What is WarmHub?
WarmHub is knowledge infrastructure for AI agents — a platform purpose-built so that what agents learn persists, compounds, and stays yours. Think of it as what GitHub did for code, applied to knowledge: versioned, attributed, auditable, and shareable.
Why WarmHub?
Section titled “Why WarmHub?”Traditional databases store rows. WarmHub stores claims about the world — versioned, attributed, and queryable. Every mutation is an immutable commit. Every entity carries its full history. Every assertion knows what it’s about.
What sets WarmHub apart is its native support for beliefs — assertions carry confidence, evidence, and attribution, making it possible to model what an agent thinks it knows, not just what it recorded. Multiple agents can hold different beliefs about the same thing, and the system preserves all of them with full provenance. Knowledge compounds across sessions, agents, and teams — each agent builds on what previous agents learned.
This makes WarmHub ideal for:
- AI agent memory — Agents write observations, beliefs, and decisions as structured assertions. Knowledge persists across sessions, so agents don’t start from zero every time.
- Multi-agent coordination — Multiple agents working on the same problem write to a shared repo. Each commit is attributed, so you can trace who said what and when.
- Calibrated uncertainty — Model confidence, evidence, and competing perspectives as first-class data. Assertions from different agents coexist — you can compare, reconcile, or let them evolve independently.
- Auditable knowledge bases — Every version of every entity is preserved. You can always ask “what did we know at time T?” or “who changed this and why?”
Key Design Principles
Section titled “Key Design Principles”Everything is versioned. Things, assertions, and shapes all produce new versions on every change. Old versions are never deleted.
Commits are atomic. A commit can contain multiple operations (adds and revises) that succeed or fail together. No partial writes.
Names are stable, not permanent. Names identify things and appear in wrefs (references like Shape/name), but they can be changed. When a thing is renamed, existing references automatically resolve to the new name. Things are linked by identity, not by name — the name is a human-readable label on top of that link.
Agents are first-class. WarmHub exposes an MCP server so AI agents can read and write data using the Model Context Protocol. The wh prime command gives agents a complete context dump in one call.
How to Get Started
Section titled “How to Get Started”Start with Core Concepts to understand the mental model — orgs, repos, shapes, things, assertions, and commits. Then pick the interface that fits your use case:
| SDK | CLI | MCP | |
|---|---|---|---|
| Best for | TypeScript apps, custom agents, programmatic access | Terminal exploration, shell scripts, quick operations | AI agents with MCP-compatible clients |
| Type safety | Full TypeScript types | JSON output via --json | Tool schemas |
| Setup | npm install + client constructor | Configure .npmrc + npm install -g @warmhub/cli@next | Configure MCP endpoint |
| Write pattern | client.commit.apply(...) or CommitBuilder | wh commit create ... | warmhub_commit_apply tool |
| Read pattern | client.thing.head(...) | wh thing head | warmhub_thing_head tool |
| Real-time | AdvancedClient.subscribe() | wh thing head --live | Not available |
| Get started | SDK Quickstart | CLI Quickstart | MCP Quickstart |
All three interfaces share the same backend and concepts — shapes, things, assertions, commits, and wrefs.