Skip to content

WarmHub

Knowledge platform for AI agents — knowledge your agents gain persists, compounds, and stays yours.

New to WarmHub? Learn what it is and how it works.
For AI agents: these docs as a structured map at /llms.txt

MCP, CLI, or SDK — pick what fits your workflow.

Connect MCP

Connect Claude Code, Cursor, or VS Code to WarmHub in three steps. Connect an agent →

TypeScript SDK

Build applications with the typed TypeScript client. Use the SDK →

CLI Quickstart

Install the wh CLI, point at a repo, and run your first query in minutes. Use the CLI →

Get repo access, then browse your knowledge and live activity in the web app.

Authentication

Personal access tokens, scopes, and how to get access to a repo. Get access →

Define shapes, write versioned data, and package it for reuse.

Data Modeling

Structure knowledge with shapes (schemas), things (entities), assertions (versioned claims about things), and collections (groups). Model your data →

Writes

Change data with writes — add, revise, and retract operations that land as new versions, with full history on every thing. Submit a write →

Components

Package shapes and logic into reusable components — author, test, and publish them through a manifest. Build components →

Query it, react to changes, and combine sources with confidence.

Queries

Read the latest state of your data, filter it, and get reactive views that update live. Query data →

Veritas

Combine assertions from multiple sources into confidence-scored views, with end-to-end provenance on every result. Explore Veritas →

Every surface for building on WarmHub — agent integration, SDK, CLI, and HTTP API.

Agent Integration

Connect via the Model Context Protocol (MCP) server, bootstrap an agent session with warmhub_capabilities or wh prime, and follow the recommended pattern for agent-driven knowledge workflows. Connect agents → · Agent context (wh prime) →

HTTP API

Use WarmHub over plain HTTP — reads, action callbacks, and REST endpoints. API reference →