Build Google ADK agents visually in Forjinn
Google ADK lets you build, validate, and deploy agent flows using Google's Agent Development Kit from Forjinn's visual canvas — full edition only.
Google ADK (AutoADK) is Forjinn's integration with Google's Automated Application Development Kit. You design agent flows on a canvas, Forjinn generates the underlying Python code, validates the flow, and executes it — all without leaving the browser.
Google ADK is not available in the open source edition. You need a full or enterprise Forjinn plan to access the Google ADK section of the sidebar. If you're on the open source build, this option will not appear in the menu.
Create a Google ADK project
In the left sidebar, click Google ADK. The dashboard lists your existing ADK projects.
Click Add New. Give your project a name and click Create.
The canvas opens with the ADK node palette on the left. Add nodes by clicking them or dragging them onto the canvas, then wire them together.
Select any node to open its settings panel. Fields vary by node type — an LLM Agent requires a model selection and an instruction; workflow nodes require connections to sub-agents.
Click Validate in the toolbar. Forjinn checks that the graph is complete and all required fields are filled before allowing execution.
Click Execute to run the flow immediately, or use the Deploy option to publish it as a persistent endpoint.
Available node types
The Google ADK canvas provides a specialized set of nodes built around Google's ADK primitives:
Agents
| Node | Description |
|---|---|
| LLM Agent | A reasoning agent backed by any chat model. Supports tools, memory, and knowledge bases. |
| Sequential Agent | Runs a list of sub-agents in order, passing state between each one. |
| Parallel Workflow | Runs multiple sub-agents at the same time and merges their outputs. |
| Loop Workflow | Repeatedly executes a sub-agent until a condition is met. |
| Multi-Agent Router | Routes an incoming request to one of several specialist agents based on content. |
Tools
| Node | Description |
|---|---|
| Google Search Tool | Web search powered by Google Search. |
| Vertex AI Search Tool | Retrieval from a Vertex AI Search data store. |
| BigQuery Analytics | Run SQL queries against BigQuery. |
| REST API Tool | Call any HTTP endpoint and return the result to the agent. |
| OpenAPI Tool | Automatically generate tools from an OpenAPI specification. |
| Built-in Tools | Pre-packaged Google Cloud tool bundles. |
| MCP Toolset | Connect to any Model Context Protocol server to expose its tools. |
| LangChain Tool | Use any LangChain-compatible tool inside an ADK flow. |
| CrewAI Tool | Reuse a tool defined in a CrewAI flow. |
| Agent as Tool | Treat another ADK agent as a callable tool. |
| Long Running Tool | Tools that run asynchronously and return when complete. |
Infrastructure and routing
| Node | Description |
|---|---|
| Session | Manage session state and conversation history. |
| Memory Store | Persist and retrieve information across agent turns. |
| GCS Artifact Service | Read and write files to Google Cloud Storage. |
| Agent Transfer | Hand off a conversation from one agent to another. |
| State Editor | Read and write flow state variables at runtime. |
| Structured Output | Enforce a schema on the agent's final response. |
| Safety | Apply content safety filters to agent inputs and outputs. |
| Grounding | Anchor agent responses to factual sources. |
| Callbacks | Attach lifecycle hooks to agent events. |
| Evaluator | Score agent outputs against criteria during a run. |
Integrations
| Node | Description |
|---|---|
| Bigtable Toolset | Query data from Google Cloud Bigtable. |
| Spanner Toolset | Query data from Google Cloud Spanner. |
| Computer Use Toolset | Give the agent control over a computer environment. |
| GKE Code Executor | Run generated code in a sandboxed Google Kubernetes Engine container. |
| Google Cloud Tools | Batch of common Google Cloud service integrations. |
Chat with an ADK agent
After deploying a project, open the Chat panel from the project dashboard. Type a message and the agent responds in real time using server-sent events for streaming output.
You can also call a deployed ADK agent over the REST API:
curl -X POST https://your-forjinn-instance/api/v1/autoadk/{project-id}/predict \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"question": "Summarize last quarter revenue"}'ADK flows support file uploads. Attach documents, images, or other files in the chat panel and the agent can reference them during execution.
Build agent workflows with Langflow in Forjinn
Langflow is Forjinn's primary canvas for building language-based agent pipelines — drag, connect, and deploy LLMs, memory, tools, and retrievers visually.
Build collaborative multi-agent flows with CrewAI
CrewAI lets you build teams of AI agents with defined roles, goals, and tasks that collaborate to complete complex multi-step work in Forjinn's canvas.