What started as a private experiment shortly turned a worldwide open supply undertaking with extraordinary momentum.
OpenClaw is a private AI assistant that runs on customers’ gadgets and connects with the messaging channels they already use. Began by Peter Steinberger as a weekend undertaking in November 2025, its GitHub repository has grown to roughly 388,000 stars, 81,000 forks, and greater than 80,000 commits by August 26, 2026.
On this video interview, filmed simply six months into the undertaking, creator Peter Steinberger and a number of other OpenClaw maintainers focus on managing a surge of pull requests, rethinking contributor belief and code assessment, addressing software program provide chain dangers, and balancing highly effective agent capabilities with safety. In addition they share safety classes from the GitHub Safe Open Supply Fund and the worth of connecting with maintainers dealing with related challenges. Watch the complete video above, then discover the important thing classes under.
Folks on this video
The next maintainers shared their experiences sustaining and securing OpenClaw.
Peter Steinberger, Creator of OpenClaw
Brad Groux, CEO, Digital Meld
Josh Avant, Member of technical workers, OpenClaw Basis
Josh Lehman, Martian Engineering
Sally O’Malley, Principal software program engineer at Crimson Hat
Val Alexander, OpenCoven
Vincent Koc, Chief architect, OpenClaw Basis
Listed here are the highest 10 classes that we took away from the dialog.
Classes 1–3: How AI modified contributions and group
1. Pull requests turned immediate requests
OpenClaw’s maintainers discovered themselves managing 1000’s of pull requests and points, with some contributors opening tons of of pull requests directly.
I don’t even name them pull requests. I name them immediate requests.
Peter Steinberger
There have been some contributors that had a number of tons of of pull requests operating these kind of automated software program factories that have been simply mining every part for points.
Josh Lehman
The problem shifted from attracting participation to discovering invaluable contributions amid a flood of exercise that might overwhelm human assessment.
2. Hold the door open for brand spanking new contributors
OpenClaw’s maintainers needed the undertaking to be welcoming to new members, whether or not they have been first-time open supply contributors, non-developers fixing a particular drawback, or folks utilizing AI brokers to assist. Quite than dismissing imperfect contributions, they seemed for promising concepts and labored with contributors to refine, rewrite, or full the ultimate adjustments themselves.
I understand how it felt when, a few years in the past, my first pull request was accepted on a undertaking.
Peter Steinberger
Among the first-time contributions that have been merged got here from folks and not using a growth background. They used an agent to create a pull request, and labored with maintainers to complete the change.
A great proportion of these first-time pull requests that bought merged are from non-developers. They’re simply those that have a particular drawback and a necessity.
Vincent Koc
3. Brokers save time however make it tougher to log off
The maintainers described two very completely different outcomes from the identical know-how: brokers may also help folks reclaim time, however they will additionally make it tougher to cease working.
I’ve seen the opposite aspect of it, the place individuals are simply so in love with it they usually notice, wow, if I don’t sleep tonight, I can do what used to take every week for me to do.
Val Alexander
I’ve three youngsters. They’re very small. OpenClaw lets me handle brokers that go and work for me so I can get again to taking part in with my youngsters.
Josh Lehman
Typically the maintainers will go on the channel and say, ‘I’m going to the touch grass now. I’m taking a number of hours off.’
Sally O’Malley
Brokers are neither good nor dangerous for work-life steadiness. However they amplify each the chance to do extra and the significance of figuring out when to step away.
Classes 4-6: How maintainers tailored
4. Earn belief by discovering the place you’ll be able to add worth
There was no single path to turning into an OpenClaw maintainer. Some contributors arrived via safety work, others via integrations or group participation, however the widespread thread was discovering a approach so as to add worth and taking possession.
Peter ignored me, so I used to be like, how else can I get his consideration? Safety.
Vincent Koc
I’m a Microsoft man, so I believed, is there a plugin for Microsoft Groups?
Brad Groux
I seemed into the group and I used to be in voice chat, and folks have been asking a number of questions, and I used to be like, nicely, how can I add worth in these conversations?
Val Alexander
5. The brand new belief sign is exhibiting your work
As contribution counts turned much less informative, the staff recognized proof that might assist a pull request stand out: agent transcripts, screenshots, testing, and an evidence of the contributor’s considering.
In the event you present us with the transcripts, we truly see the way you got here to the pull request and your dialogue with the agent. Extremely invaluable. In the event you add screenshots, you’ll be able to show that you just examined this.
Peter Steinberger
The vital query was not merely whether or not a human or an agent wrote the code. It was whether or not the contributor understood the function and had thought of the way it interacted with the remainder of the undertaking.
No person cares if you happen to wrote the code or not, however we care if you happen to truly considered this function.
