AXON

open ingest · no sdk required · postgres native

Every span. Every dollar.
Every failure — seen live.

axon is a drop-in observability spine for ai agents: runs and spans, token burn per model, latency percentiles, and failure forensics — one postgres table away. below is a real fleet, streaming right now.

llmtooldbrouterchain error

emit a span from anywhere

curl -X POST $HOST/api/axon/spans \
  -H 'content-type: application/json' \
  -d '{"agent":"my-agent","task":"triage #1","spans":[
    {"name":"llm.plan","kind":"llm","model":"gpt-5.2",
     "tokensIn":1240,"tokensOut":310,"durationMs":840}]}'

any agent, any framework, any language — json in, truth out. spans roll up into runs with cost, tokens, and status.

runs

—

spans

—

tokens

—

cost

—

p95 span

—

run errors

—

the hum — spans of the last minute, falling into history

recent runs — click for the waterfall

  • waking the fleet…

burn by model

    fleet health

      failure feed —

        span density — 1h