Debug, inspect, and visualize AI agent conversations. Analyze tokens, reasoning, tool calls, latency, and costs from OpenCode and other AI assistants â all in your browser.
npm install -g ai-conversation-analyzer
conversation-analyzer init # interactive backend + OpenCode plugin setup
conversation-analyzer # start dashboard â http://127.0.0.1:4474
From reasoning and tool execution to token usage and conversation timelines â inspect every step visually.
Follow every reasoning step from prompt to final response. See how the model thinks through complex problems.
Visualize every API and tool call with execution time, input/output payloads, and success/error status.
Input, output, reasoning, cache read/write, and total token breakdown with interactive charts.
Grouped conversation turns with expandable details, role badges, latency tags, and per-turn metrics.
Latency breakdown with p50/p95 percentiles, context accumulation curves, and acceleration detection.
Compare multiple agents, identify behavioral patterns, and understand how different models approach tasks.
Abstract connector system â add support for new AI platforms with just a connection, datasource, and parser.
All analysis runs entirely in your browser. Your conversations are never sent to any server.
Track API costs per session, per turn, and per tool call. Understand where your budget goes.
From raw conversation data to actionable analytics in seconds.
Export your AI conversation as JSON from OpenCode or any supported platform.
Upload the file or connect directly to a live session via the API.
The universal parser normalizes the data into a structured conversation model.
Visualize tokens, tools, reasoning, latency, and costs across interactive dashboards.
Every conversation turns into a rich, interactive dashboard.
Stop scrolling through JSON dumps. Start understanding.
MIT licensed. Inspect the code, fork it, extend it. No black boxes.
Runs 100% in your browser. No data ever leaves your machine.
No cloud dependency. Works offline with zero internet connection.
Works with any AI platform through pluggable connectors.
Zero build step. Pure HTML/CSS/JS. Serve and go.
Clean architecture, modular code, easy to contribute.
Found a bug? Have an idea? Want to contribute? We'd love to hear from you.