Peter Steinberger
6. Maintainers are reviewing agent code with brokers
Maintainers more and more relied on AI instruments to assist assessment AI-generated contributions, whereas additionally taking a extra hands-on strategy to bettering submitted code.
Each time I get a pull request from an AI, one factor I like to do now’s use GitHub Copilot for all of the critiques. I simply press a button proper there. It does a assessment and generates readability on all of the recordsdata which might be connected, what the recordsdata imply, and the way they modified.
Val Alexander
That is the primary undertaking the place I noticed it develop into normalized that when somebody submits a pull request, as a maintainer, you simply edit it. You simply make it proper.
Josh Lehman
Classes 7–9: Safety challenges
7. Fame turned an assault floor
Contribution historical past itself may very well be manipulated. OpenClaw’s maintainers noticed folks duplicate present pull requests, and Vincent Koc explains why.
Folks would principally duplicate different folks’s pull requests. What they have been making an attempt to do right here was to construct credibility, as a result of we had these badges, like what number of you’ve merged. So the extra merges you had, it was like a belief sign to us maintainers.
Vincent Koc
Peter described an organization utilizing an automatic pull request to advertise their product. The staff needed to establish duplicate work and decide which pull request was the unique.
The code was not the one factor the undertaking wanted to judge. Maintainers additionally needed to rethink the social alerts they used to determine what, and whom, to belief.
8. “Protected by default” is determined by who you ask
What feels secure to 1 person could really feel unnecessarily restrictive to a different.
The tradeoff was clear in apply. Tighter workspace restrictions generated person complaints, whereas fewer restrictions might expose the undertaking to safety incidents.
It’s actually typically a tough sport to seek out the suitable steadiness between making it actually handy for customers and in addition constructing one thing that’s secure sufficient as a default.
Peter Steinberger
Safe defaults should account for an agent’s capabilities, what customers perceive, and what a selected surroundings is ready to permit.
9. Know who maintains your dependencies
Current provide chain assaults pushed the maintainers to assume extra rigorously about each the dependencies they relied on and their relationship with the tasks behind them.
We went via our dependencies with a fine-tooth comb. What it’s pushed us to do is definitely scale back the core dependencies, but in addition create a relationship with the maintainers that we’ve got a dependency on.
Vincent Koc
It’s not the default that firms truly attempt to contribute again as a substitute of simply sustaining a fork and never caring.
Peter Steinberger
Lesson 10: How the GitHub Safe Open Supply Fund helped
Members described the GitHub Safe Open Supply Fund as each a safety studying expertise and a solution to join with maintainers confronting related, typically overwhelming, issues.
The presenter was like, first, go get a cup of espresso. The first step, take a breath. It linked us to the human aspect of being a maintainer.
Josh Avant
This system supplied larger consciousness of safety practices and helped collaborating maintainers perceive easy methods to immediate brokers.
Now we have brokers now, they usually can do absolutely anything that you just ask them to do, however you continue to should know what to ask them to do. Now I’ve the power to know what to ask for.
Josh Lehman
Vincent emphasised the worth of assembly different maintainers who have been experiencing the identical challenges of securing open supply tasks. This system gave members a group they may faucet into and study from as these challenges continued.
Proceed the dialog
Watch the complete dialog to listen to how OpenClaw’s maintainers are adapting when contributions scale quicker than the human techniques used to assessment, safe, and maintain them.
OpenClaw participated in Session 4 of the GitHub Safe Open Supply Fund. To study extra, learn What 50 open supply tasks taught us about safety within the AI period.
Functions for the GitHub Safe Open Supply Fund at the moment are open. In the event you’re sustaining an open supply undertaking, apply to study from safety consultants, join with fellow maintainers, and strengthen the safety of your undertaking.
Head over to the GitHub Group and ask the maintainers what it’s actually like constructing the fastest-growing open supply undertaking in GitHub historical past!
Thanks to all GitHub Safe Open Supply Fund Companions
Collectively, we’re serving to safe the open supply ecosystem for everybody!
Funding Companions: Alfred P. Sloan Basis, American Categorical, Chainguard, Datadog, Herodevs, Kraken, Mayfield, Microsoft, Shopify, Stripe, Superbloom, Vercel, Zerodha, 1Password
Ecosystem Companions: Atlantic Council, Ecosyste.ms, CURIOSS, Digital Knowledge Design Institute Lab for Innovation Science, Digital Infrastructure Insights Fund, Microsoft for Startups, Mozilla, OpenForum Europe, Open Supply Collective, OpenUK, Open Know-how Fund, OpenSSF, Open Supply Initiative, OpenJS Basis, College of California, OWASP, Santa Cruz OSPO, Sovereign Tech Company, SustainOSS
Searching for sensible safety steps you’ll be able to take to your undertaking? Arrange the safety baseline to your undertaking in a single minute >
Written by
