Entrepreneurial second acts rarely happen by design. More often, they emerge from a collision of experience, burnout, and a nagging sense that there's a better way to build technology. This is the story of how one developer went from selling his company to creating an AI assistant used by 84,000+ developers worldwide.
Act I: The PSPDFKit Years (2011-2023)
In 2011, the mobile app ecosystem was exploding. iPhone apps were everywhere, but many struggled with a surprisingly mundane problem: rendering PDF documents reliably. Peter Steinberger, a Vienna-based iOS developer, saw the opportunity.
PSPDFKit started as a side project—a high-performance PDF rendering SDK for iOS. The technical challenge was real: PDFs are notoriously complex, with thousands of edge cases, inconsistent specifications, and performance requirements that test even experienced engineers.
What started as a solo developer SDK grew into a comprehensive document processing platform used by Fortune 500 companies. Dropbox, Salesforce, SAP—major enterprises depended on PSPDFKit for document workflows. The company bootstrapped profitably for years before taking strategic funding.
"We focused obsessively on developer experience. Our SDK had to be so good that developers would advocate for it internally. Enterprise sales followed adoption, not the other way around."
By 2023, PSPDFKit employed over 100 people across multiple countries, served billions of documents annually, and generated substantial revenue. When Nutrient (formerly DocuWare) acquired the company for $116 million, it seemed like the classic success story.
The Post-Exit Paradox
Conventional wisdom says exiting your company is the dream outcome. Financial security. Validation. Freedom to do anything next. The reality proved more complex.
Peter stayed on post-acquisition, helping integrate PSPDFKit into Nutrient's broader platform. But the energy that comes from building something from scratch—the daily problem-solving, the direct customer feedback, the autonomy—was gone. Replaced by corporate processes, alignment meetings, and strategic planning cycles.
"I was financially comfortable but creatively exhausted. I'd spent 15 years solving the same problem. I needed a new challenge, something that felt urgent and mattered."
The burnout was real. Not from overwork—from the loss of purpose. When you've built something for 15 years, stepping away creates an identity vacuum. Who are you when you're no longer the founder of [Company]?
The Genesis of clawbot: A Personal Need
The spark for clawbot came from frustration with AI tools. In late 2023, ChatGPT and Claude were everywhere. Peter used them daily—for coding help, research, brainstorming. But something was missing.
"These tools are conversational miracles," Peter observed, "but they can't actually do anything. I'd ask Claude for help with a deployment issue, it would give me the perfect commands to run, and I'd copy-paste them into my terminal. Why can't the AI just execute them?"
The limitation was architectural. Cloud AI can never have system access—it's a security boundary. For AI to execute commands, it must run locally with appropriate sandboxing and permission controls. No cloud provider would grant that level of access.
The solution was obvious: self-hosted AI with native system integration. Peter started building what would become clawbot in December 2023 as a personal tool. The goal was simple: an AI assistant that could genuinely delegate tasks, not just suggest them.
First Commit
Initial prototype: Node.js server bridging OpenAI API to local shell execution. Sandboxed commands, approval workflows, audit logging.
Multi-Channel Support
Added WhatsApp integration. The ability to delegate tasks via familiar messaging apps proved transformative—suddenly AI felt accessible anywhere.
Open Source Release
MIT License. Published to GitHub. Expected modest interest from power users. Reality exceeded expectations dramatically.
Viral Growth
HackerNews front page. 10,000 stars in 48 hours. Overwhelming community response validated the need for self-hosted, privacy-first AI automation.
Skills Ecosystem
Community-contributed skills exceeded 500. ClawdHub marketplace launched. AgentSkills standard established for plugin compatibility.
Production Scale
84,000+ GitHub stars. 11,200+ forks. Deployed in startups, enterprises, personal workflows worldwide. Active contributor community of 500+.
Why Open Source?
PSPDFKit was proprietary, venture-backed, and commercially licensed. clawbot took the opposite approach: open source, community-driven, MIT License. The choice was deliberate.
Trust Through Transparency
AI with system access requires trust. Open source means anyone can audit the code, understand what it does, and verify security claims.
Community Ownership
No single company controls the roadmap. Contributors from 40+ countries shape development. The project outlives any individual or organization.
Innovation Velocity
565+ skills in the ecosystem. Community builds integrations faster than any single team could. Network effects accelerate capability.
Privacy by Default
Self-hosting means no vendor has access to your data. Open source means no hidden telemetry or data collection. You're in control.
"I built PSPDFKit as a product. I'm building clawbot as infrastructure. The difference is profound. Products are owned; infrastructure is shared. The best infrastructure becomes invisible—it just works, and everyone benefits."
The Vision: Personal AI Infrastructure
Peter believes we're entering an era where AI capability is commoditized. In 2026, dozens of frontier models offer comparable intelligence: GPT-4, Claude, Gemini, LLaMA, Mistral, and more. The differentiator won't be which model you use—it will be how you deploy it.
The future Peter envisions is one where everyone has personal AI infrastructure they control:
Not cloud AI that talks to you. Local AI that acts for you.
Not vendor-locked platforms. Interoperable systems you own.
Not subscriptions that extract rent. Infrastructure you deploy once and maintain.
Not opaque black boxes. Transparent systems you can audit and understand.
This vision resonates. The 84,000+ developers using clawbot aren't just users—they're stakeholders in a movement toward digital autonomy. They contribute skills, report bugs, improve documentation, and advocate for privacy-first AI.
What's Next: The Road Ahead
clawbot's growth has been organic—driven by word-of-mouth, not marketing budgets. The community built skills, integrated platforms, and deployed it in production environments worldwide. No VC funding. No acquisition strategy. Just sustainable, community-driven development.
Peter's focus remains on three principles:
1. Keep it open. MIT License is non-negotiable. Anyone can use, modify, and deploy clawbot without restriction.
2. Prioritize privacy. Self-hosting and local models aren't optional features—they're core to the mission.
3. Empower the community. The best ideas come from users solving real problems. The maintainers' job is to enable, not dictate.
The AI landscape will change. New models will emerge. Companies will pivot. Vendors will consolidate. Through it all, clawbot aims to be the constant: open-source infrastructure that works regardless of which models or providers dominate.
Join the Movement
clawbot isn't built by a company—it's built by a community. Developers, power users, and privacy advocates worldwide contributing to the future of personal AI infrastructure.
Star on GitHub Get Started