Get started
Product note / documentation
Build trusted knowledge systems
A self-hostable, knowledge-graph-grounded assistant and durable agent platform. A markdown knowledge base, an LLM-extracted entity graph with provenance, GraphRAG chat with citations, a provider-agnostic agent harness with guardrails and human-in-the-loop, evals with golden sets and drift detection, and client-facing visualizations — all multi-tenant and runnable on a laptop or a single VPS.
What eli.ai is
eli.ai turns a pile of markdown into a queryable, grounded knowledge system. Documents are chunked and indexed for full-text and vector search the moment you save them; embeddings and LLM entity extraction run asynchronously in the background. The result is a connected entity graph with provenance that chat and agents reason over — answering with citations that link straight back to the source chunk.
It is built for two audiences at once: personal local use — one command on a laptop, works offline — and consulting engagements, with per-client workspaces, interview-driven KB building, and gap analysis that tells you which business unit to interview next.
Explore the docs
The public docs are organized into five tabs. Start wherever your question lives.
Run eli.ai locally in one command, understand the architecture and the ideas it is built on, then get your first cited answer.
Inside — Core principles · Architecture · Quickstart · Use cases · Examples
The eight capability systems — Intake through Ports — and how a single document flows through all of them, each with concept guides and how-tos.
Inside — The capability map · concept guides · task-oriented guides
The scoped /api/v1 surface — authentication, structured query, agents & runs, manifest & governance — plus the native MCP server, OpenAPI reference, and playground.
Inside — Auth · endpoints · MCP server · OpenAPI · playground
Multi-tenant isolation and security, multi-provider AI, cost and scheduling, PDF deliverables, and deploying eli.ai on your own box.
Inside — Tenancy · providers · cost & schedules · deployment · backup
How governed knowledge, cited answers, live data, team assets, skills, and capability-locked agents come together in the product experience.
Inside — Workspace overview · end-to-end journeys
The eight capability systems
The Product tab is organized around eight named systems — one job each, carrying a document from a raw source to a governed, cited answer. Each chip links to that system's reference.
- IntakeConnect & ingest
Every source becomes one clean, deduplicated corpus.
- AtlasThe semantic layer
A canonical knowledge graph over a ten-type ontology.
- WarrantGovern & certify
Authority, lifecycle, and gated change on every fact.
- LineageProvenance & traceability
Per-document provenance and downstream impact in one trail.
- ConduitThe live-data layer
Governed live queries bound to graph entities.
- LensRetrieve & answer
Retrieval and an answer policy that knows when to abstain.
- CrucibleEvaluate & measure
Golden sets, retrieval and abstention metrics, governed judges.
- PortsProgrammatic surfaces
Every capability, exposed to your agents and tools.
Every step of the grounded answer path, listed in order below, followed by all eight systems and every documented handoff between them. On large screens the plate also draws those systems and handoffs as a map, with this path picked out.
- Intake
Normalize source material and preserve source permissions.
- Atlas
Connect evidence through typed entities and relations.
- Warrant
Apply certification, lifecycle, and content ACL policy.
- Lens
Retrieve, rerank, answer, clarify, or abstain.
- Ports
Deliver the result to a person or programmatic caller.
- Intake
- Atlas
- Warrant
- Lineage
- Conduit
- Lens
- Crucible
- Ports
All system handoffs
- Intake → Atlas Normalized documents feed typed entity and relation extraction with sentence evidence.
- Atlas → Warrant Graph concepts receive authority tiers, lifecycle state, ownership, and governed change.
- Atlas → Lens Graph neighborhoods join lexical and vector retrieval before reranking.
- Warrant → Lens Content ACLs and certification state constrain what the answer pipeline may state.
- Conduit → Lens Entitled named queries contribute live rows and auditable Data Citations.
- Lens → Ports Grounded results reach the workspace, REST clients, MCP hosts, and durable agents.
- Lens → Crucible The real answer pipeline runs against golden sets and measured quality gates.
- Crucible → Warrant Significance-gated evaluation results govern consequential configuration changes.
- Lens → Lineage Answer claims, source revisions, data calls, and policy verdicts remain traceable.
- Lineage → Intake Reverse provenance resolves a result back to the exact ingested source revision.
Reach it programmatically
The workspace is one of three ways in. Applications and agents run the same governed pipeline under the same workspace-scoped key, and get back the same cited answer.
Executable surface
The same answer over HTTP or MCP
Applications send POST /api/v1/query. MCP clients mount the workspace at /api/mcp and call the eli_query tool.
REST
one call in, a dual-cited answer out
curl -L -X POST https://eli.ai/api/v1/query \
-H "Authorization: Bearer $ELI_KEY" \
-H "Content-Type: application/json" \
-d '{"question":"Is anything stuck on the Order Platform right now?","includeData":true}'MCP
mount the workspace over streamable HTTP
claude mcp add --transport http eli https://eli.ai/api/mcp \ --header "Authorization: Bearer eli_sk_..."
Start here
Three short reads that orient you before you dive into a tab.
Who it's for
Personal knowledge, self-hosted
One command on a laptop. Your vault stays on disk, git-able and Obsidian-compatible. Point it at a local Ollama model and everything runs offline — no API cost, no data leaving the box.
Consulting engagements
Per-client workspaces with hard tenant isolation, interview-driven KB building, agents that traverse the graph with approval-gated tools, and exportable deliverables with coverage and gap analysis for the room.
Ready to run it?