Harden
Local security checks before coding-agent tool execution
Local security checks before coding-agent tool execution
Independent inspections of AI-agent alignment and failures
A Slack-based agent connected to company tools
Interactive audiovisual worlds that evolve during use
Repository-based feature flags documented in Markdown
Domain-adaptive agents for web crawling and research
Developer tools for live audiovisual AI agents
Expose code-discovered APIs through managed MCP servers
Spoken screen guidance with visual annotations
Inspectable PHP applications on persistent Swoole servers