Developers & AI agents

Build on the cited knowledge layer.

Give your app or agent grounded, cited answers over your documents — through a native MCP server, OAuth 2.0, and a REST API. Every answer carries citations and returns grounded:false instead of fabricating.

Native MCP server

Point any MCP client at one URL and call answer, search, ingest_url and more. Streamable-HTTP transport, OAuth-secured.

OAuth 2.0 + PKCE

Let users securely connect their own corpus to your product. Authorization code with PKCE, refresh tokens, fine-grained scopes, one-click revocation.

REST API v1

Full programmatic control — upload, ingest-by-URL (bulk), extract, template-deploy, export, webhooks. A stable, versioned, OpenAPI-described contract.

Use our RAG as your brain

Connect an agent in two minutes.

Drop this into your MCP client config. Your agent can now ground every response in the connected document corpus — with citations it can show the user, and an honest “not in the documents” when the answer isn’t there.

  • Citations on every answer (documentId · page · exact quote)
  • grounded:false instead of hallucinating
  • Hybrid retrieval: semantic + lexical + knowledge graph
  • Scoped to exactly the projects you authorize
~/.mcp/config.json
{
  "mcpServers": {
    "intellimento": {
      "type": "streamable-http",
      "url": "https://intellimento.com/api/v1/mcp",
      "authorization": "oauth2"
    }
  }
}

Quickstart

Call the MCP tools over JSON-RPC, or use the plain REST API. Authenticate with an OAuth token (for user-connected data) or a public API key (for your own corpus).

Ask a grounded question (MCP)
# 1) Get a token (OAuth 2.0 client-credentials-style via your app),
#    or use a public API key from the developer console.
# 2) Ask a grounded, cited question over your corpus:

curl -s https://intellimento.com/api/v1/mcp \
  -H "Authorization: Bearer $INTELLIMENTO_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0", "id": 1, "method": "tools/call",
    "params": {
      "name": "answer",
      "arguments": { "question": "What is the warranty period in the MSA?" }
    }
  }'

# → returns { answer, citations:[{documentId, page, quote}], grounded: true|false }
#   grounded:false means "not in your documents" — never a fabricated answer.
Upload + extract (REST)
# 1) Create a project first (required — a document is always uploaded into one)
curl -s -X POST https://intellimento.com/api/v1/projects \
  -H "X-API-Key: $INTELLIMENTO_KEY" -H "Content-Type: application/json" \
  -d '{ "name": "Contracts" }'          # -> { "id": "<projectId>", ... }

# 2) Upload the file WITH projectId (mandatory multipart form field; missing -> 400)
curl -s https://intellimento.com/api/v1/documents/upload \
  -H "X-API-Key: $INTELLIMENTO_KEY" \
  -F "file=@contract.pdf" -F "projectId=<projectId>"

# That's it — the document is PROCESSED AUTOMATICALLY (parsed, indexed, extracted).
# No trigger to call. Poll GET https://intellimento.com/api/v1/documents/<id>; then answer/search find it.

What you can build on

Everything the platform does is reachable programmatically. Bold capabilities, grounded in what actually ships.

Grounded, cited answers
RAG that always attributes to source passages, or abstains.
Hybrid search
Semantic + lexical + graph ranking over your corpus.
Knowledge graph
Entities and relationships extracted and queryable.
Structured extraction
Any document → typed fields with per-field confidence.
Claim verification
Per-claim grounding scores on generated content.
Webhooks
React to extraction.completed and document events.

Self-configuring discovery

An agent can configure itself against IntelliMento from the domain alone. These endpoints are live, CORS-open, and absolute-URL correct in every environment.

Ready to build?

Create a free developer account, mint an API key, and connect your first agent.

© IntelliMento · Built for regulated data