Find the best AI agent for your job

Type a job and get the AI agents our playbooks use for it, ranked by how many sourced playbooks back each one. Every result links to the playbook, its prompt and the public accounts behind it. It runs in your browser and makes no AI call.

What job do you want an agent to do?

Chapter

Agent you already use

Your skill level

Risk you accept

Results come only from our playbooks. Each one cites public accounts of people doing the job. The job you type goes into the page link so you can share the result.

The agents our 54 playbooks use most. Type a job to narrow it down.

1
Claude Code
Written into 38 of 54 matching playbooks, the first choice in 23. 186 independent sources behind them. 45 public accounts name Claude Code.
- Match receipts to bank transactions and flag VAT or sales tax questions for your accountant
  Corroboratedintermediatehigh riskYour agent plan. Accounting software and accountant fees are separate.
- Build a small internal automation with n8n's AI builder and test it before it goes live
  Corroboratedintermediatemedium riskn8n's docs say each builder message uses one AI credit. Zapier charges per task or activity.
- Find the questions AI assistants answer without naming your brand, and plan content to fix it
  Corroboratedintermediatelow risk
- And 35 more playbooks below.
Start here: the prompt in the first playbook. It takes 1-3 hours for the first month, less after that.
You approve: Approve every proposed category and every match before anything is entered into the ledger.
Watch for: The agent invents a transaction or pulls in an unrelated one so the numbers add up. Read the lesson
2
Claude
Written into 23 of 54 matching playbooks, the first choice in 6. 116 independent sources behind them. 16 public accounts name Claude.
- Match receipts to bank transactions and flag VAT or sales tax questions for your accountant
  Corroboratedintermediatehigh riskYour agent plan. Accounting software and accountant fees are separate.
- Find the questions AI assistants answer without naming your brand, and plan content to fix it
  Corroboratedintermediatelow risk
- Turn a long podcast or video into a blog post, a timestamped clips list and social posts
  Corroboratedbeginnerlow risk
- And 20 more playbooks below.
Start here: the prompt in the first playbook. It takes 1-3 hours for the first month, less after that.
You approve: Approve every proposed category and every match before anything is entered into the ledger.
Watch for: The agent invents a transaction or pulls in an unrelated one so the numbers add up. Read the lesson
3
ChatGPT
Written into 22 of 54 matching playbooks, the first choice in 8. 106 independent sources behind them. 27 public accounts name ChatGPT.
- Match receipts to bank transactions and flag VAT or sales tax questions for your accountant
  Corroboratedintermediatehigh riskYour agent plan. Accounting software and accountant fees are separate.
- Find the questions AI assistants answer without naming your brand, and plan content to fix it
  Corroboratedintermediatelow risk
- Turn a long podcast or video into a blog post, a timestamped clips list and social posts
  Corroboratedbeginnerlow risk
- And 19 more playbooks below.
Start here: the prompt in the first playbook. It takes 1-3 hours for the first month, less after that.
You approve: Approve every proposed category and every match before anything is entered into the ledger.
Watch for: The agent invents a transaction or pulls in an unrelated one so the numbers add up. Read the lesson
4
Claude Cowork
Written into 17 of 54 matching playbooks, the first choice in 6. 87 independent sources behind them. 14 public accounts name Claude Cowork.
- Brief yourself on every external meeting guest before your day starts
  Corroboratedintermediatemedium riskNeeds an agent plan with calendar and email access, and a web search tool for public research
- Research a list of target accounts and draft first emails for you to review and send
  Corroboratedintermediatehigh risk
- Triage your support inbox and draft replies that a person approves before sending
  Corroboratedintermediatehigh risk
- And 14 more playbooks below.
Start here: the prompt in the first playbook. It takes 30-60 minutes to set up, then a few minutes each morning.
You approve: Approve which calendars and which kinds of meetings the agent may read.
Watch for: Text in a calendar invite takes over the assistant. Researchers showed this against Gemini and against ChatGPT with connected tools. Read the lesson
5
Codex
Written into 12 of 54 matching playbooks. 44 independent sources behind them. 1 public account names Codex.
- Build a personal knowledge base wiki that a coding agent maintains from your notes
  Corroboratedintermediatelow riskIncluded in a Claude, ChatGPT or Gemini plan that includes the coding agent. Obsidian is free for personal use.
- Turn small tracker issues into reviewed pull requests with cloud coding agents
  Corroboratedintermediatemedium riskNeeds a paid agent plan (Copilot, Claude, ChatGPT with Codex, Jules, Cursor or Devin). GitHub Actions minutes may apply.
- Map an unfamiliar codebase into a checked architecture document before you change it
  Corroboratedbeginnerlow riskIncluded in your coding agent plan. DeepWiki is free for public GitHub repositories.
- And 9 more playbooks below.
Start here: the prompt in the first playbook. It takes 1-2 hours to set up, then a few minutes per source.
You approve: Approve the schema file before the first ingest.
Watch for: A bulk ingest or reorganisation deletes or overwrites files. Read the lesson

How to use it

1
Type the job
Use a few plain words, such as inbox, travel planning or coding, or pick an example. Narrow it by chapter, your skill level or the risk you accept.
2
Read the ranking
Each agent shows how many matching playbooks use it and the sources behind them. You also see where to start, the step you approve and the failure to watch for.
3
Open the playbook
Click a playbook for the full prompt, steps and evidence. Copy the link to share this exact result with a person or an agent.

Questions

Which AI agent is best for coding?

In our coding playbooks, Claude Code is the first choice most often. Codex, GitHub Copilot, Cursor and Gemini CLI are also written into them. Type coding above to see each playbook, its sources and the failure to expect.

Which AI agent should I use?

Start from the job, not the agent. Type the job here and the finder lists the agents our playbooks use for it, ranked by how many sourced playbooks back each one. If you already pay for an agent, pick it under Agent you already use to see if a playbook covers it.

How does the agent finder rank agents?

It counts the matching playbooks written for each agent. Ties go to the agent that is the first choice in more of them, then to more independent public sources, then to the higher trust level. No AI model picks the order and no vendor pays for a place.

What can AI agents do for me?

Our playbooks cover ten chapters: money, life admin, shopping, travel, operations, growth, research, content, data and software. Leave the job box empty to see the agents used most across all of them, or pick a chapter to browse one area.

What if no agent fits my job?

The finder says so. It recommends an agent only when a playbook backs it, so a job with no sources returns no agents. If you did that job with an agent, add the playbook and it shows up here once an editor approves it.

Claude Code vs Codex vs Cursor: which should I pick?

Type the names, for example claude code vs codex vs cursor, and the finder compares them across our playbooks. Add a job, such as claude code vs codex for seo, to compare them on that job only. Each count links to the playbooks behind it.

Sources

Every recommendation comes from the playbooks and agent profiles in the handbook. Each playbook lists its own public sources. We built this index on 2026-10-30.

Agents

Claude Code Claude ChatGPT Claude Cowork Codex ChatGPT agent

Lessons

Agents act before you approve, even when you told them to ask first

More free tools

All free tools