What is Lito Graph?

Lito Graph transforms your Markdown documentation into a typed knowledge & workflow graph that AI agents can reason over, plan with, and act on — not just search through.

The Problem

Today, agents consume docs via RAG: chunk text, embed it, retrieve the top-k similar chunks. This has fundamental limitations:

  • No relationships — RAG finds the “Create User” page but doesn’t know it acts on the User entity or is step 2 of an onboarding workflow
  • No planning — an agent can’t answer “what’s the full sequence to onboard a customer?” because steps are scattered across pages
  • No safety — nothing tells the agent “this workflow requires human approval” or “never delete workspaces during onboarding”

The Solution

Lito Graph adds a compilation layer between your docs and agents:

Markdown + Frontmatter  →  Lito Graph Compiler  →  graph.json  →  MCP Server  →  AI Agents

Your docs stay human-readable. Light annotations in frontmatter create machine-traversable structure.

Key Features

  • 5 node types — DocNode, ConceptNode, ApiNode, WorkflowNode, StepNode
  • 14 edge types — structural, semantic, capability, and procedural relationships
  • 7 MCP tools — list, get, traverse, search, workflow, entity APIs, stats
  • Backward compatible — existing Lito api: "GET /users" format auto-converts
  • Deterministic — same docs always produce the same graph

Next Steps

1 min read