Reado

What Reado MCP server does

Reado MCP server connects an agent to the review state maintained by Reado, a local-first code-reading and review application. Developers can inspect code, anchor comments to exact line ranges, classify them as bugs, refactors, performance issues, questions, or notes, and mark selected comments as tasks for an agent.

The server gives the agent access to the open review work through the reado://tasks and reado://comments MCP resources. This lets an agent read the developer’s instructions and context before editing files. Reado stores comments as Markdown under .reado/, where they remain part of the project rather than being held only in the application interface.

How it works

The workflow begins in Reado: choose working changes, a branch, or a pull request, then review the relevant files and leave comments. Findings proposed by the agent remain subject to approval, while developer-created comments can be sent back as tasks.

When a review is sent, the agent runs in Reado’s terminal and uses the reado binary to read open tasks, consult project documentation and specifications, modify files, and report task status. It can mark work complete with reado task done <id> or report a failure with reado task fail <id> "<reason>". Reado watches these changes and updates the review state as tasks close.

Comments are tied to line ranges and can be re-anchored when code moves. If the referenced code disappears, the comment becomes an orphan rather than silently pointing to an unrelated line.

Setup and configuration

Install Reado from a signed release for macOS, Linux, or Windows. The packaged application can install the command-line tool from Settings → Command-line tool, linking reado into ~/.local/bin. The binary must be available on the agent’s PATH for the MCP and CLI workflow to operate.

The application automatically wires the MCP server into each supported agent’s configuration when a project opens. The server is started with reado mcp. Source builds can use scripts/install-cli.sh to build and link the CLI. Reado does not require an account; project code and comments remain local plain files.

Tools and capabilities

Reado MCP server supports an agent workflow centered on:

• Reading open review tasks through reado://tasks.
• Reading review comments through reado://comments.
• Listing tasks with reado task list.
• Closing tasks with reado task done <id>.
• Reporting task failures with reado task fail <id> "<reason>".
• Working with comments that include line ranges, types, threads, and task or note status.

Limitations and notes

The server depends on the reado binary being installed and visible on the agent’s PATH. The provided material describes a local project workflow rather than a hosted review service. Reado supports Claude Code, Codex, Copilot, Gemini, OpenCode, and Cursor; client-specific behavior may depend on each client’s MCP configuration.

The project is MIT-licensed and supports macOS, Linux, and Windows. Reado’s on-disk format may evolve, while the reado binary is described as the stable interface for agents.