Coworker AI pricingCustomized around your enterprise
Scoped to the connectors, people and deployment you need. Access to the Coworker MCP, Work, Code, Agents, and Meet are all included.
Last updated
Enterprise
Tailored to your organization
Custom deployment, dedicated support, and implementation services, scoped with your team before you commit.
What you get
- OM2 Organizational Memory
- Model routing across 13 models from Anthropic, OpenAI, Google, Moonshot and Z.ai
- Coworker MCP and native apps: Work, Code, Agents, Meet
- 50+ connectors, plus custom
- Custom deployment, implementation services and a dedicated CSM
- SSO and custom SLAs
Everything, in one plan
Open and closed models are enabled for all customers. Closed: Anthropic (Fable 5.1, Opus 5.5, Sonnet 5.5, Haiku 5.5, Claude Opus 4.8), OpenAI (GPT 6.1 Sol, GPT 6 Astra, GPT 5.6 Terra, GPT 6 Luna), Google (Gemini Argon 4, Gemini 3.8 Flash). Open-weight: Moonshot (Kimi K3), Z.ai (GLM 5.3 Flash).
Every model runs on US infrastructure, including the open-weight ones, under SOC 2 Type II, GDPR, and CASA Tier 2.
Closed models are passed through with no markup. You pay what the provider charges.
OM2 is the organizational memory that learns continuously from your company's work, so your AI stops re-reading and starts thinking.
Learn about Organizational MemoryChoose from 13 open and closed models across five providers, picking the one that fits each task, with your company's context in every one.
Plug Coworker into the apps where your work already lives: 50+ connectors including Slack, Gmail, Salesforce, Jira, Notion, BigQuery, and Snowflake. Enterprise can add custom connectors for internal systems.
View all connectorsEvery artifact and code-execution task runs in an isolated cloud sandbox. Nothing executes on your machine and nothing leaks across orgs.
Coworker learns the structure of your tools (custom fields, workflows, terminology) so every answer and action fits how your team actually works.
Personal preferences, recent context, and recurring tasks remembered per user across sessions.
Plug any MCP-compatible AI tool (Claude Code, Cursor, your own agents) into your Coworker organization. Read context, run agents, write back to your stack.
Learn moreReads, writes and acts across your connected tools in plain language. You get sourced answers and finished work rather than a chat transcript.
Learn moreYour engineers' coding agent, in the terminal they already use. They pick the model per task, from near-free open ones to frontier, with your company's context in the session and a budget you set.
Learn moreLearns how your team works, then runs it in the background. Triggers from your connected tools, takes action across your stack, and is built in plain language with no code.
Learn moreLive transcription, summaries, and action items pulled from Zoom, Meet, and Teams, synced back to your CRM and docs.
Learn more
Save any agent, prompt, or workflow as a reusable template. Share it with your team or the whole org.
Share any artifact (deck, dashboard, doc, report) with teammates or external stakeholders via a secure link.
Dedicated implementation team to design custom workflows, build agents, train your team, and get you to value on day one.
- Dedicated CSM
Single sign-on via Okta, Azure AD, Google Workspace, and any SAML 2.0 IdP.
- Custom SLAs
Dashboards on how your team is using Coworker: adoption, time saved, and top agents.
Coworker is a context layer for enterprise AI. It connects 50+ work tools, builds a permission-aware company graph called OM2, and runs agents that act across those tools, inside Coworker or in Claude, ChatGPT and Cursor through Coworker MCP.
How can a company cut Claude and ChatGPT token spend without losing answer quality?
Coworker cuts token spend by sending models less, but more relevant, company context. In Coworker's 100-task benchmark, OM2 cost 89.1% less in tokens (9x cheaper) than Claude with native connectors, finished tasks 64% faster and was preferred 84.5% of the time. With Coworker's model routing on top, total token cost was 51x lower.
Common questions
Keep them. Coworker MCP plugs your company's context into Claude, ChatGPT, Cursor and any other MCP client, so the assistant you already pay for works across 50+ connected tools and OM2, the organizational memory that learns continuously from your company's work. Every query respects the permissions people already have in those tools. The same memory runs in Coworker's own apps, Work, Code, Agents and Meet, so context doesn't reset when the tool changes.
All of them. Work, Code, Agents, Meet, and the Coworker MCP server, with OM2 Organizational Memory underneath all of them. Nothing is sold as a per-product add-on.
Open and closed models are enabled for all customers. That is 13 models across Anthropic, OpenAI, Google, Moonshot, and Z.ai. They all run on US infrastructure, and closed models are passed through at the provider's price with no markup.
A dedicated implementation team to design your workflows, build your first agents, and train your people. After that you get a named Customer Success Manager, custom SLAs, SSO with your identity provider, and usage analytics for your admins.
It is tailored to your organization: the connectors you need, the number of people, and the deployment and support that fit. Book a demo and the team will scope it with you.
Book a demo. The team walks you through Coworker on your own stack, scopes what you need, and sets your workspace up.
Yes. Coworker is SOC 2 Type II certified, GDPR compliant, and CASA Tier 2 certified. Your data is encrypted at rest and in transit, and we never train models on your data.
See the numbers on your own data, free.
Our app runs the benchmark on your own systems, we cover the token costs, and you keep the numbers whichever way they land.
Get your free Context Impact ReportYour part: two steps
- Sign an NDA and tell us which systems to cover.
- Name one or two testers to install the Coworker benchmarking app.
Our part: everything else
Connect your sources, build your graph, run the question set against your current setup and through Coworker, score every answer blind, then tune and run it again.
Ready to put AI to work?
Connect your tools, deploy your first agent, and see results in minutes.