AI is getting very good at telling us what to do.
The next challenge is letting AI actually do it — without giving an agent uncontrolled access to everything.
That’s the idea behind the OpenVerb Hackathon.
I’m inviting developers, students, AI engineers, open-source contributors, founders, and anyone interested in agent infrastructure to build something using OpenVerb.
What is OpenVerb?
OpenVerb is an open execution layer for AI actions.
Applications expose capabilities as structured verbs:
email.send
crm.create_contact
calendar.create_event
gis.add_layer
file.organize
Instead of letting an AI generate arbitrary commands, the execution path can be structured:
Intent
↓
Verb
↓
Schema
↓
Policy
↓
Executor
↓
Receipt
The model figures out what it wants to do.
OpenVerb provides infrastructure around how that action is represented, validated, authorized, executed, and recorded.
What Can You Build?
I’m intentionally keeping the hackathon broad.
🤖 Build an Agent
Give an agent explicitly defined capabilities instead of unrestricted access to an application.
For example:
research.search
document.create
email.send
calendar.create_event
Then use OpenVerb policies to control what the agent can actually execute.
🧩 Build an Integration
Connect an existing application or API to OpenVerb.
Create verbs for:
- GitHub
- Slack
- Notion
- GIS platforms
- CRMs
- developer tools
- productivity software
- hardware
- your own API
The interesting question is:
What capabilities should this application expose to AI?
🛠️ Build Developer Tools
The OpenVerb ecosystem itself creates opportunities for tooling.
You could build:
- a verb debugger
- schema generator
- visual policy editor
- receipt explorer
- verb registry browser
- workflow visualizer
- testing framework
- new executor
- development environment
📄 Build With .ov
One of the newer parts of OpenVerb is the .ov workflow format.
The idea is to make AI-generated workflows portable:
Prompt
↓
Generate workflow
↓
Save .ov
↓
Inspect
↓
Execute
↓
Share
↓
Reuse
Instead of an automation disappearing inside a chat session, it becomes an artifact.
That means you could build a .ov generator, editor, runner, library, marketplace, converter, visualizer — or something completely different.
🗺️ Build Something Unexpected
This is probably what I’m most excited about.
Use OpenVerb for GIS.
Robotics.
Music.
Research.
Education.
Business automation.
Accessibility.
Developer infrastructure.
Games.
Scientific workflows.
Land surveying.
Or a use case nobody involved with OpenVerb has considered yet.
A Simple Example
Imagine the user says:
“Add Sarah Chen from Atlas Engineering to my CRM.”
The model could translate that into:
{
"verb": "crm.create_contact",
"input": {
"first_name": "Sarah",
"last_name": "Chen",
"company": "Atlas Engineering"
}
}
OpenVerb can then validate the request.
Policy determines whether the operation is permitted.
The appropriate executor performs the action.
A receipt records the result.
Natural Language
↓
Structured Intent
↓
crm.create_contact
↓
Validate
↓
Policy
↓
Execute
↓
Receipt
That’s the pattern I want developers to experiment with.
🛡️ The Trust Problem
The hackathon isn’t only about whether an agent can perform an action.
I want projects to explore whether it should.
Take:
email.send
A useful implementation should be able to think about questions like:
- Who is the recipient?
- Is this agent allowed to send email?
- Does this action require confirmation?
- Are certain domains prohibited?
- What information can be included?
- How do we prove what was actually sent?
As agents become more capable, these questions become infrastructure problems.
That’s why schemas, policies, executors, and receipts are important parts of OpenVerb.
What I’m Looking For
The biggest project won’t necessarily win.
I’d rather see three beautifully designed verbs than 100 poorly controlled tools.
Projects should focus on four things:
Creativity — Apply OpenVerb somewhere interesting.
Execution — Make the AI actually accomplish something.
Structure — Define clear capabilities and interfaces.
Trust — Think about validation, policies, permissions, and receipts.
The Ecosystem
OpenVerb has grown beyond the original protocol.
The ecosystem includes work around:
Protocol
Framework
Runtime
Policy
SDK
CLI
Core
.ov
Viewer
Studio
CRM
GIS
IDE
Pulse
Stdlib
Markdown
Python Core
Fellowship
You don’t need to use everything.
Choose the part that makes sense for your project.
Or build something that extends the ecosystem itself.
Why I’m Doing This
I’ve spent a lot of time thinking about what OpenVerb could become.
But that’s also a limitation.
I know what I would build with it.
An open protocol becomes much more interesting when someone else looks at the same primitives and says:
“Wait — I could use this for something completely different.”
That’s what I want this hackathon to uncover.
Build. Share. Execute.
Full dates, rules, registration, judging, prizes, and submission details will be announced separately.
For now, start thinking about one question:
What would you build if AI could reliably take action?
⚡ OpenVerb
https://openverb.dev
📦 NPM
https://www.npmjs.com/package/openverb
If you’re interested in participating, contributing, mentoring, judging, sponsoring, or partnering with the OpenVerb Hackathon, reach out.
Don’t just make AI talk. Make it act.
Top comments (0)