AI-powered logging platform: a small SDK sends structured logs; the platform runs semantic search over them and drafts fixes as pull requests.
What I did
Built the SDKs (Node, Python), the semantic log search, and an MCP server so coding agents (Claude Code, Cursor) could query production logs directly.
Grew it to 30+ active users and $11K in startup credits (Daytona, AWS); wound the product down in 2026 — its agent-infrastructure ideas became Coregit.
