Integrations

Trace LangChain (JS and Python) chains, agents, and tools with one callback handler. Every chain run, LLM call, and tool invocation becomes a Spanlens span.

Trace LangGraph node and edge execution with Spanlens. One callback handler captures the full graph topology, parallel fan-out, and tool calls.

Trace CrewAI crews, agents, and tasks with Spanlens. One install captures the full crew execution tree with per-agent cost and latency.

Trace LlamaIndex query engines and agents with Spanlens. One callback handler maps every CBEventType to a span so the trace mirrors your RAG pipeline.

Trace generateText, streamText, generateObject, and streamObject with Spanlens. Two callbacks spread into the AI SDK options log every call automatically.

Trace OpenAI Assistants API threads, runs, and steps with Spanlens. Tool calls and code-interpreter steps render as a span tree per run.

Trace Instructor structured-output calls with Spanlens. Captures the Pydantic schema, retry count, and per-retry token cost.

Connect Flowise visual flows to Spanlens. Flowise calls go through the Spanlens proxy and each node executions becomes a span.

Capture AWS Bedrock model invocations (Claude, Llama, Mistral, Titan) with Spanlens. SigV4 auth handled by your SDK; Spanlens proxies the wire.

Query Spanlens LLM cost, traces, and anomalies from Cursor, Claude Desktop, or Continue via MCP. Public-scope keys stay safe in plaintext IDE config.