Documentation/MCP

MCP Integration

Connect any MCP-compatible AI assistant to search and explore AI safety problems directly.

What is MCP?

The Model Context Protocol (MCP) is an open standard that allows AI assistants to connect to external tools and data sources. Instead of copy-pasting information, your AI assistant can directly search our database, fetch problem details, and help you find relevant work.

No API key required

Read-only access works immediately

Real-time data

Always queries the latest problems

Works with any MCP client

Standard protocol, broad compatibility

Setup

Add the AI Safety Marketplace to your MCP client configuration:

{
  "mcpServers": {
    "ai-safety-marketplace": {
      "url": "https://safetybounty.com/api/mcp"
    }
  }
}

Endpoint: https://safetybounty.com/api/mcp

Available Tools

search_problems

Search for AI safety problems by keyword, tags, difficulty, or status.

Parameters

Name	Type	Description
`query`	string	Free text search
`tags`	string[]	Filter by tag names
`difficulty`	enum	ENTRY, INTERMEDIATE, ADVANCED, EXPERT
`status`	enum	OPEN, IN_PROGRESS, SOLVED, DEPRECATED
`limit`	number	Max results (1-50, default 10)

get_problem

Get full details for a specific problem including description, bounties, resources, and metadata.

Parameters

Name	Type	Description
`id`	string	Problem ID (required)

list_tags

List all available tags grouped by category with problem counts. Useful for discovering what topics are available before searching.

No parameters required.

Example Prompts

Once connected, try asking your AI assistant:

>Search for AI safety problems about interpretability

>Find beginner-friendly problems I can work on

>What alignment problems are currently open?

>Tell me more about problem [ID]

>What categories of problems are available?

>Find problems related to deception detection

Need REST API access?

For programmatic access, custom integrations, or higher rate limits, use our REST API.

REST API Docs Get API Key