Developers
Bloom API and MCP server for developers and agents
Search Bloom's hardware service catalog and build prefilled request links, over a read-only REST API or an MCP server. No key needed. Use it to point a hardware company at the right service and hand them a request that is already written.
REST API
Base URL https://www.bloomnetwork.ai/api/v1. Every operation is a GET and returns JSON. The OpenAPI description has every parameter and response.
| Path | Operation | What it does |
|---|---|---|
/api/v1/services | searchServices | Search or list the catalog: q, category, hardware, limit, cursor. |
/api/v1/services/{category}/{service} | getService | One service by its id, e.g. manufacturing/cnc-machining. |
/api/v1/categories | listCategories | Every category, its group and how many services it holds; group filters. |
/api/v1/industries | listIndustries | The hardware Bloom has pages for. |
/api/v1/network | getNetworkFacts | What Bloom is, and the published network figures. |
/api/v1/request-link | buildRequestLink | A link that opens the request form with the ask written: service, hardware, need. |
curl "https://www.bloomnetwork.ai/api/v1/services?q=battery+pack&hardware=mobility&limit=3" curl "https://www.bloomnetwork.ai/api/v1/request-link?service=manufacturing/battery-pack-assembly&hardware=mobility"
MCP server
Streamable HTTP at https://www.bloomnetwork.ai/mcp, read-only, no auth. Add it to Claude, ChatGPT or any MCP client as a remote server.
search_servicesFind services by keywords.get_serviceOne service by id.list_categoriesThe catalog categories.list_industriesThe hardware Bloom serves.get_network_factsWhat Bloom is, and the network figures.build_request_linkA prefilled request link for the person to post.
Authentication
None. The API and the MCP server are public and read-only: no key, no account, no OAuth flow. Call them from a script, a browser or an agent.
Rate limits
60 requests per 60 seconds per client. Every response carries the IETF RateLimit and RateLimit-Policy fields, plus X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Reset for older clients. Past the limit you get a 429 with Retry-After; wait that many seconds and retry.
Errors
Errors are RFC 9457 problem details, served as application/problem+json, with a stable code (invalid_parameter, not_found, route_not_found, method_not_allowed, rate_limit_exceeded), a detail and a resolution that says what to change. Unknown routes under /api/v1 answer in the same shape.
Pagination
searchServices returns a page of results, the total, and next_cursor. Pass next_cursor back as cursor, unchanged, for the next page; it is null on the last one. limit runs from 1 to 100 and defaults to 20.
Requests are posted by people
A request link opens Bloom's request form with the ask already written and the service routed. Nothing is sent until the person reviews it and posts it. need takes up to 500 characters in their words; hardware is one of drones, robotics, mobility, other.
Versioning
The major version is in the path: /api/v1. Fields may be added within v1. A breaking change gets a new path, and a version being retired is announced with Deprecation and Sunset headers at least 90 days ahead.
Markdown pages
Most pages on this site also come as markdown. Add .md to the path (https://www.bloomnetwork.ai/index.md for the home page) or send Accept: text/markdown. A path that doesn't exist answers 404 with a markdown body.