khive

A research knowledge graph runtime for agents that need structure: typed substrates, closed taxonomies, and a verb-consolidated MCP surface.

Vector search finds similar text. A knowledge graph finds structure: lineages, dependencies, contradictions, gaps. khive gives an agent a typed, queryable graph that grows as it works.

crates.io · GitHub · License: Apache 2.0

For AI agents

  • /llms.txt: short project summary + a linked index of every doc page
  • /llms-full.txt: every doc page, concatenated, in one fetch
  • /md/*.md: raw, unconverted markdown for each page (also linked from the bottom of every page)

    Documentation

  • Getting Started: Install, connect an MCP client, and run your first khive session.
  • Request and the Verb-Dispatch DSL: The single request tool: input forms, parallel batches, chains, and the result envelope.
  • Knowledge Graph Modeling: Entity kinds, edge relations, and modeling patterns for the KG substrate.
  • Memory and Recall: Episodic vs semantic memory, salience, and decay-weighted recall.
  • Search and Retrieval: FTS5, vector search, hybrid fusion, and reranking.
  • GTD Task Management: Task lifecycle, priorities, and dependency edges.
  • Communication and Email: Actor-addressed messaging, threading, and the optional email channel.
  • Agent Sessions and Data Ingest: Session record verbs and provider transcript mirror ingest.
  • Specialized Packs: Niche packs beyond the production set, starting with the formal-math pack.
  • Prompt Cookbook: Ready-to-use verb patterns for common agent workflows.
  • Query Cookbook: Question classes mapped to the right verb, plus verified GQL/SPARQL idioms and known gaps.
  • Tips and Tricks: Query craft, DSL round-trips, param gotchas, and troubleshooting for the request tool.
  • Proof Graph Case Study: Mathlib ingested as a 320K-entity, 4.4M-edge khive graph: ingestion path and traversal at scale.
  • API Reference: Full verb catalog for all 12 production packs: params, DSL examples, response envelope.

Demos

Runnable transcripts, captured against a scratch database, in the repo’s demos/ directory:

  • research-ingest: create entities, link them, search, and traverse the graph
  • gtd-memory: task lifecycle and salience-weighted memory recall

Install 

cargo install kkernel

kkernel is the single shipped binary; kkernel mcp serves the MCP request surface. Full install and MCP client setup: Getting Started.

Reference

Raw markdown for this page: /md/index.md