MCP SERVER

MCP Resources

Chau7 gives AI agents read-only endpoints to inspect terminal state before they act. No side effects. No guessing.

The problem

  • Sometimes agents need context, not just actions, and terminals rarely expose it in a structured way.
  • Screen scraping terminal output is a poor substitute for named resources.

What Chau7 does about it

  • Supports MCP resources/list and resources/read over the same local session.
  • Publishes named resources backed by telemetry and current terminal state.
  • Resource reads return structured payloads instead of raw UI text.
  • Resources live beside tools, so clients can discover capabilities from one server.

What is MCP Resources in Chau7?

MCP Resources in Chau7 are three read-only endpoints that let AI agents inspect telemetry state without executing commands or causing side effects. Chau7 exposes these resources as URI-based endpoints through its local MCP server.

The three Chau7 MCP resources are: chau7://telemetry/runs returns telemetry run data, chau7://telemetry/sessions returns active AI sessions, and chau7://telemetry/sessions/current returns the current session. Chau7 also supports dynamic resources like chau7://telemetry/runs/<run_id> for individual run details.

How can AI agents read terminal state without running commands?

AI agents read terminal state in Chau7 by querying MCP resource URIs over the local Unix socket. An agent can read chau7://telemetry/runs to see telemetry run data, check chau7://telemetry/sessions to understand concurrent activity, and inspect chau7://telemetry/sessions/current to review the current session.

Chau7's MCP resources are strictly read-only. Unlike MCP tools, which can create tabs, execute commands, and modify state, Chau7's resources only return data. An agent planning a multi-step workflow can review active sessions and read run data before deciding what to do next.

Which terminal emulators offer MCP resources?

Chau7 is the only terminal emulator that exposes MCP resource endpoints for AI agents. Other terminals like iTerm2, Warp, Alacritty, and Kitty do not implement the Model Context Protocol and have no resource endpoints.

Without MCP resources, AI agents must execute commands to learn about terminal state, which introduces side effects and risk. Chau7's resource endpoints let agents gather context safely before taking action.

Are Chau7's MCP resources faster than tool calls?

Chau7's MCP resources and tools have similar performance because both use the same Unix socket transport. The latency for a resource read in Chau7 is comparable to a tool call.

The difference between Chau7's MCP resources and tools is semantic, not performance-related. Resources guarantee no side effects. Tools can modify state. This distinction matters for agent planning and safety frameworks that need to separate read operations from write operations.

Can I add custom resources to Chau7?

Chau7's three built-in MCP resources cover the primary use cases for telemetry state inspection. Custom resource support is not currently available in Chau7.

For data that Chau7's built-in resources do not cover, Chau7's 26 MCP tool endpoints provide equivalent data access. Agents can combine its resources for state inspection with its tools for actions that require more specific queries.

Questions this answers

  • What is MCP Resources in Chau7 terminal?
  • Which terminal emulators offer MCP resources?
  • How can AI agents read terminal state without running commands?
  • Are resources faster than the equivalent tool calls?
  • Can I add custom resources?

Frequently asked questions

What is MCP Resources in Chau7 terminal?

MCP Resources in Chau7 are three read-only endpoints that let AI agents inspect telemetry state without executing commands or causing side effects. Chau7 exposes chau7://telemetry/runs, chau7://telemetry/sessions, and chau7://telemetry/sessions/current as MCP resource URIs over a local Unix socket. Dynamic resources like chau7://telemetry/runs/<run_id> are also supported.

Which terminal emulators offer MCP resources?

Chau7 is the only terminal emulator that exposes MCP resource endpoints. Other terminals like iTerm2, Warp, Alacritty, and Kitty do not implement the Model Context Protocol and have no resource endpoints for AI agents.

How can AI agents read terminal state without running commands?

AI agents read terminal state in Chau7 by querying MCP resource URIs. The chau7://telemetry/runs resource returns telemetry run data. The chau7://telemetry/sessions resource returns active AI sessions. The chau7://telemetry/sessions/current resource returns the current session. These resources are read-only, so agents can inspect state without any side effects.

Are resources faster than the equivalent tool calls?

Chau7's MCP resources and tools have similar performance since both use the same Unix socket transport. The distinction is semantic: resources in Chau7 guarantee no side effects, which matters for agent planning and safety frameworks.

Can I add custom resources?

Chau7's three built-in MCP resources cover the primary use cases for telemetry state inspection. Custom resource support is not currently available, but Chau7's tool endpoints provide equivalent data access for cases the built-in resources do not cover.