An AI agent is only as capable as the systems it can interact with. An agent that can ‘think’ is great, but an agent that can manage code repositories, trigger workflows, or query databases is transformative.<\/p>\n
This shift from chat to action is exactly what Agent Development Kit (ADK)<\/a> was built for. As an open-source framework, ADK provides the flexible primitives developers need to build, test, and deploy these kinds of complex, production-grade agentic workflows.<\/p>\n That\u2019s why we are thrilled to announce a significant expansion to the Agent Development Kit (ADK) ecosystem. We have partnered with leading developer platforms to bring a broad suite of third-party tools and integrations directly into ADK.<\/p>\n With just a few lines of code, these integrations allow your agents to interact seamlessly with the real world. For example, watch how easily an agent can search Hugging Face to discover models and generate code:<\/p>\n Here are the partner integrations available today in the ADK Tools and Integrations<\/a> documentation:<\/p>\n Code & development<\/p>\n Agents can connect directly into your development tools to help you manage code, test APIs, and execute scripts in secure environments:<\/p>\n Daytona<\/a>: Execute code, run shell commands, and manage files in isolated sandboxes.GitHub<\/a>: Analyze code, manage issues and PRs, and automate workflows.GitLab<\/a>: Perform semantic code search, inspect CI\/CD pipelines, and manage merge requests.Postman<\/a>: Manage API collections, test APIs, and generate client code.Restate<\/a>: Resilient agent execution with durable sessions and automatic recovery.Project management<\/p>\n Give your agents access to your team’s project tracking and documentation:<\/p>\n Asana<\/a>: Manage projects, tasks, and goals for team collaboration.Atlassian<\/a>: Manage Jira issues, search Confluence pages, and update team content.Linear<\/a>: Manage issues, track project cycles, and streamline development workflows.Notion<\/a>: Search workspaces, create pages, and manage tasks and databases.Databases<\/p>\n Connect your agents to databases and vector search engines:<\/p>\n Chroma<\/a>: Store documents and retrieve information using semantic search and metadata filtering.MongoDB<\/a>: Query collections, manage databases, and analyze schemas.Pinecone<\/a>: Store data, perform semantic search, and rerank results across vector indexes.Memory<\/p>\n Give your agents persistent, searchable memory across conversations:<\/p>\n GoodMem<\/a>: Automatic or agent-controlled semantic memory with multimodal document support.Qdrant<\/a>: Semantic memory layer for storing and retrieving information with vector search.Observability<\/p>\n Monitor, debug, and optimize your agents in production:<\/p>\n AgentOps<\/a>: Session replays, metrics, and monitoring for ADK agents.Arize AX<\/a>: Production-grade observability, debugging, and evaluation of LLM applications.Freeplay<\/a>: End-to-end observability, prompt management, and evaluation for AI agents.MLflow<\/a>: Ingest OpenTelemetry traces for agent runs, tool calls, and model requests.Monocle<\/a>: Open-source observability, tracing, and debugging of LLM applications.Phoenix<\/a>: Open-source, self-hosted observability, tracing, and evaluation of LLM applications.W&B Weave<\/a>: Log, visualize, and analyze model calls and agent performance.Connectors<\/p>\n Connect your agents to external apps and services through integration platforms:<\/p>\n n8n<\/a>: Trigger automated workflows, connect apps, and process data.StackOne<\/a>: Connect agents to 200+ SaaS providers through a unified integration gateway.AI models & datasets<\/p>\n Give your agents direct access to the AI\/ML ecosystem:<\/p>\n Hugging Face<\/a>: Access models, datasets, research papers, and thousands of Gradio AI applications.Payments<\/p>\n Let your agents work with you to handle your financial operations and billing:<\/p>\n PayPal<\/a>: Manage payments, send invoices, and handle subscriptions.Stripe<\/a>: Manage payments, customers, subscriptions, and invoices.Speech & audio<\/p>\n Add voice and audio capabilities to your agents:<\/p>\n Cartesia<\/a>: Generate speech, localize voices across languages, and create audio content.ElevenLabs<\/a>: Generate speech, clone voices, transcribe audio, and create sound effects.Email and messaging<\/p>\n Let your agents send, receive, and manage email:<\/p>\n AgentMail<\/a>: Dedicated email inboxes for agents to manage conversations and attachments.Mailgun<\/a>: Send emails, track delivery metrics, and manage mailing lists.<\/p>\n ADK also includes built-in integrations with Google Cloud services including BigQuery, Spanner, Pub\/Sub, and more. See the full list in the ADK Tools and Integrations<\/a> documentation.<\/p>\n A unified developer experience<\/p>\n Your agent’s core logic shouldn’t be coupled to its tools. By using the McpToolset<\/a> primitive or plugin<\/a> architecture, you can add any of these third-party integrations into your agent with just a few lines of code. ADK allows you to experiment and find the best tool for the job without ever having to refactor your agent. You don’t need to worry about the underlying implementation details – just copy the configuration for the tool you want, and the ADK handles the rest.<\/p>\n Here is an example of adding the GitHub integration to an ADK agent in Python, which instantly helps you empower your agent to manage repositories and automate workflows:<\/p>\n from google.adk.agents import Agent
\n Python\n <\/p>\n And here\u2019s what the ADK agent looks like in action as it queries and searches through GitHub:<\/p>\n Get Started<\/p>\n Ready to give your AI agents new skills?<\/p>\n
\nfrom google.adk.tools.mcp_tool import McpToolset
\nfrom google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPServerParams
\n
\nGITHUB_TOKEN = “YOUR_GITHUB_TOKEN”
\n
\n# Initialize your agent
\nroot_agent = Agent(
\n model=”gemini-3-flash-preview”,
\n name=”github_agent”,
\n instruction=”Help users get information from GitHub”,
\n # Add a third-party integration by configuring McpToolset
\n tools=[
\n McpToolset(
\n connection_params=StreamableHTTPServerParams(
\n url=”https:\/\/api.githubcopilot.com\/mcp\/”,
\n headers={
\n “Authorization”: f”Bearer {GITHUB_TOKEN}”,
\n “X-MCP-Toolsets”: “all”,
\n “X-MCP-Readonly”: “true”
\n },
\n ),
\n )
\n ],
\n)<\/p>\n