignored automatically):\n\n```\n.axme-code/\n  oracle/           # stack.md, structure.md, patterns.md, glossary.md\n  decisions/        # Architectural decisions with enforce levels\n  memory/           # Learned mistakes (feedback/) and validated patterns (patterns/)\n  safety/rules.yaml # git + bash + filesystem guardrails\n  backlog/          # Persistent cross-session task tracking\n  sessions/         # Session metadata\n  plans/            # Handoff files for session continuity\n  worklog.jsonl     # Structured event log\n```\n\nHuman-readable markdown and YAML. No database, no external dependencies.\n\n## Requirements\n\n- [Claude Code](https://docs.anthropic.com/en/docs/claude-code) (CLI or VS Code extension)\n- Node.js >= 20\n- Claude subscription (for LLM-powered setup and session auditor)\n\n## Alpha Status\n\nThis is alpha software (v0.x.y). What that means:\n\n- **Breaking changes** may happen in minor version bumps (0.x.0 -> 0.x+1.0)\n- **Storage format** may change between versions (`.axme-code/` is local and gitignored, so upgrades won't affect your repo)\n- **Bug reports are welcome** — [open an issue](https://github.com/AxmeAI/axme-code/issues)\n- **Feedback** — we're actively iterating based on user feedback. Tell us what works and what doesn't\n\n## Auto-Sync\n\nThis repo is auto-synced from the main [axme-code](https://github.com/AxmeAI/axme-code) repository on each release. **Do not edit files here directly** — changes will be overwritten.\n\n## License\n\n[MIT](LICENSE)\n\n---\n\n[Website](https://code.axme.ai) · [Main Repo](https://github.com/AxmeAI/axme-code) · [Issues](https://github.com/AxmeAI/axme-code/issues) · contact@axme.ai\n",
  "bytes": 4906,
  "sha": "afa903b202146bcdbad1218c3f9e42956e6eb50be87529f5521e354c06d7a54d",
  "repo_slug": "axmeai/axme-code-plugin",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_axmeai_axme_code_plugin_axme_code_3b28e00f/readme"
}