DOCSEARCHbyAlgolia
Submit Docs

LangSmith API Reference Documentation

website·Indexed Apr 17, 2026

https://docs.langchain.com/

Official API documentation for LangSmith, covering v2 endpoints for agent connections, authentication services (OAuth providers and tokens), deployment management, listener configuration, and integrations. Includes reference for v1 GitHub integrations and Attribute-Based Access Control (ABAC) policies for self-hosted deployments.

Tokens
1.3M
Snippets
4.6K
Records
5.8K
Agent score
50%

What's inside LangSmith API Reference

  1. Overview of LangSmith Deployment and Agent Server

    LangSmith Deployment is a workflow orchestration runtime designed for agent workloads, providing managed infrastructure for durable execution, real-time streaming, and horizontal scaling. It supports the full lifecycle from local development to production deployment and is framework-agnostic (works with LangGraph and other frameworks).

    Key components include:

    • Agent Server: The core engine providing durable execution, assistants (config management), threads (state persistence), and runs (workload execution).
    • Capabilities: Real-time streaming, human-in-the-loop pauses, concurrent input handling, and agent-to-agent (A2A) or tool (MCP) connections.

    Prerequisites: A LangSmith Plus plan or above is required. For BYOC (Bring Your Own Cloud) or premium support, contact sales.

  2. Overview of LangSmith Sandboxes and Access

    LangSmith Sandboxes provide isolated environments for agents to safely execute potentially risky operations, such as running arbitrary code or interacting with the filesystem, without affecting your main infrastructure. This feature is currently in private preview, meaning APIs and features may change. To access Sandboxes, navigate to the LangSmith homepage and select the Sandboxes option from the menu. You can sign up for the waitlist to get early access.

  3. Overview of LangChain Python Tool Integrations

    LangChain Python provides a framework for integrating external tools with AI models. Tools are utilities designed to be called by a model, where inputs are generated by the model and outputs are passed back to it. A toolkit is a collection of tools meant to be used together. The available integrations are categorized into domains such as Search, Code Interpreter, Productivity, Web Browsing, Database, Finance, and Integration Platforms.

  4. Overview of LangChain Chat Model Integrations

    LangChain Python provides integrations for chat models, which accept a sequence of messages as input and return messages as output. This page lists featured models, routers/proxies, and compatibility with the Chat Completions API. Featured models support various advanced features like tool calling, structured output, and multimodal inputs. Routers and proxies allow access to multiple providers through a single API and credential, simplifying billing and enabling automatic fallbacks.

  5. Overview of LangSmith-managed ClickHouse architecture

    LangSmith-managed ClickHouse is a fully managed ClickHouse instance monitored by the LangSmith team, designed to store traces and feedback for self-hosted LangSmith instances. It is recommended for easier management and scaling compared to a fully self-hosted ClickHouse setup.

    Key architectural differences:

    • Private Network: You must establish a private network connection between your LangSmith instance and the managed ClickHouse instance to ensure data security.
    • Sensitive Data Handling: Sensitive fields (inputs, outputs, errors, manifests, extras, events) are stored in cloud object storage (S3 or GCS) within your VPC, not in ClickHouse. Only non-sensitive metadata is stored in ClickHouse.
    • Monitoring: The LangSmith team monitors the instance for metrics like run-ingestion delay and query performance.

  6. Overview of LangChain JavaScript Tool Integrations

    LangChain JavaScript allows models to interact with external utilities called 'tools'. Tools are designed with inputs generated by models and outputs passed back to models. A 'toolkit' is a collection of tools meant to be used together.

    The documentation provides access to a wide range of pre-built tools and toolkits, including search engines (DuckDuckGo, Tavily, SerpApi), cloud services (Azure, AWS Lambda, Google), and automation platforms (Composio, Stagehand). For integration platforms offering unified access to multiple services, Composio is highlighted with 500+ integrations, OAuth handling, and event-driven workflows.

    If you need to extend functionality, you can create your own tools or contribute a new integration to the LangChain ecosystem.

    # Key Integration Platforms
- **Composio**: 500+ integrations, OAuth handling, event-driven workflows, multi-user support.

