Weftd API & MCP server
Weftd's own AI Presence can be reached by programs as well as people: an HTTP API that answers questions about Weftd, and an MCP server for AI assistants. Both are public and need no key.
At a glance
| Surface | URL | Description |
|---|---|---|
| HTTP API | https://weftd.weftd.space/api/v1 | Ask a question in natural language, get JSON or plain text. weftd.com/api/v1 redirects here; /api is an alias of the current version. |
| OpenAPI | https://www.weftd.com/openapi.json | OpenAPI 3.1 description of the HTTP API. |
| MCP server | https://weftd.weftd.space/mcp | Streamable HTTP endpoint for AI assistants (Claude, ChatGPT, Cursor…). |
Authentication
None for Weftd's public space: call the API and connect the MCP server without a key or an account.
The same API exists for every AI Presence built on Weftd. A private space requires an API key, created in that space's Developer API settings and sent as Authorization: Bearer <key> or X-API-Key: <key>.
Endpoint: GET /api/v1
Ask Weftd a question. POST /api/v1 takes the same fields as a JSON body.
| Parameter | Required | Values | Description |
|---|---|---|---|
| query | yes | text | The question, in natural language. |
| mode | no | context (default), search | context: an AI-written answer with its sources. search: the knowledge entries ranked by relevance, without AI. |
| type | no | txt (default), json | Plain text or structured JSON. |
| limit | no | 1–50, default 4 | Maximum number of knowledge entries used or returned. |
Example requests
curl "https://weftd.weftd.space/api/v1?query=How%20much%20does%20Weftd%20cost%3F&mode=search&type=json&limit=2"curl -X POST https://weftd.weftd.space/api/v1 \
-H "Content-Type: application/json" \
-d '{"query": "What is an AI Presence?", "mode": "context", "type": "json"}'{
"results": [
{ "id": "…", "title": "Weftd Plans and Pricing: Free to Enterprise", "snippet": "Weftd starts for free…", "relevance": 0.011 }
],
"total": 1,
"query": "How much does Weftd cost?"
}Errors and rate limits
Errors are JSON, with an error field and, when useful, a code and a message: 400 for a missing query or a limit outside 1–50, 403 for a private space without a valid key, 404 for an unknown path, 429 when the rate limit is reached.
Requests are rate-limited per client IP. Every response carries the IETF RateLimit header fields: RateLimit-Policy gives the quota and its window, RateLimit the requests left and the seconds until the window resets. A 429 also carries Retry-After: wait that many seconds before retrying.
Versioning and deprecation
The version is in the path: /api/v1. Changes that keep existing requests working (a new optional parameter, a new response field) ship within v1. A breaking change ships as a new path, /api/v2, and v1 keeps answering.
A version being retired says so on every response: a Deprecation header (RFC 9745), a Sunset header (RFC 8594) with its retirement date, at least six months ahead, and a Link header with rel="successor-version" pointing at its replacement. The OpenAPI document states the same policy in x-api-lifecycle.
Connect the MCP server
Add https://weftd.weftd.space/mcp to your assistant as a custom connector. It lets the assistant search Weftd's knowledge and send a message to the team.
claude mcp add --transport http weftd https://weftd.weftd.space/mcp{
"mcpServers": {
"weftd": { "type": "http", "url": "https://weftd.weftd.space/mcp" }
}
}Machine-readable files
- /openapi.json — The OpenAPI 3.1 document of the HTTP API.
- /.well-known/api-catalog — API catalog (RFC 9727) pointing at that document.
- /.well-known/mcp/server-card.json — Identity, version and endpoint of the MCP server.
- /.well-known/ard.json — Agentic Resource Discovery catalog.
- /llms.txt — Index of the site for language models, with a Markdown version of every page.