Distributed Tracing
Follow complete agent chains across model calls, tools, and memory steps with parent-child context.
Open-source observability for AI agents. Trace every step, attribute every dollar, score every output.
Self-host in minutes. Cloud access by waitlist.
Switch views instantly to isolate slow steps, expensive calls, or failing outputs.
Follow complete agent chains across model calls, tools, and memory steps with parent-child context.
Break down spend by agent, task, and provider to identify runaway loops before they hit budgets.
Attach eval scores and confidence signals to every run so teams track behavior, not just latency.
Works with OpenAI Agents SDK, LangChain, CrewAI, AutoGen, and custom orchestrators via OTLP.
Run locally or in your VPC with a single Docker Compose command and keep sensitive traces in-house.
Trigger alerts on cost spikes, score drops, and anomalous agent behavior with actionable run context.
Bring your own infra while keeping the same trace and quality model as managed cloud.
docker compose up agentscope $0
From $99/mo
Custom
Tell us your stack and we will invite teams in rollout order.