Quickstart: Get Your First Agent Trace in 10 Seconds

Getting your first trace into Veritrix takes about ten seconds of actual coding. You install one package, call one function, and every LLM call, tool execution, and agent handoff in your pipeline starts flowing into the dashboard automatically. This guide walks you through the full setup from account creation to your first live trace.

Create your free account

Go to app.veritrix.xyz and sign up. The Free tier gives you one project, 10,000 traced events per month, and 7-day retention — no credit card required.

If you’re setting up Veritrix for a team, start with the 14-day Team trial so you can explore alerting and multi-seat access from day one.

Once you’re logged in, create a new project and copy your API key from the project settings page.

Install the SDK

Install the Veritrix SDK into your Python environment.

pip install veritrix

Veritrix requires Python 3.8 or later. It works with all major agent frameworks out of the box — no additional adapters needed.

Initialize tracing in your agent

Add two lines at the entry point of your agent — before any LLM calls are made. Pass your project name to trace.init so spans are grouped correctly in the dashboard.

from veritrix import trace

trace.init("my-agent")

If you want to pass your API key explicitly (for example, in a CI environment), use the api_key parameter:

from veritrix import trace

trace.init("my-agent", api_key="vx-your-api-key-here")

You can also set your API key via the VERITRIX_API_KEY environment variable and omit the api_key argument entirely — the SDK picks it up automatically.

Run your agent

Run your agent exactly as you normally would. Veritrix instruments in the background — there’s nothing else to change.

from veritrix import trace
from agents import Agent, Runner

trace.init("research-agent")

agent = Agent(name="ResearchAgent", instructions="You are a helpful research assistant.")
result = Runner.run_sync(agent, "Summarise the latest news on LLM reliability.")
print(result.final_output)

from veritrix import trace
from langchain.agents import initialize_agent, load_tools
from langchain.llms import OpenAI

trace.init("langchain-agent")

llm = OpenAI(temperature=0)
tools = load_tools(["serpapi", "llm-math"], llm=llm)
agent = initialize_agent(tools, llm, agent="zero-shot-react-description")
agent.run("What is the square root of the number of days in a leap year?")

from veritrix import trace
from crewai import Crew, Agent, Task

trace.init("crewai-pipeline")

researcher = Agent(role="Researcher", goal="Find accurate information", backstory="You are a diligent researcher.")
task = Task(description="Research OpenTelemetry best practices.", agent=researcher)
crew = Crew(agents=[researcher], tasks=[task])
crew.kickoff()

View your traces in the dashboard

Open app.veritrix.xyz and navigate to your project. You’ll see a live trace waterfall showing every span from your agent run.Each trace entry shows:

Span name — the operation that ran (e.g. planner.plan, researcher.search_docs)
Duration — how long the span took in milliseconds
Token counts — input and output tokens consumed
Status — ok, fail, or retry
Payload — the full input and output for inspection

Here’s an example of what a completed trace looks like in the API and dashboard:

{
  "trace_id": "0x9f3a·b21e·c704",
  "span_count": 6,
  "duration_ms": 4812,
  "total_tokens": 7052,
  "cost_usd": 0.083,
  "spans": [
    {
      "span": "planner.plan",
      "duration_ms": 412,
      "input_tokens": 487,
      "output_tokens": 325,
      "status": "ok"
    },
    {
      "span": "researcher.search_docs",
      "duration_ms": 890,
      "input_tokens": 0,
      "output_tokens": 0,
      "status": "ok"
    },
    {
      "span": "researcher.rerank",
      "duration_ms": 320,
      "input_tokens": 384,
      "output_tokens": 256,
      "status": "ok"
    },
    {
      "span": "coder.write_diff",
      "duration_ms": 1280,
      "input_tokens": 1488,
      "output_tokens": 992,
      "status": "ok"
    },
    {
      "span": "coder.run_tests",
      "duration_ms": 640,
      "input_tokens": 0,
      "output_tokens": 0,
      "status": "ok"
    },
    {
      "span": "reviewer.parse_json ×3",
      "duration_ms": 1140,
      "input_tokens": 1872,
      "output_tokens": 1248,
      "status": "fail",
      "error": "JSONDecodeError at line 3"
    }
  ]
}

When a span fails, Veritrix generates a plain-language root-cause summary — for example:

Reviewer failed to parse Coder’s output as JSON on all 3 retries. The Coder’s tool schema expects diff: string, but it returned diff: object.

Click any span in the waterfall to inspect its full payload — the exact prompt sent to the model, the raw tool input and output, and any error details.

Next steps

Now that you have tracing running, explore the rest of the platform:

How It Works

Understand the three-phase Instrument → Trace → Diagnose model and what Veritrix captures automatically.

Integrations

Set up framework-specific instrumentation for OpenAI Agents SDK, LangGraph, CrewAI, AutoGen, and more.

Get Started

Core Concepts

Tutorials

Integrations

Features

Account & Billing

Quickstart: Get Your First Agent Trace in 10 Seconds

Next steps

How It Works

Integrations

​Next steps

How It Works

Integrations

Next steps