# Coworker

> Coworker is an MCP server for agent tooling. Enterprise context layer connecting 50+ work apps to AI assistants, by demo only. It is available on API, Web and Self-hosted.

- **Kind:** MCP server
- **Pricing:** Contact sales
- **Platforms:** API, Web, Self-hosted
- **Website:** https://coworker.ai/contact
- **MCP Registry name:** ai.coworker/mcp
- **Categories:** [agent tooling](https://ai-powered.com/categories/agent-tooling)
- **Listing updated:** 2026-11-18

## About

The Coworker MCP server exposes an organization's connected work tools to MCP-compatible AI clients such as Claude, ChatGPT and Cursor, so those assistants can read company context and run agents. The context comes from OM2, an organizational memory that the vendor says builds a permission-aware graph from the 50+ connectors a company uses, with each query respecting the access people already have in those systems.

Beyond the MCP layer, the vendor's plans bundle its own apps for Work, Code, Agents and Meet, along with model routing across 13 open and closed models. Closed models are passed through at provider prices, and the company states all models run on US infrastructure. Deployment can be in Coworker's cloud or a customer's VPC, and the registry entry lists no installation transports or packages.

This is aimed at enterprise IT and operations teams rather than individuals. Access requires booking a demo, pricing is scoped per organization, and the pricing page describes implementation services and a dedicated customer success manager as part of the offer.

## Features

- **Permission-aware organizational memory:** OM2 builds a memory of company work that the vendor says respects existing user permissions in connected tools.
- **50+ app connectors:** Connectors include Slack, Gmail, Salesforce, Jira, Notion, BigQuery and Snowflake, with custom connectors available on Enterprise.
- **MCP access for AI clients:** Any MCP-compatible tool can read company context, run agents and write back to connected systems.
- **Model routing across providers:** Choose among 13 open and closed models from five providers for each task.
- **Enterprise security controls:** Offers SSO via SAML 2.0 and listed providers, SOC 2 Type II, GDPR and CASA Tier 2 compliance claims.

## Good for

- Give Claude or ChatGPT access to internal Slack, CRM and ticketing context through MCP
- Automate cross-tool workflows with background agents triggered by connected apps
- Summarize and sync meeting notes from Zoom, Meet and Teams into CRM and docs
- Reduce token spend on AI assistants by supplying company context instead of raw retrieval

## FAQ

### Is there a self-serve plan?

No. The pricing page describes an Enterprise plan tailored to the organization, and the contact page directs visitors to book a demo.

### Which models are included?

The pricing page lists 13 open and closed models from Anthropic, OpenAI, Google, Moonshot and Z.ai, enabled for all customers.

### Where does the product run?

The contact page says it runs in Coworker's cloud or the customer's VPC, and the pricing page says models run on US infrastructure.

### Is the data used for training?

The FAQ states that customer data is encrypted and that models are never trained on customer data.

### What are the best alternatives to Coworker?

The closest alternatives to Coworker are Guildworks and AIOS.

## Where this comes from

- Description: written from the project's website and its MCP Registry entry.

## Alternatives

- [Guildworks](https://ai-powered.com/mcp/guildworks): Guildworks is an MCP server for agent tooling.
- [AIOS](https://ai-powered.com/mcp/aios): AIOS is an MCP server for agent tooling.

---
Source: https://ai-powered.com/mcp/coworker
