INTEGRATION / FRAMEWORKS

See inside every step of your agent graph.

A LangGraph agent that misbehaves is a graph problem: which node, which state transition, which loop. Obsivara traces execution node by node — every LLM call, every tool invocation, every edge taken — so debugging an agent stops being archaeology.

LangGraphEXAMPLE
AGENT RUNS TRACED / 24H3,184
STEPS PER RUN (AVG)11.4
LOOPS FLAGGED / 24H6
ILLUSTRATIVE · EXAMPLE DATA
WHAT YOU GET

LangGraph, fully observable.

Node-level execution tracing

Every node in the graph traced with inputs, outputs, duration, and the LLM calls it made — the full picture of how a run actually flowed.

Loop and recursion detection

Agents that cycle burn money silently. Obsivara flags abnormal loop counts and runaway recursion before a run consumes a day's budget.

State transition visibility

Watch the graph's state evolve across steps and see exactly where a run diverged from the happy path.

Per-graph cost attribution

Cost rolled up per node, per run, and per graph — so you know which agent designs are efficient and which are expensive by construction.

HOW TO CONNECT

Three steps. No code changes.

1STEP 01

Install the Obsivara tracer

Add Obsivara's lightweight callback handler to your LangGraph setup — a few lines, no changes to your graph structure.

2STEP 02

Runs stream in automatically

Every graph execution appears in Obsivara with full node-level traces, timing, and cost.

3STEP 03

Set agent guardrails

Alert on loop counts, step budgets, latency, and failure rates per graph.

WHAT WE MONITOR

Signals tracked out of the box.

  • Node execution time and status
  • Edges taken per run (path analysis)
  • Loop iterations and recursion depth
  • LLM calls and tokens per node
  • Cost per run and per graph
  • State size growth across steps
INTEGRATION FAQ

Common questions.

Do I need to restructure my graphs to use Obsivara?

No. Obsivara attaches via a callback handler — your graph definitions, nodes, and edges stay exactly as they are.

Can Obsivara trace multi-agent LangGraph setups?

Yes. Sub-graphs and multi-agent handoffs are traced hierarchically, so you can drill from an orchestrator down into each agent's steps.

Does it work with both the Python and JS versions?

Yes. Obsivara ships tracers for LangGraph Python and LangGraph.js with identical trace semantics.

Connect LangGraph in minutes.

No credit card · 5-minute setup