Native 1-line instrumentation for every top agent framework
Whether you're compiling stateful LangGraph state machines, coordinating CrewAI crews, or executing multi-tool OpenAI Assistants, Splyntra automatically intercepts execution loops, tool calls, and model completions.
# 1. LangGraph / LangChain
from splyntra import instrument
instrument() # Auto-instruments all compiled StateGraphs
# 2. CrewAI
from crewai import Crew
from splyntra.integrations.crewai import instrument_crew
instrument_crew(my_crew)
# 3. OpenAI Agents SDK
from splyntra.integrations.openai import instrument_openai
instrument_openai()Engineered for high-throughput autonomous agents
Every capability is built into the OpenTelemetry streaming pipeline with sub-millisecond ingestion overhead.
LangGraph StateGraph & Node Tracing
Inspect state transitions, conditional edge evaluations, loop counters, and parallel node branches directly on visual timelines.
- ✓Captures checkpointed state mutations between nodes
- ✓Traces human-in-the-loop breakpoint interrupts
- ✓Calculates cumulative cost per graph invocation
CrewAI Crew & Agent Task Delegation
Trace hierarchical multi-agent crews. See which agent delegated a sub-task, what prompt was used, and how output was synthesized.
- ✓Agent role, goal, and backstory metadata attached to spans
- ✓Tool execution latency and error recovery tracking
- ✓Multi-agent loop detection and memory inspection
OpenAI Agents & Assistants SDK
Full visibility into assistant threads, function calling schemas, code interpreter executions, and file search vector queries.
- ✓Inspect streaming run steps and tool call payloads
- ✓Prompt caching and reasoning token cost calculation
- ✓Real-time DLP redaction on tool responses
Quickstart Guide for Frameworks
from splyntra import instrument
from langgraph.graph import StateGraph
# Initialize Splyntra
instrument(api_key="splyntra_live_...", project="support-agent")
# Your standard LangGraph workflow
workflow = StateGraph(AgentState)
workflow.add_node("agent", call_model)
workflow.add_node("action", call_tool)
app = workflow.compile()
# Every app.invoke() is automatically traced!
response = app.invoke({"messages": [("user", "Help with my order")]})Frequently Asked Questions
Do I need to rewrite my agent logic to use Splyntra?▼
Can I use Splyntra with custom agent loops not built on a framework?▼
Related Platform Features & Guides
Ready to monitor and secure your AI agents?
Get started in under 3 minutes with zero credit card required. Free tier includes up to 5 projects and community telemetry.