OpenAI Agents SDK（Python）

The OpenAI Agents SDK is a lightweight yet powerful framework for building multi-agent workflows. It is provider-agnostic, supporting the OpenAI Responses and Chat Completions APIs, as well as 100+ other LLMs.

课时

• 介绍
• 快速开始
• config
• 文档
  • agents
  • 沙盒智能体
    • 快速入门
    • 概念
    • 沙箱客户端
    • 智能体记忆
  • 实时智能体
    • quickstart
    • guide
  • 语音智能体
    • quickstart
    • pipeline
    • tracing
  • 模型
  • tools
  • guardrails
  • running_agents
  • streaming
  • multi_agent
  • handoffs
  • results
  • human_in_the_loop
  • 会话
    • index
    • sqlalchemy_session
    • advanced_sqlite_session
    • encrypted_session
  • context
  • usage
  • mcp
  • tracing
  • 测试
  • visualization
  • repl
  • release
• Codex Execution Plans (ExecPlans)
• .agents
  • SDK Maintainer References
    • Agent Definition and Run Context
    • Conversation State Ownership
    • Function and Output Schema
    • Local MCP Server Lifecycle
    • Model and Provider Boundaries
    • Realtime Session Lifecycle
    • Realtime Tracing Architecture
    • Run Item Lifecycle
    • Runner Lifecycle
    • RunState Schema and Resume Boundary
    • Sandbox Runtime Boundary
    • Session Persistence
    • Tool Execution Lifecycle
    • Tool Identity and Routing
    • Tracing Lifecycle
    • Voice Pipeline Lifecycle
  • skills
    • code-change-verification
      • Code Change Verification
    • docs-sync
      • Docs Sync
      • references
        • Doc Coverage Checklist
    • examples-run-analysis
      • Examples Run Analysis
    • final-release-review
      • Final Release Review
      • references
        • Release Diff Review Checklist
    • implementation-final-review
      • Implementation Final Review
      • references
        • High-risk independent review
        • Independent Reviewer Brief
    • implementation-kickoff
      • Implementation Kickoff
    • implementation-strategy
      • Implementation Strategy
    • maintainer-review
      • Maintainer Review
      • references
        • Maintainer Evaluation Framework
    • openai-knowledge
      • OpenAI Knowledge
    • pr-draft-summary
      • PR Draft Summary
    • release-candidate-prep
      • Release Candidate Preparation
    • runtime-behavior-probe
      • Runtime Behavior Probe
      • references
        • Common Error Cases
        • OpenAI Runtime Patterns
        • Reporting Format
        • Validation Matrix
    • sensitive-logging-audit
      • Sensitive Logging Audit
      • references
        • Python sensitive logging validation
    • test-coverage-improver
      • Test Coverage Improver
• .github
  • Publishing a release
  • scripts
    • Repository skill tests
• OpenAI Agents SDK
  • Agents
  • Configuration
  • Context management
  • Examples
  • Guardrails
  • Handoffs
  • Human-in-the-loop
  • Model context protocol (MCP)
  • Agent orchestration
  • Quickstart
  • Release process/changelog
  • REPL utility
  • Results
  • Running agents
  • Quickstart
  • Streaming
  • Testing
  • Tools
  • Tracing
  • Usage
  • Agent visualization
  • Models
    • LiteLLM
  • realtime
    • Realtime agents guide
    • Quickstart
    • Realtime transport
  • Agents module
    • Memory
    • Sandbox
    • extensions
      • Handoff prompt
      • LiteLLM Models
      • experimental
        • codex
          • Codex Options
          • Output Schema File
          • Thread Options
          • Turn Options
        • hosted_multi_agent
          • Model
      • memory
        • AdvancedSQLiteSession
        • Async Sqlite Session
        • DaprSession
        • EncryptedSession
        • MongoDBSession
        • RedisSession
        • SQLAlchemySession
    • memory
      • OpenAI Conversations Session
      • OpenAI Responses Compaction Session
    • models
