Faster iteration
Understand behavior and improve your agents quickly.
Real-time observability for AI applications
See your AI applicationexecute in real time.
Beacon turns OpenTelemetry traces from your application into a live execution graph. Follow tool calls, model operations, and execution steps as they happen.
Works with any OpenTelemetry-instrumented application.
Live execution
run_01f8c2e7
Streaming
10:24:01planner selected 4 execution paths
10:24:03llm call completed successfully
Nodes
10
Edges
9
Latency
842ms
Understand every execution
From black box agents to clear execution traces.
Beacon shows what your application is doing, which operations it performs, how the steps connect, how long they take, and where an execution goes wrong.
Real-time traces
See agent executions as they happen.
Tool & LLM visibility
Inspect every tool call, prompt, and response.
Debug failures
Find where complex executions go wrong.
Open standards
Built on OpenTelemetry and designed for your stack.
BUILT FOR REAL ENGINEERING WORKFLOWS
Build with visibility.
Ship with confidence.
Whether you're building an AI workflow, tool-calling application, coding workflow, or another multi-step system, Beacon helps you understand how it executes.
Higher reliability
Catch errors and edge cases before your users do.
OK
Better performance
See latency breakdowns across your entire stack.
Works with your stack
OpenTelemetry native. Framework agnostic.
01span.ingest202
02trace.process18ms
03node.completeok
Open source observability
Built in the open.
Designed for your stack.
Beacon is built around open standards so your agent telemetry stays inspectable, portable, and close to the systems you're already using.
View on GitHub
OpenTelemetry native
Live telemetry
OpenTelemetry → Beacon
$ trace.ingest
trace_id 7f91a2
status received
$ span.process
duration 18ms
status processing
$ node.created
type tool_call
status ✓ complete
$ trace.complete
nodes 04
status ✓
traces / spans / nodes
streamingFrequently asked questions
What you need to know.
A few things you might want to know before sending your first trace to Beacon.
Beacon is an open-source observability tool for AI agents. It helps you see agent executions, tool calls, LLM calls, and the flow between each step in real time.
Beacon collects execution telemetry such as traces, spans, nodes, timing information, and execution status so you can understand what happened during an agent run.
Yes. Beacon is built around OpenTelemetry and is designed to work with telemetry generated through open standards.
Yes. Beacon is designed to fit into existing AI systems without requiring you to rebuild your agent architecture around it.
Yes. Beacon is open source and developed in the open.
You can use Beacon to inspect agent executions, debug failures, understand tool and LLM behavior, identify slow steps, and monitor complex agent workflows.
LLM Calls
Tool Execution
Agent Steps
User Requests
System Events
Clearer debugging
See where executions go wrong.
Faster iteration
Understand behavior as you build.
More reliable AI systems
Reason about complex executions.