Pricing
Start free. Scale as your agents grow.
All plans include 30-day retention · No credit card for Hobby
Hobby
For solo developers
- ✅ 5,000 traces / month
- ✅ Step-by-step trace timeline
- ✅ Token & cost breakdown
- ✅ Failure replay
- ✅ All frameworks supported
- — Multi-agent topology
- — Alerts & notifications
- — Team collaboration
Pro
For small teams
- ✅ 50,000 traces / month
- ✅ Everything in Hobby
- ✅ Multi-agent topology
- ✅ Real-time alerts (Slack/webhook)
- ✅ Token budget auto-stop
- ✅ 3 team members
- ✅ 90-day retention
Enterprise
For organizations
- ✅ Unlimited traces
- ✅ Everything in Pro
- ✅ Self-hosted deployment
- ✅ SSO / SAML
- ✅ Custom retention
- ✅ Dedicated support
- ✅ Data residency options
Frequently Asked Questions
Can I self-host Agent Debugger? ▼
Yes. Enterprise plan includes self-hosted deployment. We provide a Docker image with D1-compatible SQLite backend.
What frameworks do you support? ▼
AutoGen, CrewAI, LangGraph, LangChain, and custom agent frameworks. Our SDK works with any Python or TypeScript agent via a simple decorator.
How is this different from LangSmith? ▼
LangSmith is tied to LangChain and focuses on single-chain LLM observability. Agent Debugger is framework-agnostic, provides true multi-agent topology visualization, and includes deterministic failure replay.
What is "trace replay"? ▼
Replay shows you a step-by-step playback of any recorded agent session — exactly what the agent saw, decided, and output at each step. Use it to debug failures, audit decisions, and identify hallucination patterns.