Langchain Docs

Langchain — Reference Skill-Graph

Comprehensive reference documentation for Langchain Docs.

Sources: https://docs.langchain.com

🧭 How to use this skill-graph

This is a full reference corpus for Langchain — a manual at your disposal. Treat it as ground truth: quote it, don't paraphrase from memory.

• Look something up: scan the Table of Contents (or index.json for a machine-readable map), open the specific reference/… file, quote it + link it.
• Stay grounded: never invent APIs/flags — verify against the reference and cite the file. sources.json tracks provenance + freshness.

📚 Table of Contents

• 📁 0004-agent-server-changelog/ (4)
  • Agent Server changelog
  • 0004 Agent Server Changelog Pt2
  • 0004 Agent Server Changelog Pt3
  • 0004 Agent Server Changelog Pt4
• 📁 0015-langgraph-cli/ (4)
  • LangGraph CLI
  • 0015 Langgraph Cli Pt2
  • 0015 Langgraph Cli Pt3
  • 0015 Langgraph Cli Pt4
• 📁 0035-evaluate-a-rag-application/ (3)
  • Evaluate a RAG application
  • 0035 Evaluate A Rag Application Pt2
  • 0035 Evaluate A Rag Application Pt3
• 📁 0070-organization-and-workspace-operations-reference/ (3)
  • Organization and workspace operations reference
  • 0070 Organization And Workspace Operations Reference Pt2
  • 0070 Organization And Workspace Operations Reference Pt3
• 📁 0088-self-hosted-langsmith-changelog/ (4)
  • Self-hosted LangSmith changelog
  • 0088 Self Hosted Langsmith Changelog Pt2
  • 0088 Self Hosted Langsmith Changelog Pt3
  • 0088 Self Hosted Langsmith Changelog Pt4
• 📁 0112-user-management/ (3)
  • User management
  • 0112 User Management Pt2
  • 0112 User Management Pt3
• 📁 0119-backends/ (3)
  • Backends
  • 0119 Backends Pt2
  • 0119 Backends Pt3
• 📁 0128-claude-specific-guidance/ (4)
  • +
  • 0128 Claude Specific Guidance Pt2
  • 0128 Claude Specific Guidance Pt3
  • 0128 Claude Specific Guidance Pt4
• 📁 0141-subagents/ (3)
  • Subagents
  • 0141 Subagents Pt2
  • 0141 Subagents Pt3
• 📁 0163-long-term-memory/ (3)
  • Long-term memory
  • 0163 Long Term Memory Pt2
  • 0163 Long Term Memory Pt3
• 📁 0167-models/ (3)
  • Models
  • 0167 Models Pt2
  • 0167 Models Pt3
• 📁 0180-build-a-sql-agent/ (3)
  • Build a SQL agent
  • 0180 Build A Sql Agent Pt2
  • 0180 Build A Sql Agent Pt3
• 📁 0192-graph-api-overview/ (3)
  • Graph API overview
  • 0192 Graph Api Overview Pt2
  • 0192 Graph Api Overview Pt3
• 📁 0199-use-the-graph-api/ (3)
  • Use the graph API
  • 0199 Use The Graph Api Pt2
  • 0199 Use The Graph Api Pt3
• 📁 0219-claude-specific-guidance/ (5)
  • +
  • 0219 Claude Specific Guidance Pt2
  • 0219 Claude Specific Guidance Pt3
  • 0219 Claude Specific Guidance Pt4
  • 0219 Claude Specific Guidance Pt5
• 📁 0271-streaming/ (3)
  • Streaming
  • 0271 Streaming Pt2
  • 0271 Streaming Pt3
• Create Connection
• Human-in-the-loop using server API
• Overview
• Agent Server
• Distributed tracing with Agent Server
• Alerts in LangSmith
• Custom instrumentation
• Manage feedback & annotation queues programmatically
• Assistants
• Audit logs
• Authentication & access control
• Self-hosted LangSmith on Azure
• Manage billing in your account
• LangSmith Cloud changelog
• Cloud (SaaS)
• How to compare experiment results
• Include HTTP headers in server logs
• LangSmith control plane
• Cost tracking
• Create a prompt
• ... can add custom routes if needed.
• Manage bulk export destinations
• Data purging for compliance
• Run the agent and wait for the final response
• Enable LangSmith Deployment, Fleet, Insights, and Chat
• Self-host standalone servers
• Deploy your app to cloud
• Find and fix your agent's failures with LangSmith Engine
• LangSmith for Enterprise
• ... database connection and query code
• Run evaluation
• Remember that langgraph graphs are also langchain runnables.
