AI agents in production lack the observability they need. When something breaks, developers have no trace of which call failed, what the model received, or why costs spiked. AgentLens solves this. Full session traces across every LLM call ā costs, latency, errors, and prompt/completion logs. Framework-agnostic. Proxy-based integration requires zero code changes. Self-hostable via Docker Compose. TypeScript and Python SDKs included. MIT licensed.
Hey Product Hunt š
Developers building AI agents in production face a problem
most observability tools don't solve ā visibility into full
agent runs, not just individual API calls.
AgentLens is built for that gap. The proxy approach means
complete observability with a single environment variable
change. No SDK required to get started. Works with OpenAI,
Anthropic, LangChain, LlamaIndex, or any custom agent.
What's included:
ā Full session trace viewer with span hierarchy
ā Cost analytics by model, agent, and date
ā Real-time live feed via WebSocket
ā Session replay for any past agent run
ā PII scrubbing before data leaves your infrastructure
ā Slack/email failure alerts
ā TypeScript + Python SDKs
Fully self-hostable. MIT licensed. No vendor lock-in.
Would love feedback from anyone running agents in
production ā what visibility matters most to your team?
GitHub: https://github.com/farzanhossan/...
Docs: https://agentlens.techmatbd.com
Report
No reviews yetBe the first to leave a review for AgentLens