Announcing LangChain Native Support
VigilAgent now automatically traces every LangChain run without writing any manual span decorators.
We are thrilled to announce that VigilAgent now offers first-class, native support for LangChain.
When we first launched VigilAgent, integrating it into your application required manually wrapping your LLM calls and tools with our SDK decorators. While this provided granular control, it was tedious, especially for developers using high-level frameworks like LangChain where the actual API calls are abstracted away.
Zero-Code Tracing
Starting today, if you are using LangChain in Python or TypeScript, you can trace your entire agent execution with exactly one line of code.
By simply initializing the VigilAgentTracer at the start of your application, we automatically hook into LangChain's callback system. We will capture:
- Every PromptTemplate resolution
- Every LLM and ChatModel generation
- Every Tool and Agent execution
- Memory reads and writes
Visualizing the Chain
Because we understand LangChain's internal abstractions, the VigilAgent dashboard will now display your runs exactly as you architected them. You'll see the hierarchical tree of your Chains, Agents, and Tools, rather than a flat list of HTTP requests.
This makes debugging complex AgentExecutors incredibly intuitive. If an agent fails to extract the right information using the SearchTool, you can click directly into the span, see the exact query it generated, the raw HTML returned by the search, and the subsequent error parsing it.
To get started, simply update your @vigilagent/sdk to version 2.0 and follow the new quickstart guide in our documentation!