start · MCP Codex Worker MCP: Risk & Policy

WHAT IT DOES

What start does on MCP Codex Worker

AI agents invoke start to trigger actions in MCP Codex Worker. 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.

WHY IT NEEDS A POLICY

Why start needs a policy

This tool executes code jobs asynchronously in a sandboxed environment. While sandboxed, the tool's ability to run arbitrary Claude Code jobs, verify results, and auto-revise based on failures means an AI agent could trigger unintended computations, resource consumption, or side effects depending on job specifications.

From the tool's definition Tool description states it 'Start[s] an asynchronous Claude Code job' with 'reasoning layer verifies the result' and 'auto-revise on concrete failures'.

Attacks that exploit this kind of access

Recommended rule Ready

Verdict

start Rate-limited

start 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.

View as policy code

policy.json

{
  "version": "1",
  "default": "deny",
  "tools": {
    "start": {
      "limits": [
        {
          "counter": "start_rate",
          "window": "minute",
          "max": 10,
          "scope": "grant"
        }
      ]
    }
  }
}

RATE-LIMIT THIS TOOL →

Instant setup, no code required.

PolicyLayer is an MCP gateway: every tool call is checked against your rule before it reaches MCP Codex Worker, so you can run agents against it safely.

FAQ

Questions about start

What does the start tool do? +

Start an asynchronous Claude Code job inside SANDBOX_ROOT. A deterministic Mythos-style reasoning layer verifies the result (checks, scope, exit, calibration) and can auto-revise on concrete failures. It is categorised as a Execute tool in the MCP Codex Worker MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.

How do I enforce a policy on start? +

Register the MCP Codex Worker MCP server in PolicyLayer and add a rule for start: 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 MCP Codex Worker. Nothing to install.

What risk level is start? +

start is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.

Can I rate-limit start? +

Yes. Add a rate_limit block to the start 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.

How do I block start completely? +

Set action: deny in the PolicyLayer policy for start. 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.

What MCP server provides start? +

start is provided by the MCP Codex Worker MCP server (sorryorc/mcp-codex-worker). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.