Keep the AI your company already uses.
MCP.computer is not another chatbot or a required agent runtime. Claude, ChatGPT, Codex, OpenClaw, custom applications, and future AI tools can all use the same company-approved resources.
The workspace is the control point. Start with clear defaults, then open Advanced for exact providers, model aliases, connection keys, MCP servers, tools, parameters, environments, budgets, and diagnostics.
Choose the right models
Give apps stable names like Smart, Fast, and Low cost while you choose the provider and model behind each one.
Approve tools and skills
Decide which capabilities are available, which actions are allowed, and how each person connects their account.
Share trusted knowledge
Turn reviewed company information into canonical context that people and AI can retrieve consistently.
See and control usage
Manage access, budgets, connections, and activity without storing prompts, responses, tool inputs, or credentials.
One workspace, two paths.
Choose and govern
The dashboard stores your desired setup: approved resources, aliases, access, budgets, and customer-visible history.
Use and enforce
Requests go directly through the managed gateway, where the effective model, budget, MCP server, tool, and parameter policy is enforced.
Make one useful path work first.
Create a workspace, then choose a path. Connect an AI app your team already uses to approved tools and knowledge without changing its model, or connect software to governed company models. Add the other capabilities when the workflow needs them.
Follow the getting-started guide