Vessel
See what your agents are actually doing
Start new thread

Vessel - See what your agents are actually doing

by•
Vessel is a local-first observability proxy for LLM traffic: capture, search, replay, and compare every prompt across Ollama, OpenAI, and Anthropic and all other backends. One binary, no account, nothing leaves your machine. It also hosts a small MCP server so your coding agents can query traces in real-time - very useful for troubleshooting

Add a comment

Replies

Best
Maker
📌
I built it because I was running multiple local agents and wanted to track what they were actually doing. Now each one is tagged, measured, and comparable. It's not a replacement for the likes of Langfuse/Langsmith/Logfire and other observability platforms. Vessel is more like a debugging proxy. Fiddler/Proxyman/DevTools-network-tab territory, but for LLM traffic. No SDK, no code changes: you point a base_url at it, so it also works with clients you can't modify (Open WebUI, IDE plugins, whatever). Local web UI to inspect what actually went over the wire, plus replaying a captured request against a different model to compare.