Live in beta — free during early access

Debug Your AI Agents
Like Chrome DevTools

The agent debugging tool that traces every step — thinking, tool calls, decisions, and outputs. Debug single agents or visualizemulti-agent collaboration across any framework.

Start Debugging Free →Watch Demo

No credit card · 30-day retention · AutoGen, CrewAI, LangGraph

Built for teams building with

LangChainAutoGenCrewAILangGraphCustom Agents

Your Agents Are Black Boxes

AI agents make decisions, call tools, and consume tokens — but you have no idea what's happening inside.

🔍

Where's the bug?

Your agent returned wrong results but you can't trace which step failed. Hours of log spelunking.

💰

Token cost unknown

You see the total bill but not which tool calls or thinking steps consumed most tokens.

🔗

Multi-agent chaos

Agents pass messages, hand off tasks, and fail silently. No tool shows the full collaboration picture.

Every Step, Every Token, Every Decision

The agent debugging tool that gives you full visibility into your agent's behavior.

trace_7f3a2b — user_42 — 23 steps — 4.2s — $0.008FAILED
01
🤔 Think: Determine user intent
"User wants to find pricing for enterprise tier..."
150 tok
02
🔧 Tool: search_pricing_page
search(query="enterprise pricing", limit=3)
320 tok
03
⚠️ Tool result: 404 — page not found
Page was moved to /pricing/enterprise
85 tok
04
❌ Output: "No pricing found" (incorrect)
Should have retried with different query
210 tok
Failed at step 03 (tool result not handled)
Step-by-Step

See every thinking step, tool call, and decision in a timeline

Per-Step Cost

Token and cost breakdown for every step, not just the total

Failure Replay

Replay any failed trace to understand exactly what went wrong

独家功能 · Multi-Agent

See How Your Agents Talk to Each Other

Most monitoring tools only see a linear chain. Agent Debugger visualizes multi-agent collaboration — who called whom, what messages passed between them, and where the handoff failed.

Agent topology mapCross-agent tracesCollaboration errorsMessage context viewer

Works With Any Framework

One SDK. Drop-in instrumentation for every major agent framework.

AutoGen
pip install ad-autogen
CrewAI
pip install ad-crewai
LangGraph
npm i @ad/langgraph
Custom
from ad import trace

Start Debugging Your Agents

Free during beta. No credit card. Full access to traces, replay, and multi-agent visualization.

Join 200+ developers debugging their agents