AI agents invoke run_command to trigger actions in Code. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
Shell command execution (run_command) permits running arbitrary programs with the privileges of the process, including potentially destructive operations (rm, dd), financial operations (curl to payment APIs), or lateral movement attacks. While destructive/financial outcomes depend on the specific command argument, the capability itself is Execute.
From the tool's definition Tool explicitly 'Execute[s] a shell command and return its output.' Server description confirms it 'enables Claude AI to run terminal commands.' Shell command execution is fundamentally Execute—arbitrary code may be invoked depending on arguments.
Documented attack patterns abuse exactly the kind of access run_command gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and Code, and nothing reaches the server without passing your rules. This is the rule we recommend for run_command:
{
"version": "1",
"default": "deny",
"tools": {
"run_command": {
"limits": [
{
"counter": "run_command_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}run_command stays usable, but rate-capped — a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
Free to start. No card required.
Execute a shell command and return its output. It is categorised as a Execute tool in the Code MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Code MCP server in PolicyLayer and add a rule for run_command: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Code. Nothing to install.
run_command is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the run_command rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for run_command. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
run_command is provided by the Code MCP server (54yyyu/code-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Deterministic rules across all 9 Code tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.
Free to start. No card required.
9 Code tools catalogued and risk-classified — across an index of 42,500+ MCP servers.