Start

Get started with 2Signal

2Signal is a testing and reliability platform for AI agents. Instrument your agent with the Python, TypeScript, Go, Java, or Ruby SDK, run evaluations on every trace, and see what broke and why.

Quickstarts

Log your first trace

Install the Python SDK and send your first trace to 2Signal in under 5 minutes.

Run your first eval

Set up automated evaluations to measure and improve your AI agent quality.

Workflow

Follow 2Signal's structured workflow to build, evaluate, and improve AI agents:

Step 1

Instrument

Add the @observe decorator and wrap your LLM providers to capture traces.

Step 2

Evaluate

Configure evaluators to automatically score traces for quality, cost, and latency.

Step 3

Observe

Monitor your agents in production with the dashboard, CLI, or TUI.

Core Concepts

Traces

A trace represents one complete execution of your agent — from input to output. Traces contain spans.

Spans

Spans represent individual operations within a trace: an LLM call, a tool invocation, a retrieval step. Spans can be nested (a span can have child spans).

Evaluators

Evaluators score your traces automatically. 2Signal includes 8 evaluator types: LLM Judge, Contains, Regex Match, JSON Schema, Similarity, Latency, Cost, and Custom (via API). Evaluators run asynchronously — they never slow down your agent.

Scores

Each evaluator produces a score (0–1) with an optional label ("pass"/"fail") and reasoning. Scores are attached to traces or individual spans.

Three Interfaces

Use whichever fits your workflow — they all share the same auth and data:

  • Web Dashboard — full GUI at app.2signal.dev
  • CLI — non-interactive commands for scripting and CI/CD
  • TUI Dashboard — interactive terminal dashboard with real-time monitoring

Next Steps

Have questions? Join our community!

Connect with other developers and the 2Signal team.

Join Discord