Coding agents and MCP

benchbar mcp is a Model Context Protocol server on stdio, so Claude Code, Cursor and other agents can see and drive your benches:

claude mcp add benchbar -- benchbar mcp

For another client, add a stdio server whose command is benchbar with the argument mcp.

Tools

Tool	What it does
benchbar_list	Every bench benchbar knows: path, default site, sites, ports, whether its service is installed
benchbar_status	Live state of one bench, with pid, ports, sites and site ping
benchbar_doctor	The read only health report of one bench, every check with its level and fix
benchbar_logs_tail	The last lines of the bench log, one process if asked
benchbar_site_list	The sites of one bench, the default, the hosts line and the ping code
benchbar_up	Start a bench and wait for its default site
benchbar_down	Stop a bench and keep it stopped, also across reboots
benchbar_restart	Restart every process of a bench

benchbar_list, benchbar_status, benchbar_doctor, benchbar_logs_tail (last lines, one process if asked) and benchbar_site_list read; benchbar_up, benchbar_down and benchbar_restart act. Each one runs benchbar ... --json and returns what the CLI printed. Nothing that repairs, installs or needs sudo is offered. It needs only python3, which the Command Line Tools provide.

Every tool but benchbar_list takes an optional bench, the absolute path from benchbar_list; without it the tool acts on the default bench. The JSON each one returns is described in the JSON schema.

Repairs and installs

Repairs and installs are deliberately not tools. An agent runs them in a terminal, with you, and shows the plan first:

benchbar doctor --json
benchbar repair --dry-run
benchbar repair --yes

AGENTS.md in the repository is written for agents that are pointed at BenchBar with “set up Frappe on my Mac” or “my bench is broken, fix it”: how to start from the facts, the rules they must keep (never rm -rf inside a bench, never bench update unless asked, never print the MariaDB root password), and how to read the output.