Let Claude Code pick the right loop.

Hook up one MCP server and Claude Code can browse the library and choose the right loop for the job.

browse→pick→run→stop with evidence

claude code

# 30-second install: one file, stdlib only, read-only
$ curl -fsSL ailooplibrary.com/mcp/server.py -o ~/.ai-loop-library/server.py
$ claude mcp add ai-loop-library -- python3 ~/.ai-loop-library/server.py
$ claude
> pick a loop to speed up our CI, then run it
⏺ pick_loop_for_goal("speed up CI") → CI Optimization
⏺ render_run_protocol("ci-optimization") → 8 rounds max, verify every round▊

Value prop

Why this is an MCP server and not a docs search

Retrieval is not the hard part of agent loops; packaging and judgment are. The server does the packaging; judgment stays with the calling agent, which knows the operator's repo, data, and constraints:

Step	Tool	What you get
1. Browse	`browse_catalog()`	The whole library as a ~2k-token digest: id, one-line use_when, verifier strength.
2. Pick	`pick_loop_for_goal(goal)`	Lexically ranked shortlist with an honest confidence signal, never a single blind verdict.
3. Run	`render_run_protocol(id, goal)`	Executable markdown: done contract, one change per round, the same verification every round, state-file skeleton, stop conditions, budget, approval boundary, paste-ready prompt.
4. Lint / design	`critique_loop`, `design_loop`	Lint a loop design against the anti-pattern rubric, or scaffold a new spec when nothing fits.

Every loop has a stop condition and a budget. If your task has no verifiable end state, the library says so instead of pretending.

30 seconds

Install for Claude Code

One Python file, standard library only, Python 3.9+. Read-only: no write tools, no auth, no shell execution.

mkdir -p ~/.ai-loop-library
curl -fsSL https://ailooplibrary.com/mcp/server.py -o ~/.ai-loop-library/server.py
claude mcp add ai-loop-library -- python3 ~/.ai-loop-library/server.py

Verify before trusting it: run the offline self-test.

python3 ~/.ai-loop-library/server.py --self-test   # 43 checks, no network

Third-party checks: listed in awesome-mcp-servers under Knowledge & Memory, and graded on Glama for tool definition quality and server coherence.

Install for Cursor and other MCP clients

Add to ~/.cursor/mcp.json (or your client's MCP config). Point AI_LOOP_LIBRARY_CATALOG_PATH at a downloaded catalog.json to run fully offline.

{
  "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"
      }
    }
  }
}

Use it without MCP

Option one: the Claude Code skill, which prefers MCP tools when connected and falls back to fetching the catalog over HTTPS.

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

Option two: zero install. The catalog is public JSON; paste this into any tool-using 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.

Full install guide CLAUDE.md drop-in MCP server on GitHub

Tool reference

Tool	Arguments	Returns
`browse_catalog`	category?	~2k-token digest of every loop; the agent judges it directly
`search_loops`	query, category?, limit?=8	Ranked loops with id, title, category, url, one-line why-matched
`get_loop`	id_or_slug	Full loop spec + canonical URL, verifier strength, loop kind
`pick_loop_for_goal`	goal, constraints?, limit?=5	Ranked shortlist + confidence signal; the calling agent makes the final call
`render_run_protocol`	id_or_slug, goal?, risk_posture?, kind?=auto\|session\|scheduled-tick, max_rounds?, max_minutes?	Executable protocol + state-file skeleton + paste-ready prompt
`critique_loop`	loop_description	0–10 lint against the anti-pattern rubric, with per-check fixes
`design_loop`	goal, constraints?, cadence?, context?	Draft loop spec: verifier suggestion, MVL check, nearest catalog loops
`list_categories`	—	Category counts with library filter URLs
`catalog_stats`	—	Loop count, featured loops, last_updated, catalog source

Resources: ailooplibrary://catalog and ailooplibrary://loop/{id}. Full server docs at /mcp/README.md.

Demo

What picking looks like

Ask for an outcome, get a loop with reasons, then render it and go.

> pick_loop_for_goal("speed up CI")

{
  "goal": "speed up CI",
  "confidence": "high",
  "confidence_note": "top match covers most goal terms and leads clearly",
  "shortlist": [
    {
      "id": "ci-optimization",
      "title": "CI Optimization",
      "category": "Engineering",
      "url": "https://ailooplibrary.com/loops/ci-optimization/",
      "verifier_type": "deterministic-leaning",
      "why_matched": "title matches ci; tags matches ci; use when matches ci"
    },
    {
      "id": "test-flake-stabilizer",
      "title": "Test Flake Stabilizer",
      "verifier_type": "deterministic-leaning",
      "why_matched": "tags matches ci; goal terms (ci, speed) map to the Engineering category"
    }
  ],
  "your_call": "A ranked shortlist, not a verdict. You have operator context this server lacks; pick accordingly, or pick none."
}

The heuristic is documented in the server source and labels itself in every response; a 20-query golden eval holds it to a ≥85% top-3 hit rate. The calling agent always makes the final call.

What's in the library

68bounded loop patterns

10categories

21engineering loops

7Paul originals

Every loop ships with a trigger, objective, allowed actions, verification, stop condition, budget, and approval boundary: the seven parts that keep an agent from spending your tokens looking busy.

Browse the library Loop linter Loop template Loop engineering