Project
Stackglass
Observability and verification layer for AI-assisted development — evidence instead of guesses.
active · ai / developer-tools / systems · TypeScript
Evidence plane
tests
observed
contracts
observed
failures
observed
docs drift
flagged
Overview
Observability and verification layer for AI-assisted development — evidence instead of guesses.
Why it exists
Agents write fluently and still guess about state, tests, failures, history, and contracts. Stackglass feeds them evidence.
Problem
Without an observability plane, coding agents invent repository truth from incomplete context.
Approach
Local-first evidence about state, tests, failures, configuration, contracts, and documentation — consumable by humans and agents.
Architecture
- CLI and MCP surfaces for agent tooling
- Evidence collection over repo and runtime signals
- Verification-oriented outputs for Cursor workflows
Current status
Active open-source agent observability work.
Related
AftermathExecution-backed verification layer for coding agents — durable receipts over narration.CartographerTypeScript-native navigation system for software architecture — know where you are in the code.ArcframeLocal-first repository intelligence for Cursor MCP — graph, blast radius, evidence-backed context.ChimeraDecentralized peer-to-peer distributed computation and rendering grid in Rust.ParallaxPolyglot migration and universal execution runtime — capture state in one language, restore in another.