AgenticLens

See inside your AI agents.

AgenticLens is a developer tool for visualizing and debugging AI agent workflows from JSONL logs.
It converts raw traces into an interactive workspace with flow graphs, event trees, timelines, replay controls, and step-level inspection.

Beta: AgenticLens is currently in beta.
If you find a bug or unexpected behavior, please report it in GitHub Issues.

Why AgenticLens

Agent runs are hard to reason about from raw logs alone:

• Flows are implicit
• Nested tool chains are noisy
• Latency and token issues are easy to miss

AgenticLens helps you:

• Understand reasoning and tool usage
• Visualize full execution flow
• Detect slow or high-token steps quickly
• Replay sessions step by step
• Inspect each event with metadata and raw JSON

Quick Start

Run directly with npx:

npx agenticlens path/to/logs.jsonl

The app starts and opens in your browser.

You can also launch and upload files/folders directly in the UI.

Product Demo

Watch the full walkthrough:

• AgenticLens Demo Video (MP4)

Screenshots

Flow Canvas

Visual graph of each run with connected steps, slow/high-token badges, and clear transitions between queue ops, user turns, thinking, tools, and results.

Event Tree

Hierarchical breakdown of events and sub-steps so you can expand branches and inspect execution structure in context.

Timeline

Chronological event stream with timing deltas and performance signals for debugging ordering and latency gaps.

Flow + Inspector

Flow canvas paired with inspector details (metadata + raw JSON) so topology and payload stay aligned.

Core Features

• Flow View: graph of execution steps and transitions
• Tree View: nested event hierarchy
• Timeline View: ordered event feed with deltas
• Replay: step through session progression
• Session Summary: totals for time, tokens, steps, nodes, and issues
• Inspector Panel: detailed event metadata and raw JSON
• Anomaly Signals: slow spans, token-heavy steps, bottlenecks

Supported Logs

Current:

• Claude Agent SDK JSONL logs

Planned:

• OpenAI Agents
• Custom agent frameworks
• SDK-based integrations

Use Cases

• Debugging agent workflows
• Understanding reasoning and tool usage
• Optimizing latency and cost
• Comparing different runs
• Building more reliable AI systems

Roadmap

• SDK integration (agenticlens.init())
• Chrome DevTools extension
• Multi-provider support
• Run comparison
• Cloud dashboard

Local Development

npm install
npm run build
node cli/index.js sample-logs.jsonl

License

MIT