Breadcrumb: Open-Source LLM Tracing Built for Agent Visibility
Breadcrumb is a self-hosted, open-source LLM tracing platform designed for developers who need to understand what their AI agents are doing — without the enterprise overhead of LangFuse or LangSmith.
The platform captures every prompt, response, token count, and cost for every request. An AI monitor watches all traces and automatically flags issues: wrong tool calls, looping agents, oversized models, and cost spikes. The Explore tab lets you query your traces in natural language and get real charts back.
Setup requires just three lines of code. Traces are stored in ClickHouse for fast querying even at scale. Deployment options include one-click Railway deploy or any self-hosted infrastructure. The entire codebase is open source with no usage fees or vendor lock-in.
Breadcrumb integrates natively with Vercel AI SDK and positions itself as the lighter-weight alternative for teams that want trace visibility without the complexity of enterprise observability platforms.
https://breadcrumb.sh
GitHub: https://github.com/joshuaKnauber/breadcrumb
← Back to all articles
The platform captures every prompt, response, token count, and cost for every request. An AI monitor watches all traces and automatically flags issues: wrong tool calls, looping agents, oversized models, and cost spikes. The Explore tab lets you query your traces in natural language and get real charts back.
Setup requires just three lines of code. Traces are stored in ClickHouse for fast querying even at scale. Deployment options include one-click Railway deploy or any self-hosted infrastructure. The entire codebase is open source with no usage fees or vendor lock-in.
Breadcrumb integrates natively with Vercel AI SDK and positions itself as the lighter-weight alternative for teams that want trace visibility without the complexity of enterprise observability platforms.
https://breadcrumb.sh
GitHub: https://github.com/joshuaKnauber/breadcrumb
Comments