Using the Lobster Tool in Openclaw: Automate Workflows with Shell Pipelines

Table of contents: [Show]

Quick Start

The Lobster tool in Openclaw lets you run shell-based automation workflows directly from your agent. It is perfect for chaining commands, processing data, and automating repetitive tasks using familiar Unix pipelines.

Installation

Because Lobster runs as a local subprocess, you must first install the Lobster CLI on the same host that runs your Openclaw Gateway. Ensure that lobster is accessible on your system PATH.

Enable the Tool

Add Lobster to your tool allowlist in ~/.openclaw/openclaw.json:

{
  "tools": {
    "alsoAllow": ["lobster"]
  }
}

You can also configure this per-agent under agents.list[].tools.alsoAllow.

Running Workflows

Lobster executes shell pipelines with built-in timeout and resume support. Here is an example of a run call that processes an inbox:

{
  "action": "run",
  "pipeline": "inbox list --json | inbox categorize --json | inbox apply --approve",
  "timeoutMs": 30000
}

This safely chains multiple commands, passing JSON output between steps.

Resuming Workflows

If a workflow pauses for approval, you can resume it with a token:

{
  "action": "resume",
  "token": "",
  "approve": true
}

Best Practices

Always set timeoutMs to prevent hung processes
Use JSON output flags for structured data between pipeline steps
Store resume tokens if your workflow requires human approval
Test pipelines manually before automating them through Openclaw

For more automation options, check out our guide on Exec Tool Openclaw. If you are new to Openclaw, you might also want to read about Elevated Mode Openclaw for advanced permissions.

Troubleshooting

Command not found: Verify Lobster is installed and on your PATH
Permission denied: Check your tool allowlist configuration
Timeout errors: Increase timeoutMs for long-running pipelines
Resume fails: Ensure the resume token is valid and not expired