Building AI in the Open
Open-source tools, open standards, open distribution. We built the framework, the registry, the trust standard, and the UI protocol, then open-sourced everything. This is how and why.
Explore the topics
Declarative AI Applications
Build AI apps with schemas and skills instead of code. Upjack is the open-source framework that makes AI applications declarative, define what the app does, not how.
5BAI Tool Distribution
Package, scan, and distribute MCP servers safely. mpak is the open-source registry with security scanning, like npm for AI agent tools.
5CAI Security Standards
The MCP Trust Framework is the first security standard for AI agent tools. 25 automated controls, trust scoring, and enterprise governance for MCP server deployment.
5DProtocol-Native Interfaces
UIs that emerge from protocols, not custom code. Synapse renders agent interfaces from MCP tool schemas. SkillThis captures domain expertise visually.
What we believe about this
Applications Won't Be Coded. They'll Be Declared.
JSON Schema defines data. Markdown defines logic. The AI agent is the runtime. This is the post-code application model.
Thesis #2MCP Apps: How We'll Build Software in the AI Era
Applications will be composed from MCP servers, not coded from scratch. The traditional stack is being replaced by a composition layer that wires tools together through an AI runtime. Upjack proves it.
Thesis #8Open Source Is the Only Trust Model for AI Infrastructure
Proprietary AI infrastructure is a liability. The Windsurf incident proved it. The only safe infrastructure is inspectable, forkable, self-hostable. NimbleBrain open-sources everything.