I built GhostCode because coding harnesses got smarter and I understood them less. Small open-source terminal harness you can read and fork — not a Claude Code clone.
Local-first, your API keys. Bun + AI SDK. PLAN + BUILD modes, custom TUI, prompts, tools, Ghost.md memory, JSONL sessions.
Good for developers who want to own their harness, not rent one.
AgenticLens helps you understand what your AI agent actually did.
Debugging agent workflows today means reading long, messy logs and guessing what went wrong. There’s no clear flow, no visibility into latency, and no easy way to replay execution.
AgenticLens turns your logs into a visual workspace:
• Flow view
• Timeline
• Replay
It works locally with your existing logs (Claude Agent SDK supported today), with no setup and no data leaving your machine.
