Loading the catalog…
Loading the catalog…
About
Don't just read about AI agents. Run them. AI Playbook Gems, at aiplaybookgems.com, is a directory of playbooks you can reproduce: prompt, steps, difficulty, a cost bucket, and a time-saved line that is an estimate. Social-post roundups are useful raw material. This site is the layer that says what to paste and whether anyone has checked it.
Seed entries name agents with public docs or product pages, including Claude Code, Cursor, Codex CLI, ChatGPT's computer-use tool, Grok Bot, OpenClaw, n8n, Manus, Devin, GitHub Copilot, Perplexity, CrewAI, LangGraph, Zapier Agents, and Lindy. The ChatGPT Agent page cites the OpenAI API computer-use guide and says so in the source note. It is not a customer story.
Windsurf is not a separate section. On 15 Oct 2026, windsurf.com and the Windsurf getting-started docs redirected to Devin Desktop, so a Windsurf chapter would be a stale name.
Cost buckets are filters: free means no paid plan is required to attempt the playbook, low means one individual plan is usually enough, medium means a team product or steady usage, high means a dedicated platform. Prices change. The pages tell you to check the vendor.
Rankings on "best agent for" pages follow catalog coverage. Ranked by how many matching playbooks are in this catalog, then by how many of those are marked easy, then by lower cost buckets. This is coverage inside AI Playbook Gems, not a benchmark and not a vote. The arrow on a card starts from an editorial signal stored with the playbook. A vote or a save stays in this browser. It is not a global leaderboard.
Submissions land in a JSON endpoint. With SUBMIT_WEBHOOK_URL they are forwarded. Otherwise the API accepts the payload and, when SUBMIT_MAILTO is set, offers a mailto draft. Nothing is published until a JSON file is added under content/usecases.
A prompt you can paste, the setup steps around it, who it is for, and a cost bucket. It is a playbook you can execute, not a testimonial.
The seed catalog was written for this site from public product documentation. Nobody re-ran each playbook and recorded the result. Verified stays false until that happens.
Only when a page was fetched and it is a fair citation for the product capability. If the URL is unknown, the field is empty. We do not invent links.
Ranked by how many matching playbooks are in this catalog, then by how many of those are marked easy, then by lower cost buckets. This is coverage inside AI Playbook Gems, not a benchmark and not a vote.