Getting Started with OpenProse for AI Workflows

I have spent too many hours manually copying data between different AI chat windows. It is exhausting to act as the middleman, taking the output from a research prompt and pasting it into a writing prompt. I wanted a way to define these steps in a file so the agents could just talk to each other.

OpenProse solves this. It is a portable, markdown-first format for orchestrating AI sessions. It runs as a plugin in OpenClaw, giving you a /prose command to run multi-agent programs. You can use it for research, code reviews, or content pipelines without doing the manual handoffs yourself.

What You’ll Need

Before you start, make sure you have these pieces ready:

• OpenClaw: The runtime where the plugin lives.
• Tool Access: Your tool allowlist must permit sessions_spawn, read, write, and web_fetch.
• Plugin Files: Either the bundled plugin or a local checkout of the OpenProse extension.

Quick Start

You can get OpenProse running in about five minutes. Here is the path to your first automated workflow.

1. Enable the plugin

OpenProse is bundled with OpenClaw but starts disabled. Run this command to turn it on:

openclaw plugins enable open-prose

2. Restart the Gateway

The changes won’t take effect until you restart your OpenClaw Gateway. Do that now.

3. Check the command

Verify it is working by typing this in your terminal:

/prose help

4. Create your first .prose file

Save the following code into a file named research.prose. This script runs two agents at the same time and then merges their work.

# Research + synthesis with two agents running in parallel.

input topic: "What should we research?"

agent researcher:
  model: sonnet
  prompt: "You research thoroughly and cite sources."

agent writer:
  model: opus
  prompt: "You write a concise summary."

parallel:
  findings = session: researcher
    prompt: "Research {topic}."
  draft = session: writer
    prompt: "Summarize {topic}."

session "Merge the findings + draft into a final answer."
context: { findings, draft }

5. Run it

Execute your program with the run command:

/prose run research.prose

Troubleshooting

If things go wrong, it is usually related to permissions or missing binaries.

• Program Fails Immediately: Check your Skills config. If your tool allowlist blocks sessions_spawn, OpenProse cannot start sub-agents. It also needs read and write for file operations.
• Experimental State Errors: If you try to use the sqlite or postgres state modes, ensure you have the sqlite3 or psql binaries installed on your system.
• Remote Fetch Errors: Using /prose run <handle/slug> requires the web_fetch tool. If this tool is disabled, remote programs will not load.
• Database Credentials: If you use the experimental Postgres backend, be aware that credentials appear in sub-agent logs. I recommend using a dedicated database with limited privileges.

If you need help setting up your specific environment, check the AI Setup Assistant.

What’s Next

• Plugins Documentation
• Plugin Manifest Details
• Skills Overview
• Skills Configuration