Skip to documentation
eli.aidocs

Get started

Examples

Five first steps you can copy, paste, and run against a live eli.ai instance — each one a real request, not pseudocode. Swap https://eli.ai for your instance, export your key as $ELI_KEY, and every one returns exactly the shape shown. When you want the full story behind a call — every field, every error, every knob — follow its full guide link.

Before you start

Every example authenticates with Authorization: Bearer $ELI_KEY and needs the scope shown on its badge — scope always comes from the key, never a URL or body. A workspace key (eli_sk_…) carries its workspace implicitly; a user key (eli_uk_…) additionally needs an X-Workspace-Id header. No key yet? Getting started with the API mints one in a minute.

1 · Ask your first grounded question

Bearer · agents:run

One POST in, a structured, dual-cited answer out: answer with inline [Sn] markers, the answer decomposed into atomic claims, the sources registry each marker resolves to, and a dataCalls registry (empty here — add includeData: true to ground in live rows as [Dn]).

POST /api/v1/querybash
curl -s -L -X POST https://eli.ai/api/v1/query \
  -H "Authorization: Bearer $ELI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"question":"How do we handle stuck orders on the Order Platform?"}'

Full guide: Search & ask →

2 · Start an agent run

Bearer · agents:run

Agents are addressed by slug. With stream: false the request blocks and returns the final RunResult as JSON — the simplest integration. Flip to stream: true (the default) for a Server-Sent Events stream of every step ending in a terminal event.

POST /api/v1/agents/{slug}/runsbash
curl -s -L -X POST https://eli.ai/api/v1/agents/budget-analyst/runs \
  -H "Authorization: Bearer $ELI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input":"What changed in the FY26 budget?","stream":false}'

Full guide: Run agents →

3 · Generate a report snapshot

Bearer · reports:write

One POST freezes the live state of the knowledge base into an immutable, client-ready snapshot. It comes back complete — 201 with the coverage matrix and narrative summary already computed, since reports are assembled from the persisted graph without a model call. Fetch it again any time with GET /api/v1/reports/{id}, or list metadata rows with GET /api/v1/reports.

POST /api/v1/reportsbash
curl -s -L -X POST https://eli.ai/api/v1/reports \
  -H "Authorization: Bearer $ELI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"title":"Northwind — Q3 engagement report"}'

Full guide: Build reports →

4 · Connect a content source

Bearer · kb:write

Register an external source as a connector — Confluence, SharePoint, or Slack. The credential is encrypted at write and never returned (reads show only hasCredential); the source starts active and does nothing until you enqueue its first sync.

POST /api/v1/connectorsbash
curl -s -L -X POST https://eli.ai/api/v1/connectors \
  -H "Authorization: Bearer $ELI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "confluence",
    "config": { "baseUrl": "https://acme.atlassian.net/wiki", "spaceKeys": ["ENG"] },
    "credential": { "email": "svc-eli@acme.com", "apiToken": "<atlassian-api-token>" }
  }'

Full guide: Connect Confluence, SharePoint & Slack →

5 · Call eli.ai from an MCP client

Bearer · eli_sk_

eli.ai is itself an MCP server at POST /api/mcp. Point Claude or any MCP client at your workspace with the same eli_sk_ key and it gets the KB, graph, and governed data layer as native, read-only tools — kb_search, kb_entity_lookup, and eli_query (the same pipeline as POST /api/v1/query), among others.

claude mcp addbash
claude mcp add --transport http eli https://eli.ai/api/mcp \
  --header "Authorization: Bearer eli_sk_..."

Then ask naturally — "Using the eli tools, what do we know about the Order Platform, and is anything stuck right now?" — and the client chains the tools for you. The endpoint-by-endpoint reference is in MCP server.

Full guide: MCP quickstart →

Where to next

These are the opening moves. Browse every operation — and fire real requests — in the API console, or start from the API overview for the full map of task guides.