CoWorker Protocol
VerifiedTurn private skills into secure remote APIs that never expose your code or prompts.
What is CoWorker Protocol?
CoWorker Protocol converts local skill definitions into protected network services. Your analysis logic, prompts, and supporting files execute exclusively on your hardware, while remote users interact only through defined input and output schemas over end-to-end encrypted messaging.
The system layers access controls on top of this architecture. Four trust levels, time-limited permissions that auto-revoke after tasks complete, and the option to hide skills entirely prevent unintended exposure. All communication happens peer-to-peer without central servers or code transmission.
It targets professionals who have developed proprietary methodologies and want to collaborate without risking distillation or theft. Teams needing temporary access to specialized agent behaviors will find the permission model and audit-friendly design practical.
Capabilities
What you can build with CoWorker Protocol
Protected Methodology Sharing
Share industry analysis skills with colleagues or clients while keeping the underlying reasoning and data sources completely private on your device.
Temporary Cross-Team Access
Grant time-bound access to internal agent capabilities for a project, then let permissions automatically expire once the work finishes.
Secure External Consulting
Offer paid analysis services through an agent interface without ever sending your prompts, knowledge base, or evaluation rules to the customer.
Install CoWorker Protocol
pip install agent-coworkerpip install agent-coworker
coworker init --name my-agent
coworker bridge start
coworker demo- 1Install the package with pip install agent-coworker.
- 2Prepare a folder containing your SKILL.md or define skills in Python code.
- 3Set your LLM API key as an environment variable and run coworker serve on the skill directory.
- 4Generate an invitation code with coworker invite and share it with intended users.
- 5Users connect via coworker connect and call the skill remotely to receive results.
CoWorker Protocol: pros & cons
Pros
- +Strong architectural privacy since code and prompts never leave the host machine
- +Built-in trust management and automatic permission expiration reduce long-term exposure risk
- +Works over the open internet with end-to-end encryption and no central server required
- +Simple CLI flow to turn existing skills into callable endpoints quickly
Cons
- –Requires the host machine to stay online and reachable for calls to succeed
- –Depends on the XMTP network for messaging, adding an external dependency
- –Skill authors must manage their own LLM API keys and local runtime resources
Frequently asked questions
No. Only the function schema and final return values are transmitted; the implementation stays local.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other agent frameworks options worth comparing.
