MateMCP

Local by design Work executes on your own device.

Least privilege Access is scoped to what you allow.

Human in the loop Sensitive actions can require approval.

Auditable Actions leave a clear operational trail.

Get more useful work from the AI you already have.

MateMCP does not increase a provider's usage quota. It gives an available AI chat another way to work: through scoped tools on the computers and projects you already own.

Use the AI plan you already pay for.

If a provider's built-in coding or agent tool reaches its usage limit while chat remains available, connect that conversation to MateMCP and keep working through scoped local tools.

See and operate interfaces.

Use semantic browser and native desktop controls when a workflow cannot be completed through an API or CLI.

Keep humans in control.

Gate sensitive operations with explicit approval flows instead of handing an assistant unrestricted machine access.

Use credentials without exposing them.

Secrets are resolved on the Agent and injected into approved destinations without returning plaintext values to the model.

Stay connected through real networks.

Reconnect-aware transport keeps long-running workflows recoverable when devices or networks briefly disappear.

A narrow bridge between AI and your machine.

The assistant talks MCP. MateMCP turns that into scoped local capabilities, while the control plane handles identity, policy, and approvals.

01 AI client ChatGPT, Claude, Grok, or another MCP client

02 MateMCP relay OAuth, routing, scopes, policy, approvals

03 Your Agent Local files, shell, browser, desktop, secrets

Powerful access should still feel deliberate.

MateMCP is built around explicit boundaries rather than pretending local computer access is risk-free.

Scoped projects File and shell access can be constrained to registered workspaces and capabilities.

Approval gates Sensitive operations can pause for a human decision before execution.

Local secret resolution Credentials can stay on the Agent instead of appearing in model context or tool arguments.

Traceable actions Approvals and important actions are designed to be inspectable and auditable.

Project access

Scoped

MateMCP C:\Projects\MateMCP

Read filesAllowed

Write filesAllowed

ShellApproval aware

Outside workspaceBlocked

Least privilege is easier to trust.

One workflow across macOS and Windows.

Install the Companion on the computers you want AI to work with, then manage access from one account.

⌘ macOS Companion + local Agent

Windows Companion + local Agent