Lookspan

Local-first observability for AI agents. MCP-native. See every span your agents emit — on your machine, with zero infra.

Getting started View on GitHub

npx lookspan          # → http://127.0.0.1:3100

Lookspan dashboard demo — trace list with health signals, trace timeline & conversation, multi-agent delegation graph, and replay & LLM-judge

When an AI agent misbehaves — fails, stalls, or quietly burns more tokens than expected — there’s no native way to see what happened step by step. Existing observability tools are cloud-first: they want accounts, API keys, and shipping your production data to someone else’s servers.

Lookspan takes the opposite approach: everything runs on your machine, data never leaves it, and infra cost is zero. Instrument your agent with an adapter (or just POST JSON) and open the dashboard in your browser.

Agent (MCP · LangGraph · CrewAI · OpenTelemetry · HTTP)  →  POST /api/ingest  →  SQLite  →  real-time dashboard

What you can do

Instrument in one line

Drop-in SDKs for the OpenAI and Anthropic clients, an MCP wrapper, Python adapters for LangGraph/CrewAI, or a native OpenTelemetry receiver — no proxy, no code rewrite.

Replay & diff

Re-run a trace’s captured prompt against the same or a different model and diff cost, latency and output.

LLM-as-judge & datasets

Score a response 0–1 with a judge model, or scale it to a whole test set.

Cost, alerts & sessions

Track cost per span, alert on failures or budget overruns, and follow multi-agent causality.