Real-time Tracing
Capture tool calls, file operations, shell commands, and artifacts as they happen. Monitor long-running agents with live event streaming.
Execution Timeline
Visualize agent behavior through interactive timelines and execution graphs. Click any event to see full parameters, results, and diffs.
Framework Integrations
LangChain callback, Pydantic AI capability, or plain OpenTelemetry. Chains, model calls, and tools traced automatically — often with a single line.
Self-Hosted
Full data ownership with FastAPI backend and SQLite/PostgreSQL support. No external telemetry, no cloud dependency. Your data stays with you.
Instrument in minutes, not days
A few lines of Python, or none at all with OpenTelemetry. That's all it takes to get full observability.
- LangChain
- Pydantic AI
- OpenTelemetry
- SDK
from trace_agent_sdk import TraceAgentClient
from trace_agent_langchain import TraceAgentLangChainCallback
client = TraceAgentClient("http://localhost:8000")
run = client.start_run("langchain-agent", "Answer user question")
callback = TraceAgentLangChainCallback(run) # one line
agent.invoke({"input": "What's the weather?"}, callbacks=[callback])
run.finish({"status": "success"})
from pydantic_ai import Agent
from trace_agent_sdk import TraceAgentClient
client = TraceAgentClient("http://localhost:8000")
run = client.start_run("my-agent", "What is the weather in Madrid?")
agent = Agent(
"openai:gpt-4o",
capabilities=[run.as_pydantic_ai_capability()], # one line
)
result = agent.run_sync("What is the weather in Madrid?")
run.finish(result.output)
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
# No TraceAgent code in your agent: just export OpenTelemetry traces
provider = TracerProvider()
provider.add_span_processor(BatchSpanProcessor(
OTLPSpanExporter(endpoint="http://localhost:8000/v1/traces")
))
from trace_agent_sdk import TraceAgentClient
client = TraceAgentClient("http://localhost:8000")
run = client.start_run("my-agent", "Fix the login bug")
# Every action is automatically traced
run.record_file_write("auth.py", before_content="...", after_content="...")
run.record_command(["pytest", "tests/"], exit_code=0)
run.finish({"status": "success"})
Modular by design
Install only what you need. Each package works independently. The server also ingests OpenTelemetry traces directly, with no extra package.
SDK
Python client for recording agent actions
pip install trace-agent-sdkuv add trace-agent-sdkLangChain
Automatic tracing for LangChain agents
pip install trace-agent-langchainuv add trace-agent-langchainPydantic AINew
One-line capability for Pydantic AI agents
pip install trace-agent-pydantic-aiuv add trace-agent-pydantic-aiServer
FastAPI backend with SQLite/PostgreSQL
pip install trace-agent-serveruv add trace-agent-serverUI
Interactive dashboard and timeline
pip install trace-agent-uiuv add trace-agent-uiReady to trace your agents?
Get your first trace running in under 5 minutes. Free, open source, self-hosted.