      • OpenAI Agent Registration
      • OpenAI Chat Completions model
      • Reasoning Content Replay
    • realtime
      • Realtime Configuration
      • Realtime Events
    • run_internal
      • Model Provider Lifecycle
    • sandbox
      • Workspace entries
      • Manifest
      • Permissions
      • Runtime Agent Preparation
      • SandboxAgent
      • SnapshotSpec
      • capabilities
        • Capabilities
        • Capability
        • Compaction
        • Filesystem
        • Memory
        • Shell
        • Skills
        • tools
          • Apply Patch Tool
      • sandboxes
        • Docker sandbox
        • Unix local sandbox
      • session
        • Base Sandbox Session
        • Manifest Application
        • Sandbox clients
        • SandboxSessionState
        • SandboxSession
    • voice
      • models
        • OpenAI Model Provider
        • OpenAIVoiceModelProvider
  • sandbox
    • Sandbox clients
    • Concepts
    • Agent memory
  • Sessions
    • Advanced SQLite sessions
    • Encrypted sessions
    • SQLAlchemy sessions
  • voice
    • Pipelines and workflows
    • Quickstart
    • Tracing
  • zh
    • 代码示例
    • 会话
    • models
      • LiteLLM
    • realtime
      • 实时传输
• Running the example suite
  • Common agentic patterns
  • Financial Research Agent Example
  • mcp
    • MCP Filesystem Example
    • MCP getallmcptools Example
    • MCP Git Example
    • MCP Manager Example (FastAPI)
    • MCP Prompt Server Example
    • MCP SSE Example
    • MCP SSE Remote Example
    • MCP Streamable HTTP Remote Example
    • Custom HTTP Client Factory Example
    • MCP Streamable HTTP Example
    • MCP Tool Filter Example
  • Model provider examples
  • realtime
    • Realtime Demo App
    • Twilio SIP Realtime Example
    • Realtime Twilio Integration
  • Research bot
    • sample_outputs
      • Comprehensive Guide on Best Surfboards for Beginners: Transitioning, Features, and Budget Options
      • OpenAI Agents SDK（Python）
  • Sandbox examples
    • 文档
      • Credit Note Example Repo
        • Task
      • skills
        • credit-note-fixer
          • Credit Note Fixer
    • Cloud Sandbox Extension Examples
      • daytona
        • NASA Spending Text-to-SQL Agent
          • schema
            • USAspending Glossary
            • OpenAI Agents SDK（Python）
            • tables
              • spending
      • Temporal Sandbox Agent
    • Healthcare support
      • policies
        • Auth Review Queue Routing
        • Billing After Consult FAQ
        • Blue Cross Benefits Reference
        • Blue Cross PPO Prior Authorization
        • Blue Cross Referral Rules
        • Commercial Eligibility Checklist
        • Human Escalation Policy
        • Knee Surgery Medical Necessity
        • Orthopedic Imaging Policy
        • Outbound Fax Packet Requirements
        • Patient Messaging Guidelines
        • Referral Pending SOP
        • Scheduling Hold Policy
      • skills
        • prior-auth-packet-builder
          • Prior Auth Packet Builder
    • tutorials
      • Dataroom metric extract
      • Dataroom Q&A
      • Repo code review
      • Sandbox resume
      • Vision UI reproduction
        • skills
          • playwright
            • Playwright
  • tools
    • skills
      • csv-workbench
        • CSV Playbook
        • CSV Workbench
  • voice
    • Streamed voice demo
• Packaged integration tests
• 源码
  • agents
    • Realtime
    • sandbox
      • instructions
        • How you work
      • memory
        • prompts
          • OpenAI Agents SDK（Python）
          • OpenAI Agents SDK（Python）
          • OpenAI Agents SDK（Python）
          • OpenAI Agents SDK（Python）
• Tests
  • fixtures
    • RunState compatibility corpus
  • Realtime session test groups

开始学习 → Agent Definition and Run Context