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.

Langflow is the default workspace for building AI workflows in Forjinn. You connect node types on an infinite canvas to create pipelines that range from a single LLM with a prompt to multi-step reasoning chains with memory and document retrieval. Everything runs in the browser — no code required to get started.

Create a Langflow workflow

In the left sidebar, click Langflow. The Langflow dashboard lists all your existing flows.

Click Add New in the top-right corner. A blank canvas opens.

Click the + button or right-click the canvas to open the node picker. Search for the node type you want and click it to place it on the canvas.

Drag from an output handle on one node to an input handle on another. A colored line appears when the connection is valid.

Click any node to open its configuration panel on the right. Fill in the required fields — model selection, prompts, credentials, and other settings depend on the node type.

Click Save in the top toolbar, then open the Chat panel (speech bubble icon) to send a test message and see the response.

Available node types

Forjinn's node picker groups nodes by category. The most commonly used categories in a Langflow workflow are:

Chat models

Connect to any LLM provider — OpenAI, Anthropic, Google, Mistral, Ollama, and more. Each model node exposes settings like temperature and max tokens.

Chains

Pre-built sequences for common patterns such as conversational retrieval, summarization, and question-answering over documents.

Memory

Persist conversation history across turns. Options include buffer memory, window memory, and summary memory.

Retrievers

Retrieve relevant chunks from a vector store or document store and pass them as context to your LLM.

Tools

Give your agent capabilities — web search, code execution, API calls, calculators, and custom tools you define.

Output parsers

Structure the model's response into JSON, lists, or custom formats your downstream system can consume.

You can also use Agents nodes (such as Tool Agent or ReAct Agent) inside a Langflow canvas when you want the LLM to autonomously decide which tools to call at runtime.

Deploy your flow

Once your flow works in the test chat, you can expose it as an API endpoint.

Click Deploy in the top toolbar. Forjinn builds the flow and activates an endpoint.

Open the API panel. Copy the prediction URL — it looks like /api/v1/prediction/{flow-id}.

Call the endpoint with a POST request and a JSON body:

{
  "question": "What can you help me with?",
  "chatId": "session-123"
}

Flows are private by default. To allow public access without an API key, open the flow's Settings panel and enable Public Access.

You can embed a Langflow chatbot directly on any website using the Embed option in the flow's share menu. Forjinn generates a script tag you paste into your HTML.

To restrict access, set an API key on the flow from Settings → API Key. Callers must include the key in the Authorization header.