Brought to you by Volant Partners. The lab publishes open tools and research for graph-native AI work; Volant Partners helps teams apply them in production.
Vellis · Apache-licensed knowledge infrastructure

Structured knowledge. Controlled change. Durable systems.

Vellis is an open-source context graph engine for AI agents: a typed, locally owned system for knowledge that must be shared, changed, and recovered without losing meaning. MCP-capable agents operate against the same explicit graph contract.

  • Clone the repo and run the graph engine locally
  • Connect Claude, Codex, and other MCP-capable tools
  • Model schemas, provenance, and decisions as a typed graph
  • Query, write, validate, migrate, snapshot, and replay
One typed graph that agents can read, write, and validate against.
How it works — from clone to shared knowledge
1

Clone

Pull the repo and run the graph engine on your machine.

2

Model

Define typed nodes, links, and constraints for the work.

3

Connect

Point MCP-capable agents at one shared, graph-backed knowledge model.

4

Iterate

Read, write, migrate, and validate changes with schema checks.

5

Reuse

Each pass leaves explicit, durable knowledge for the next one.

Builder · production champion

Build with the open graph

Vellis starts as a repo you can run and shape yourself. Capable teams can carry it a long way; when the graph needs operating controls, audit paths, approval workflows, or production hardening, Volant Partners can help.

Stay close

Follow the lab work

Read the Thesis, browse Perspectives, or send a note if you want to shape the next open project.

Browse Perspectives