Use with AI
Put Agentailor to work with your AI assistant.
Give your coding agent practical guidance, let your assistant search the articles and repo docs, or try a prompt with an assistant that can read the web.
Choose your starting point
| What you want | Start here |
|---|---|
| Design tools, write agent instructions, or build eval cases. | Install skills |
| Search Agentailor articles and read public repo docs as you work. | Connect via MCP |
| Explore the guides with an assistant that can fetch web pages. | Try a prompt |
Install skills
Skills give your coding agent reusable guidance for a specific task. MCP gives it access to articles and repo docs. You can use both together.
- Tool Design
Clearer tool definitions, focused outputs, and a checklist to review them.
- Agent Prompt Engineering
Instructions with clearer decisions, boundaries, and stopping conditions.
- Agent Eval Cases
Focused cases with appropriate graders, based on failures you have actually seen.
Connect via MCP
For ongoing access to Agentailor articles and public repo docs, connect your assistant over the Model Context Protocol. It can search for relevant guides and read them as needed.
Use a client that can run a local MCP server, with Node.js 20 or newer and npx available. Public content needs no API key. Merge the entry into your existing config if you already have other servers.
Run from your project directory. Check the connection with /mcp inside Claude Code.
claude mcp add agentailor -- npx -y @agentailor/mcp
Check the connection
Once your client shows the server as connected, try this. You should get article links and a repo list.
Use the Agentailor MCP tools to find articles about tool design, read the most relevant one, and list the public Agentailor repos. Include source links and tell me if any tool call fails.
More setup options and troubleshooting: MCP repo README.
Hosted MCP is planned. For now, use the local setup on this page; a hosted endpoint URL is not available yet.
Try a prompt
No Agentailor installation needed. Use an assistant that can fetch web pages. Start with the index so it can choose relevant articles before reading them in full.
Read https://blog.agentailor.com/llms.txt. I'm new to building AI agents. Choose and read three articles that will help me get started, explain the order, and suggest one small project I can build next. Link to each source.
More to try
Read https://blog.agentailor.com/llms.txt and the relevant articles and open-source repo docs. Design a structured four-week curriculum for a developer new to AI agents. Order the readings from fundamentals to production, link to the sources, and tell me what to build each week.
Read the content directly
- llms.txt — start with an index
- Discover what to read, then fetch the pages relevant to your task.
- Markdown pages
- These pages and each individual path have a Markdown version. For blog articles, follow the Markdown URLs in the blog index.
- llms-full.txt — expanded index
- An expanded content index with article summaries, metadata, and series context. Follow the linked Markdown URLs to read the articles in full.
curl https://agentailor.com/skills.md
Tools for in-browser agents (WebMCP)
If you browse agentailor.com with an AI agent that drives the page, this site hands it real tools instead of making it read the layout. No setup on your side: the page registers them itself.
- find_your_path
- Runs the fitting and recommends which of the four ways to build an agent suits you. The agent can run it silently, visit the fitting temporarily, or leave it visible.
- get_path
- Returns one path in full and can optionally visit or leave its detail page visible.
- search_glossary
- Searches the glossary and can run silently, visit the filtered glossary temporarily, or leave it visible.
All three are read-only and marked as such, so an agent can run them without asking you to confirm. WebMCP is an experimental standard: support today is the Chrome origin trial and Edge. The site detects it and does nothing at all in browsers without it.
//New to building agents? Start with what an agent is and the four paths.