# AI Loop Library agent pack — install

Three ways in, from most to least wired. Nothing here needs npm, an account, or a key.

## 1. MCP server (Claude Code, Cursor, any MCP client)

One Python file, stdlib only, read-only. Requires Python 3.9+.

```bash
mkdir -p ~/.ai-loop-library
curl -fsSL https://ailooplibrary.com/mcp/server.py -o ~/.ai-loop-library/server.py
python3 ~/.ai-loop-library/server.py --self-test    # 21 offline checks should pass
```

**Claude Code:**

```bash
claude mcp add ai-loop-library -- python3 ~/.ai-loop-library/server.py
```

**Cursor** (`~/.cursor/mcp.json`) **or any MCP client:**

```json
{
  "mcpServers": {
    "ai-loop-library": {
      "command": "python3",
      "args": ["/Users/you/.ai-loop-library/server.py"],
      "env": {
        "AI_LOOP_LIBRARY_CATALOG_URL": "https://ailooplibrary.com/catalog.json"
      }
    }
  }
}
```

Offline / air-gapped? Point `AI_LOOP_LIBRARY_CATALOG_PATH` at a downloaded copy of
`catalog.json` instead.

## 2. Claude Code skill (no MCP)

```bash
mkdir -p .claude/skills/ai-loop-library
curl -fsSL https://ailooplibrary.com/agent-pack/claude-code/SKILL.md \
  -o .claude/skills/ai-loop-library/SKILL.md
```

The skill fetches `https://ailooplibrary.com/catalog.json` over HTTPS when no MCP server
is connected — same loops, same run protocol.

## 3. Zero install

Paste the short drop-in from
[`agent-pack/claude-code/claude-md-snippet.md`](https://ailooplibrary.com/agent-pack/claude-code/claude-md-snippet.md)
into your project `CLAUDE.md`, or just tell your agent:

```
Fetch https://ailooplibrary.com/catalog.json, pick the loop that fits this goal, and run it
bounded: one change per round, the same verification every round, state in
docs/loops/<id>/, stop on verifier pass / budget / blocker / approval boundary.
```

## Future

`npx` / `uvx` packaging is on the list; for now the single-file curl install is the
supported path. The server file is versioned with the site — re-run the curl command to
update.