See what your AI agents actually did.
snowdash is an open flight recorder for autonomous AI agents. Your local agent tracks execution streams, logging thought chains, system prompts, tool calls, and terminal results as a visual timeline waterfall.
Interactive Trace Waterfalls.
Stop guessing how your agents spent their tokens. snowdash decomposes the agent execution loop into discrete, timed spans. Inspect system prompts, raw tools payload (e.g. bash commands, python scripts, file reads), and model reasoning trajectories side-by-side.
Launch ConsoleStandardized Agent Capabilities.
Monitor, analyze, and optimize every aspect of your agentic workflows in one unified, real-time dashboard.
Structured Database
Every single trace, prompt, and tool call is stored in a clean Postgres database, completely portable and queryable at any time.
Agent Authentication
Issue secure access tokens for your autonomous agents. Control endpoints, restrict scopes, and isolate trace logs with Row-Level Security.
Real-time Tool Tracing
Capture filesystem mutations, git commits, bash outputs, and Python runtime evaluations as they happen. Zero black boxes.
Cost & Token Analytics
Track token usage and dollar spend in real time across OpenAI, Anthropic, and local LLMs. Detect runaway loops instantly.
Latency Waterfalls
Pinpoint exactly which prompt or tool execution is slowing down your agent. Optimize prompt structures and system instructions easily.
Dataset Generator
Export successful agent trace trajectories directly as fine-tuning datasets with a single click. Turn runtime traces into training data.