agent-sandbox

agent-sandbox

An MCP (Model Context Protocol) server that routes shell commands to either the host machine or a Docker Compose container, based on operator-configured allow patterns.

Install

go install github.com/ynagai/mcp-command-router@latest

Configuration

Copy config.example.toml to config.toml and edit:

[mcp]
command_output_dir = "/tmp/mcp-output"

[sandbox.container]
build_context = "./docker/sandbox"
dockerfile = "Dockerfile"
image = "myapp"

[sandbox.command]
allow = [
  "git *",
  "make *",
]

Usage

Start the project sandbox from your project root:

agent-sandbox sandbox-up -d --config agent-sandbox.toml

Start the MCP server:

agent-sandbox command-router --config agent-sandbox.toml

Stop the current project sandbox:

agent-sandbox sandbox-down --config agent-sandbox.toml

Remove all Docker containers and networks that appear to be managed by agent-sandbox:

agent-sandbox sandbox-prune

sandbox-prune is destructive. It removes every container labeled cr.managed=true and every Docker network whose name starts with cr-sandbox-.

Check whether external dependencies are usable on this host:

agent-sandbox doctor

doctor verifies that nono is on PATH, that docker compose version works (which also accepts compatible CLIs like colima or podman that alias docker), and that the Docker daemon is reachable. Exits 0 when all checks pass, 1 otherwise.

Register as an MCP tool in your Claude Code settings.

Route a single command through the router from the shell (streams output live):

agent-sandbox exec --config agent-sandbox.toml -- git status

Install the Claude Code PreToolUse hook so Bash and Monitor commands route through the sandbox automatically:

agent-sandbox install-hook

This merges a PreToolUse hook into .claude/settings.json (matchers Bash and Monitor) that rewrites each command to agent-sandbox exec -- <command> via agent-sandbox hook. agent-sandbox must be on PATH.

Tool mode

tool_mode in agent-sandbox.toml selects how the agent's commands reach the router:

• mcp (default): the claude launcher disables the Bash and Monitor tools, and the agent routes commands through the run_command MCP tool.
• hook: the launcher leaves Bash and Monitor enabled, and the PreToolUse hook rewrites each command to agent-sandbox exec -- <command>. This requires the hook to be installed (agent-sandbox install-hook); agent-sandbox claude refuses to start in hook mode if the hook is missing for either the Bash or Monitor matcher.

How It Works

• Commands matching a drop pattern are refused — neither the host nor the container runs them; the MCP response carries exit code 1 and a stderr file containing dropped: command matches drop pattern "<pattern>".
• Commands matching an allow pattern are executed on the host (after shell-safety validation).
• All other commands are routed to the configured Docker Compose service.
• Drop wins over allow.
• Output is always written to separate stdout/stderr files; the MCP response returns file paths and exit code only.

Migrating from an older config

The configuration was reorganized; old keys are no longer accepted.

The deny routing axis is gone. Patterns that previously forced a host-allowed command into the sandbox now have two options: leave them out of allow (so they default to the sandbox), or add them to drop if they should be refused entirely.