# Agentboard Skills

> Agentboard Skills is a free MCP server for browser automation. Find a verified web workflow and replay it with zero model calls, with failures shown. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** API
- **Website:** https://agentboard-olive-pi.vercel.app/skills
- **Repository:** https://github.com/jai-rathore/agentboard
- **MCP Registry name:** io.github.jai-rathore/agentboard-skills
- **Categories:** browser automation, AI models and LLM tooling
- **Listing updated:** 2026-11-14

## About

Agentboard sells avoided work. A painful web task is done once by an agent; later, other agents look up the recorded workflow and replay it with new inputs, get a signed, verified result, and in the best case spend no model calls. Before running anything, an agent can call check_workflow, which is free, read-only and never executes; if nothing verified matches, it returns an explicit empty match list rather than a guess.

The README's headline benchmark is the US Treasury OFAC sanctions search, a form with no keyless API. A cold agent took 7 model calls, about $0.029 and 29 seconds, while a second agent that found the skill via find_skill replayed it with 0 model calls and finished in 3.8 seconds. Each discovery result carries a trust card built from signed records: last verified run, 7 and 30-day pass and fail counts with the latest failure reason, and median time and cost.

The maintainers are candid that most of the catalogue is easy lookups, and describe a candidate skill they declined to publish because every run would need human approval. Featured skills cover OFAC search and Florida and New Jersey licence checks. The project is built on Meta models, and a keyless install is one claude mcp add command.

## Features

- **Free workflow check:** check_workflow reports a match and trust card without executing anything.
- **Replay with no model calls:** Reuses a recorded trace so repeat runs can cost nothing in model usage.
- **Signed results:** Each run returns a verified, signed outcome.
- **Trust cards:** Show recent pass and fail counts, failure reasons, time and cost.
- **Verified no-hit:** Reports a verified empty search result using the site's own wording.
- **Failures shown:** Failures stay visible in the catalogue.

## Good for

- Look up a name on a government sanctions list without driving the form yourself
- Check a Florida or New Jersey professional licence status
- Avoid paying a model to redo a known web task
- Compare cold and reused agent cost and time

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http agentboard-skills https://agentboard-api.onrender.com/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://agentboard-api.onrender.com/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "agentboard-skills": {
      "url": "https://agentboard-api.onrender.com/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "agentboard-skills": {
      "type": "http",
      "url": "https://agentboard-api.onrender.com/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "agentboard-skills": {
      "serverUrl": "https://agentboard-api.onrender.com/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "agentboard-skills": {
      "type": "http",
      "url": "https://agentboard-api.onrender.com/mcp"
    }
  }
}
```

## Tools

- `find_skill`: Finds a reusable verified workflow for a task.
- `check_workflow`: Free read-only check for a verified workflow match.

## FAQ

### Do I need an API key?

The README gives a keyless install: claude mcp add --transport http agentboard https://agentboard-api.onrender.com/mcp.

### What does check_workflow do?

It reports whether a verified workflow exists, with executed false, 0 model calls and $0 cost.

### What if no skill matches?

It returns matches: [] instead of the nearest guess.

### Which models does it use?

The README says Meta models only.

### Is Agentboard Skills free?

Yes. Agentboard Skills is free to use.

### How do I add Agentboard Skills to Claude Code?

Run this command in your terminal: claude mcp add --transport http agentboard-skills https://agentboard-api.onrender.com/mcp

### What are the best alternatives to Agentboard Skills?

The closest alternatives to Agentboard Skills are Browser Use, Scrapling MCP Server, Claude Code Ultimate Guide, Oya Browser and Mantis.

## Alternatives

- [Browser Use](https://ai-powered.com/mcp/browser-use): Browser Use is a freemium MCP server for browser automation.
- [Scrapling MCP Server](https://ai-powered.com/mcp/scrapling-mcp-server): Scrapling MCP Server is an open-source MCP server for web search and scraping.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Oya Browser](https://ai-powered.com/mcp/oya-browser): Oya Browser is a freemium MCP server for browser automation.
- [Mantis](https://ai-powered.com/mcp/mantis-2): Mantis is an open-source MCP server for browser automation.
- [signafAIde](https://ai-powered.com/mcp/signafaide): signafAIde is a freemium MCP server for security.
- [Toolkien](https://ai-powered.com/mcp/toolkien): Toolkien is an open-source MCP server for ai models and llm tooling.
- [Lugh](https://ai-powered.com/mcp/lugh): Lugh is an open-source MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/agentboard-skills
