Your AI agents don’t know how your company works.

World Model builds a persistent, structured understanding of your organization. Your agents query it in real time instead of starting from zero every session.

Data sources

SlackLinearGitHubGoogle DriveMeetingsJirasoonNotionsoonConfluencesoonEmailsoon

World Model

Entities, relationships, ownership, decisions, history. Continuously synthesized into a structured organizational model.

Agents

Claude CodeCursorClaudeCustom agentsAny MCP client

More than connecting tools to agents.

MCP lets agents call APIs. That solves plumbing. It does not solve understanding. World Model is the layer that sits between raw data and the agents that need to reason about it.

Inference over time

World Model doesn’t just snapshot your tools. It watches how things change: who picked up a project, when a decision shifted, how teams reorganized. Agents get context that spans weeks and months, not just the current state.

Organizational structure

Reporting lines, team boundaries, ownership, areas of expertise. World Model understands the shape of your company so agents can route questions, assign work, and respect boundaries without being told every time.

Cross-system reasoning

A Slack thread, a Linear ticket, a PR, and a design doc might all be about the same decision. World Model connects them into a coherent picture. Agents see relationships, not silos.

Synthesized, not searched

This isn’t RAG over your company wiki. World Model continuously processes and structures information into entities and relationships. When an agent asks “who owns billing?”, the answer is already modeled.


What it looks like

An agent working in your codebase asks World Model a question. This is the kind of answer it gets back.

> who owns the billing service?

Billing is owned by the Payments team (Sarah Chen, lead).

Recent: migrated to Stripe Connect (merged 3 days ago)

Related: LINEAR-482 (open), 4 Slack threads this week

Active PR: acme/billing#217 (review requested from @jpark)

On-call: David Kim (rotation ends Friday)

> what decisions led to the Stripe migration?

The team evaluated Stripe Connect vs. building in-house in Q4.

Decision doc: "Payments platform strategy" (Google Docs, Nov 12)

Key factor: multi-currency support needed by March for EU launch

Decided by: Sarah Chen, Marcus Webb, approved by Jamie Liu

Discussion: #payments-platform (Slack, 38 messages over 2 weeks)


How to set it up

1

Connect your data sources

Point World Model at the tools your team uses. Run worldmodel connect slack to add a source. Each connector syncs incrementally after the initial import.

2

Let it build

World Model processes your data into structured entities: people, teams, projects, decisions, dependencies. The first build takes minutes. After that, it updates continuously.

3

Add the MCP server to your agent

Add World Model to your mcp_servers config in Claude Code, Cursor, or any MCP-compatible client. Agents can immediately start querying organizational context.

4

Deploy however you want

Run it locally for personal use, or deploy to shared infrastructure with Docker Compose so your whole team’s agents share the same model. Your data never leaves your control.


Get started

Initialize World Model in your project. It sets up the MCP server and connects to your first data source.

# Install and initialize
npx @worldmodel-dev/core init

# Or run with Docker Compose
curl -sSL https://worldmodel.dev/install.sh | sh

Early development

World Model is in early development. We’re building in public and looking for contributors and early adopters. Star the repo to follow along, or join the community to help shape the project.