# Available Tool Categories
- Search: DuckDuckGo, Tavily, SerpApi, SearchApi, Searxng
- Cloud: Azure Container Apps, AWS Lambda, Oracle AI Database
- Google Suite: Gemini Native Tools, Gmail, Calendar, Places, Routes, Scholar, Trends
- Automation: Stagehand AI, Connery, Decodo, GOAT
- Utilities: Python Interpreter, Wikipedia, WolframAlpha, Web Browser

  7. Overview of Deep Agent Harness capabilities

    The Deep Agent Harness is a framework for building long-running agents by combining several core capabilities. These include planning with structured task lists, virtual filesystem access, task delegation via subagents, context management, code execution (with sandboxes), and human-in-the-loop controls. Additionally, agents utilize Skills and Memory for context and instructions.

  8. Overview of Deep Agents SDK

    Deep Agents is a standalone library built on top of LangChain's core building blocks for agents, using the LangGraph runtime for durable execution, streaming, and human-in-the-loop capabilities. It is designed for building agents that can plan, use subagents, and leverage file systems for complex, multi-step tasks. Key features include built-in task planning, file system context management, subagent spawning, and long-term memory. It is suitable for tasks requiring context isolation, permission-controlled file access, and execution in sandboxes.

  9. Overview of Sandbox Integrations in LangChain JavaScript

    Sandbox integrations in LangChain JavaScript allow you to integrate with providers that offer isolated execution environments for running agent-generated code safely. This prevents potential security risks by isolating code execution from the main application environment.

    Available sandbox providers include:

    • Modal: A cloud platform for running code.
    • Daytona: A development environment provider.
    • Deno: A secure runtime for JavaScript and TypeScript.
    • LangSmith: LangChain's platform for building and debugging LLM applications.

    To contribute a new sandbox integration, refer to the guide on implementing sandbox integrations.

  10. Overview of LangChain Python Sandbox Integrations

    Sandbox integrations in LangChain Python provide isolated execution environments for running agent-generated code safely. This prevents untrusted code from affecting the host system. The following sandbox providers are available for integration:

    • Daytona: An open-source development environment platform.
    • Modal: A serverless compute platform.
    • Runloop: A sandbox provider for code execution.
    • AWS AgentCore: AWS's agent core service.
    • LangSmith: LangChain's observability platform which also offers sandbox capabilities.

    To use a specific provider, navigate to its dedicated integration page (e.g., /oss/python/integrations/sandboxes/daytona). If you need to support a new provider, you can contribute by implementing a sandbox integration following the contribution guide.

  11. Overview of LangSmith integrations for LLMs and frameworks

    LangSmith provides integrations for tracing and monitoring applications built with popular LLM providers, agent frameworks, voice AI frameworks, and developer tools. These integrations allow you to automatically trace your application's execution without complex setup in many cases.

    Key integration categories include:

    • LLM Providers: Amazon Bedrock, Anthropic, DeepSeek, Google Gemini, LangChain, Mistral, OpenAI.
    • Agent Frameworks: AutoGen, Claude Agent SDK, CrewAI, Deep Agents, Google ADK, LangGraph, Mastra, Microsoft Agent Framework, OpenAI Agents, OpenTelemetry, PydanticAI, Semantic Kernel, Vercel AI SDK.
    • Voice AI Frameworks: Livekit, Pipecat.
    • Developer Tools: Claude Code, Instructor, n8n, Temporal.

    If you are using LangChain, you can switch between 100+ LLM providers by setting environment variables. LangSmith will automatically trace your application when you initialize a model using LangChain.

  12. Overview of LangSmith Fleet and Agent Builder migration

    LangSmith Fleet is a no-code platform for creating and managing AI agents. It allows users to create agents from templates, connect accounts, and automate routine tasks like drafting emails, summarizing updates, and organizing information. Agents can be used in chat or integrated into workflows (e.g., Slack) with simple approval controls for important actions.

    Important Migration Note: The previous 'Agent Builder' has been rebranded to 'LangSmith Fleet'. All existing agents, configurations, and integrations continue to work without any action required.