• How to evaluate an application's intermediate steps
• Run an evaluation with multimodal content
• Application-specific evaluation approaches
• Evaluation quickstart
• Event streaming API
• Query traces using the SDK
• Frequently asked questions
• Access & oversight
• Read SKILL.md from disk once; middleware injects into state on first turn.
• Manage agent settings
• Integrate Slack with an agent
• Self-hosted LangSmith on GCP
• Granular billable usage
• Setup guide (legacy)
• Interrupt concurrent
• LangSmith CLI
• LangSmith MCP Server
• Set up the LLM auth proxy
• Custom model providers
• Log LLM calls
• Log traces to a specific project
• Metadata parameters reference
• You can tag a specific dataset version with a semantic name, like "prod"
• How to create and manage datasets programmatically
• If production tag points to commit a1b2c3d4, this is equivalent to:
• Manage LangSmith with Terraform
• Managed Deep Agents CLI reference
• Researcher
• 4. Deploy the agent.
• Managed Deep Agents SDKs
• The trace produced will not have anonymized inputs and outputs
• How to return categorical vs numerical metrics
• How to return multiple scores in one evaluator
• Tracing quickstart
• Set up composite online evaluators
• Redact sensitive data with the OpenTelemetry Gateway architecture
• Profile configuration
• Prompt engineering quickstart
• How to run evaluations with pytest
• Role-based access control
• Keep our test users from the previous tutorial
• Run (span) data format
• Set a sampling rate for traces
• Sandbox SDK usage
• LangSmith Sandboxes
• Enable blob storage
• Disaster recovery for self-hosted LangSmith
• Connect to an external ClickHouse database
• Connect to an external Redis or Valkey database
• Pull the images from the public registry
• Set up SSO with OAuth2.0 and OIDC
• Use an existing secret for your installation (Kubernetes)
• Install Mission Control
• MCP endpoint in Agent Server
• my_agent/agent.py
• Delete Chart
• Get vendor account
• Get a SCIM token
• Streaming API
• Test
• Test a ReAct agent with Pytest/Vitest and LangSmith
• Configure threads
• Trace Claude Code applications
• Trace Gemini Live applications
• Trace query syntax
• Instrument CrewAI and OpenAI
• The above chain will be traced as a child run of the traceable function
• Trace Mistral applications
• Create and run a LangChain application
• Run the pipeline, saving the recording on the way out
• Trace Vercel AI SDK applications (JS/TS only)
• How to evaluate your agent with trajectory evaluations
• How to upload experiments run outside of LangSmith with the REST API
• How to use Studio
• Use webhooks
• View traces
• Configure webhook notifications for rules
• Memory overview
• Contributing to code
• Contributing integrations
• Async subagents
• Changelog
• Pin an exact version for reproducible installs across the fleet
• Applies conventions without prompting
• the task does NOT have access to the command output
• Model providers
• Build a content builder agent
• Context engineering in Deep Agents
• Customize Deep Agents
• Build a deep research agent
• Sandbox
• Todo list
• Going to production
• Harness capabilities
• Interpreters
• Models
• openai.yaml
• Sandboxes
• Poor: too vague for reliable matching
• arxiv-search
• Streaming
• Tools
• export LANGSMITH_API_KEY="your-api-key"
• export LANGSMITH_API_KEY="your-api-key"
• ChatGoogleGenerativeAI integration
• Chat model integrations
• export LANGSMITH_API_KEY="your-api-key"
• Document loader integrations
• Embedding model integrations
• All LangChain JavaScript integration providers
• OpenAI integrations
• Anthropic integration
• OpenAI integration
• Agents
• Component architecture
• LangSmith Deployment
• Human-in-the-Loop
• OpenUI
• Message queues
