How LLM Observer Works

Install the CLI and start the local proxy and dashboard with a single command. LLM Observer auto-detects your existing AI tool history (Claude Code, Cursor, Aider, GitHub Copilot, Windsurf, Cline, and OpenAI Codex CLI) and boots the dashboard at http://localhost:4001 - no account, no cloud sign-up. Prefer a native app? Download the signed desktop installer from GitHub Releases instead.

npm install -g llm-observer
llm-observer start

The Control Room is your home screen: today's burn, active subscriptions, budget health score, top apps by spend, and your most expensive sessions - all in one glance. Toggle between Today, Week, and Month to see spending trends.

Every local AI coding conversation is parsed automatically into a searchable session list. Filter by tool or session type, then expand any row to see per-session duration, estimated cost, and token breakdown - all computed on-device from your local history files.

Point your IDE or agent at the local proxy to see every request stream in real time - model, provider, token counts, cost, latency, and status - as it happens, over a live SSE connection.

The optimization engine continuously scans your traffic for waste: oversized models on simple tasks, cacheable duplicate prompts, and anomalous spend spikes. Each recommendation shows the exact monthly savings it would unlock.

Track rolling rate-limit windows for your API providers, with live countdown timers and a 24-hour activity heatmap, so you know exactly how much headroom you have left before you hit a ceiling.

Create a project, issue it a scoped API key, and set a daily budget threshold. LLM Observer tracks spend against that budget in real time and can alert you before you go over.