khive
A research knowledge graph runtime. Typed substrates, closed taxonomies, and a verb-consolidated
MCP surface — built for agents that need structure, not just vectors.
Vector search finds similar text. A knowledge graph finds structure — lineages, dependencies,
contradictions, gaps. khive gives your research agent a typed, queryable graph that grows as it
works: read a paper and entities + edges fall out; make a connection and it's traversable
immediately; come back next session and the graph remembers what you built.
No Neo4j. No SPARQL endpoint to deploy. SQLite on disk, MCP over stdio, cargo test in 4 seconds.
What you get
| Capability | How |
|---|
| Typed entities | 6 closed kinds: concept, document, dataset, project, person, org |
| Typed edges | 13 closed relations in 6 categories (structure, derivation, dependency, impl, lateral, annotation) |
| Typed notes | 5 closed kinds: observation, insight, question, decision, reference |
| Hybrid search | FTS5 trigram (CJK-safe) + sqlite-vec embeddings + reciprocal rank fusion |
| Graph traversal | BFS with depth/direction/relation filters, bidirectional shortest path |
| GQL + SPARQL queries | Parse to SQL, run against the same SQLite backend |
| Salience-weighted notes | Notes carry importance scores; search ranks by semantic relevance × salience |
| Cross-substrate links | Notes annotate entities (and vice versa) via the same edge system |
| Soft delete + supersede | History-preserving: old records stay, newer ones supersede via graph edges |
| Namespace isolation | Tenant scoping on every operation — share one DB, isolate many agents |
The three substrates
Everything in khive is one of three things (ADR-004):
| Substrate | What it is | Mutability | Example |
|---|
| Entity | A graph node with typed edges | Mutable + soft-delete | LoRA (concept), arxiv:2106.09685 (document) |
| Note | A temporal observation about the world | Mutable + soft-delete | "FlashAttention gains scale with seq len, not batch" |
| Event | An audit log entry | Immutable | create(kind="entity", ...) was called at T |
Entities are things. Notes are what you think about things. Events are what happened.
The MCP verb surface
11 tools in v0.1, verb-shaped (ADR-023):
CRUD: create get list update delete merge
Graph: link traverse neighbors query
Search: search
create, list, search take kind=entity|note (or kind=edge for list).
get, update, delete, merge are UUID-only — they auto-detect the record type.
Agents reach khive via MCP stdio — Python, TypeScript, Rust, or any MCP-compatible client.
No language SDK to learn.
Architecture
┌──────────────────────────────────────────────────────────────┐
│ crates/khive-mcp — Rust binary (stdio MCP server) │
│ The only binary you need. Embeds khive-runtime. │
└──────────────────────────────────────────────────────────────┘
↕ in-process
┌──────────────────────────────────────────────────────────────┐
│ crates/ — Rust storage + query core │
│ khive-types, khive-score, khive-storage, khive-db, │
│ khive-query, khive-runtime │
└──────────────────────────────────────────────────────────────┘
Native sqlite-vec for vector search, FTS5 trigram tokenization (CJK-safe), concurrent connection
pooling, memory-local graph traversal. One binary, one DB file, no services to run.
HTTP gateway, CLI, and visual frontend are planned for future releases.
Crates
