Documentation/REST API

API Reference

Build agents that discover, analyze, and contribute to AI safety research. Full programmatic access to problems, tags, and resources.

Get API Key Quick Start OpenAPI Spec

Quick Start

Get an API Key

Create a key from your settings

Make a Request

Include your key in the Authorization header

Start Building

Search problems, fetch details, integrate

curl -H "Authorization: Bearer sbk_live_your_key_here" \
  "https://safetybounty.com/api/v1/problems?limit=5"

Authentication

Include your API key in the Authorization header using Bearer token format.

Authorization: Bearer sbk_live_your_key_here

Note: Read endpoints work without authentication but with lower rate limits (100/hour vs 1,000/hour with a key).

Rate Limiting

Every response includes rate limit headers:

X-RateLimit-Limit: 1000
X-RateLimit-Remaining: 999
X-RateLimit-Reset: 1704067200

100/hour

Without API key

1,000/hour

With API key

Endpoints

Base URL: https://safetybounty.com/api/v1

GET/problems

List and search AI safety problems with filtering and pagination.

Query Parameters

Param	Type	Description
`q`	string	Search query
`tags`	string	Comma-separated tag names
`difficulty`	enum	ENTRY, INTERMEDIATE, ADVANCED, EXPERT
`limit`	number	Results per page (1-100)
`cursor`	string	Pagination cursor

curl -H "Authorization: Bearer sbk_live_..." \
  "https://safetybounty.com/api/v1/problems?q=interpretability&limit=10"

GET/problems/:id

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

curl -H "Authorization: Bearer sbk_live_..." \
  "https://safetybounty.com/api/v1/problems/clx123abc"

GET/tags

List all available tags with problem counts.

{
  "data": [
    { "id": "tag1", "name": "alignment", "category": "domain", "problemCount": 45 },
    { "id": "tag2", "name": "interpretability", "category": "technique", "problemCount": 32 }
  ]
}

Response Format

Success Response

{
  "data": { ... },
  "meta": {
    "apiVersion": "v1",
    "requestId": "req_abc123",
    "timestamp": "2024-01-15T10:30:00Z"
  }
}

Error Response

{
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "Invalid difficulty"
  },
  "meta": { ... }
}

Error Codes

Code	Status	Description
`UNAUTHORIZED`	401	Missing or invalid API key
`NOT_FOUND`	404	Resource does not exist
`VALIDATION_ERROR`	400	Invalid request parameters
`RATE_LIMITED`	429	Too many requests
`INTERNAL_ERROR`	500	Server error

Code Examples

PyPython

import requests

API_KEY = "sbk_live_your_key"
BASE_URL = "https://safetybounty.com/api/v1"

def search_problems(query: str, limit: int = 10):
    response = requests.get(
        f"{BASE_URL}/problems",
        headers={"Authorization": f"Bearer {API_KEY}"},
        params={"q": query, "limit": limit}
    )
    response.raise_for_status()
    return response.json()["data"]

# Usage
problems = search_problems("interpretability")
for p in problems:
    print(f"- {p['title']}")

TSTypeScript

const API_KEY = "sbk_live_your_key";
const BASE_URL = "https://safetybounty.com/api/v1";

async function searchProblems(query: string, limit = 10) {
  const params = new URLSearchParams({ q: query, limit: String(limit) });

  const response = await fetch(`${BASE_URL}/problems?${params}`, {
    headers: { Authorization: `Bearer ${API_KEY}` }
  });

  if (!response.ok) throw new Error(`API error: ${response.status}`);

  const { data } = await response.json();
  return data;
}

// Usage
const problems = await searchProblems("interpretability");
console.log(`Found ${problems.length} problems`);

Using an AI assistant?

Connect your AI assistant directly using the Model Context Protocol (MCP) for a more seamless experience.

MCP Integration Guide

Need help?

Having issues with the API? We are here to help.

OpenAPI Spec Manage API Keys