{
  "markdown": "# Gyst\n\n**Give your AI agent your team's memory. In 90 seconds.**\n\n```bash\nnpm install -g gyst-mcp && gyst init\n```\n\n---\n\n## The problem\n\nAI coding agents are fast — but stateless. Every session your agent starts fresh: no memory of why the auth service keeps timing out, no knowledge of the deploy-on-Fridays rule, no record of decisions your team made last month. Your agent figures things out and then forgets them. The team never gets smarter.\n\n---\n\n## The solution\n\n- **Self-populating context layer** — mined automatically from your git history, code comments, markdown docs, and session transcripts\n- **Every session starts ready** — your team's rules, conventions, and past decisions are injected before the first prompt\n- **Works with every major agent** — Claude Code, Cursor, Codex CLI, Windsurf, Gemini CLI\n\n---\n\n## 90-second quickstart\n\n```bash\nnpm install -g gyst-mcp && gyst init\n```\n\n**Example output:**\n\n```\n┌─────────────────────────────────────────────────────┐\n│  gyst init                                          │\n├─────────────────────────────────────────────────────┤\n│  ✓ Database initialized      .gyst/wiki.db          │\n│  ✓ MCP server registered     Claude Code            │\n│  ✓ MCP server registered     Cursor                 │\n│  ✓ Hooks installed           post-commit            │\n│  ✓ Scanning codebase ...     42 files               │\n│  ✓ Loaded conventions        8 entries              │\n│  ✓ Loaded decisions          3 entries              │\n│  ✓ Ghost knowledge generated 5 entries              │\n├─────────────────────────────────────────────────────┤\n│  Restart your AI tool. Context will inject on the   │\n│  next session start.                                │\n└─────────────────────────────────────────────────────┘\n```\n\nRestart your AI tool. That's it.\n\n---\n\n## What your agent gets\n\n| Context type | What it is | Example |\n|---|---|---|\n| **Ghost Knowledge** | High-confidence facts about your codebase the agent must know | \"Never deploy on Fridays — batch job runs at 22:00 UTC and will conflict\" |\n| **Conventions** | Coding standards scoped to files and directories | \"All API responses use the `ApiResponse<T>` envelope shape\" |\n| **Decisions** | Architecture choices and the reasons behind them | \"Switched from Prisma to raw SQL — N+1 queries were killing dashboard load time\" |\n| **Error Patterns** | Known failure signatures with their fixes | \"`SQLITE_BUSY` on parallel writes — use WAL mode + retry with backoff\" |\n\n---\n\n## Supported agents\n\n| Agent | Auto-detected | Context injection | Hook coverage |\n|---|:---:|:---:|:---:|\n| Claude Code | ✓ | ✓ | 12 hooks |\n| Cursor | ✓ | ✓ | 4 hooks |\n| Codex CLI | ✓ | ✓ | 4 hooks |\n| Windsurf | ✓ | ✓ | 4 hooks |\n| Gemini CLI | ✓ | ✓ | 4 hooks |\n\n---\n\n## How it works\n\n**Mining:** `gyst init` scans your codebase and git history to build a context layer — zero config, zero manual writing. It extracts conventions from your source files, decisions from commit messages and ADRs, error patterns from `TODO`/`FIXME` comments, and structural knowledge from the files your team touches most. The post-commit hook keeps it current automatically after every commit.\n\n**Injection:** At every session start, Gyst injects your team's ghost knowledge and top conventions into the agent's context window. Subagents get the same injection. When a session ends, anything the agent learned gets distilled back into the context layer, so the next session starts even stronger.\n\n---\n\n## Learn more\n\n- [Technical reference](docs/ADVANCED.md) — MCP tool API, CLI flags, hook system, search pipeline, configuration\n- [3-minute demo script](docs/DEMO_SCRIPT.md) — step-by-step pitch walkthrough\n\n---\n\n## Dashboard\n\n```bash\ngyst dashboard   # opens at localhost:3579\n```\n\nBrowse all context entries, inspect the knowledge graph, review low-confidence entries, and track context economics (leverage ratio, zero-result rate, intent mix).\n\n---\n\n## Team mode\n\n### Git-sync (zero infrastructure)\n\nThe context layer lives in your repo. `git pull` syncs it automatically via the `post-merge` hook.\n\n### Shared HTTP server\n\n```bash\n# Admin — set up once\ngyst create team \"Acme Engineering\"\nGYST_API_KEY=gyst_admin_... gyst team invite\ngyst serve --http --port 3456\n\n# Each developer — one command\ngyst join gyst_invite_abc123... \"Alice\" --server http://your-host:3456\n```\n\nEvery developer's AI tools automatically point at the shared context layer. Knowledge grows as the team grows.\n\n---\n\n## License\n\nMIT — see [LICENSE](LICENSE).\n",
  "bytes": 4507,
  "sha": "d61a280ededfd98a47c8fc4f1fc8e83be04e20745b38bafcf6ecebd1bc2d3c86",
  "repo_slug": "chaydavs/gyst",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_chaydavs_gyst_7ae578ad/readme"
}