• Structured output
• Guardrails
• Build a semantic search engine with LangChain
• Model Context Protocol (MCP)
• Prebuilt middleware
• Custom middleware
• Custom workflow
• Build customer support with handoffs
• Multi-agent
• Build a multi-source knowledge base with routing
• Build a SQL assistant with on-demand skills
• Subagents
• Build a personal assistant with subagents
• LangSmith Observability
• Quickstart
• Build a RAG agent with LangChain
• Retrieval
• Short-term memory
• Streaming
• console.log(result);
• Unit testing
• Tools
• Build a voice agent with LangChain
• Memory
• Build a custom RAG agent with LangGraph
• Case studies
• Choosing between Graph and Functional APIs
• Fault tolerance
• Graph execution
• Install LangGraph
• Run a local server
• Quickstart
• Streaming
• Test
• Use the functional API
• Subgraphs
• Use time-travel
• Workflows and agents
• LangChain v1 migration guide
• Changelog
• Context overview
• Contributing to code
• Run tests to ensure your changes don't break existing functionality
• Using standard tests
• Backends
• Per-thread: storage scoped to a single conversation
• Or set it empty so Deep Agents Code ignores a key exported in your shell
• Data locations
• Deep Agents Code
• the task does NOT have access to the command output
• Model providers
• Comparison with Claude Agent SDK
• Context engineering in Deep Agents
• Customize Deep Agents
• Append a system-prompt suffix whenever gpt-5.5 is selected.
• Build a deep research agent
• Sandbox
• Todo list
• Inside your agent's tool:
• Harness capabilities
• Interpreters
• Models
• Profiles
• Grading rubrics
• Sandboxes
• Skills
• arxiv-search
• Skip internal middleware steps - only show meaningful node names
• ❌ Bad: Too many tools
• For large datasets, lazily load documents
• All LangChain Python integration providers
• Fireworks integrations
• Microsoft integrations
• Ollama integrations
• Tool integrations
• Vector store integrations
• Agents
• Component architecture
• Context engineering in agents
• Build a data analysis agent from scratch
• Headless tools
• CopilotKit
• Join & rejoin streams
• Reasoning tokens
• Tool calling
• Stream agent progress and LLM tokens until interrupt
• Long-term memory
• Extract structured content from tool messages
• List of standard content blocks
• Method 3: Custom detector function
• Custom middleware
• Models
• []](reference/0258-doc.md)
• Handoffs
• Agent will call go_back_to_warranty and restart the warranty verification step
• Build a multi-source knowledge base with routing
• Build a SQL assistant with on-demand skills
• Subagents
• Build a personal assistant with subagents
• LangChain overview
• Quickstart
• Build a RAG agent with LangChain
• If desired, specify custom instructions
• Custom state can be passed in invoke
• Build a SQL agent
• Structured output
• Agent Evals
• Access custom state fields
• - Email: foo@langchain.dev
• Build a voice agent with LangChain
• Pass context at invocation time
• Create store with semantic search enabled
• Below is a minimal helper for demonstration purposes.
• Case studies
• Find the checkpoint where an interrupt occurred
• Event streaming
• Separate run and idle limits
• Graph execution
• Functional API overview
• and this is also supported
• Final state: {'vals': ['a:answer for question_a', 'b:answer for question_b']}
• LangGraph overview
• Quickstart
• Redefine the tool node to use the interrupt version
• Use version="v2" for a unified StreamPart format
• Define the structure for email classification
• Continue execution
• 3. Pass in configuration at runtime:
• Define edges
• Outer agent with checkpointer
• Get the tool call
• LangChain v1 migration guide
• LangGraph v1 migration guide
• Unified access to content blocks