Framework-agnostic tracing

Debug AI agents with a clean, visual trace.

Trace LLM calls, tool calls, latency, logs, and failures. Send raw events from any stack—TraceLens turns them into an explorable timeline + graph.

Works with OpenAI, Anthropic, Gemini, LangChain, custom stacks, and more.

Live trace preview

timeline • metrics • logs

LLMToolsTokensLatencyLogs

Built for tracing and debugging agentic workflows

When runs branch, retry, and call tools, logs aren’t enough. TraceLens gives you a clean graph and the details you need to ship.

Timeline-first UI

See steps laid out by time. Fit-to-screen by default. Click any step to inspect inputs, outputs, tokens, and metadata.

Tool-agnostic

HTTP calls, function calls, and agent events are captured as raw events. The backend turns them into consistent steps for the UI.

Multi-tenant workspaces

Keep traces separated by workspace. Invite teammates with roles. Add tags, annotations, and share read-only links.

Secure ingest

Per-workspace API keys, strict payload validation, rate limiting, audit logging, and org-scoped trace access.

Exports & sharing

Export a trace as JSON or a graph as PNG. Create a share link for external review without giving account access.

Usage visibility

Org usage dashboard, monthly limits, and top users—built-in primitives for SaaS plan enforcement.

Create a workspace

Instrument your agent

Wrap your entrypoint with trace_agent. Your code stays generic; the SDK collects events automatically.

Debug in TraceLens

Browse traces in timeline or graph view. Inspect steps, logs, and metrics. Share a link when needed.

Create a workspace, generate an ingest key, and wrap your agent with trace_agent. TraceLens handles the rest.