Bluefield
Web data API for AI agents
Bluefield is a web data API that exposes search, scrape, map, crawl, and extract as tools your agents call over REST or MCP, returning markdown and structured JSON ready for retrieval pipelines.
What is a web data API for agents?
A web data API for agents wraps live web access as callable tools so language models can fetch and reason over current page content instead of stale training data.
- MCP server
- Model Context Protocol integration that registers Bluefield tools in agent configs.
- Structured extraction
- LLM backed field extraction against a JSON schema you provide.
- changeTracking
- Diff against the previous scrape in git diff or JSON patch form.
Bluefield capabilities
| Capability | Detail |
|---|---|
| MCP server | npx @bluefields/mcp-server with search, scrape, map, crawl, extract |
| Markdown output | Clean text ready for chunking and embedding |
| POST /extract | Bulk structured extraction per URL |
| Sessions | Named cookie and header bundles for authenticated pages |
| Webhooks | Async job completion and watch mode delivery |
Frequently asked questions
- How do agents call Bluefield?
- Via the REST API with any HTTP client or through the official MCP server registered in your agent configuration.
- What output format works best for RAG?
- Markdown from POST /scrape is the default. Use POST /extract when you need a fixed JSON schema.
- Can agents monitor pages for changes?
- Yes. POST /subscriptions enables watch mode with webhook delivery on meaningful content changes.