Skip to main content
ServFlow is an agent builder: you define AI agents and the workflows they run, and a Go engine serves them. The dashboard is the visual way to do that — you create an agent, add workflows to it, and wire each workflow up on a canvas. In this tutorial you’ll create an agent, give it a “Hello World” HTTP workflow, and call the endpoint.
Prefer configuration files? See Deploy your first API declaratively. Both produce the same objects.

​
Prerequisites

ServFlow Pro running with the dashboard enabled:
If you haven’t installed it yet, see Installation — and note that the first time you open the dashboard you’ll be asked to complete setup and create an account.

​
What you’ll build

An agent named hello-agent with one workflow that answers GET /hello-world with a JSON greeting.

​
Step 1: Open the dashboard and sign in

Navigate to:
The dashboard is served under /dashboard on the same port as the engine — there is no separate dashboard port.
Sign in with the account you created during first run. On a brand-new instance you’ll be prompted to complete setup and register instead; see First run.

​
Step 2: Create an agent

Agents are the top-level object in ServFlow — a workflow always belongs to one. On the landing screen choose Create New. The New agent dialog appears. Fill it in:
  1. Agent name — enter hello-agent
  2. Under New, choose the option to start from scratch
The same dialog lists Templates — prebuilt agents you can start from instead of an empty one. They’re the fastest way to see a working multi-workflow agent.
Confirm to create the agent. ServFlow then prompts you to add its first workflow.

​
Step 3: Add a workflow

In the workflow dialog:
  1. Name — enter hello-world
  2. Entry type — choose API (HTTP)
  3. Agent — leave it set to hello-agent
The entry type decides how the workflow is reached, and how it is attached to the agent: Since this is an HTTP endpoint, API (HTTP) attaches it to the agent as a webhook.

​
Step 4: Configure the entry

The workflow opens on the builder canvas with an HTTP entry node already created, pre-filled from the workflow name:
  • Method — GET
  • Listen path — /hello-world
Click the entry node to change either. The listen path is the URL the workflow answers on.

​
Step 5: Add a response

Click the + button below the entry node. You’re offered three kinds of step:
  • Actions — do work: call an AI model, query a database, make an HTTP request
  • Branch — route conditionally
  • Responses — return something to the caller
Choose Responses, then configure it:
  1. Set the status code to 200
  2. Set the response body to:
Make sure the entry node connects to this response node — the connection is what makes it run.

​
Step 6: Save

Click Save. The workflow is stored and served immediately; there’s no separate deploy step and no restart.

​
Step 7: Call your endpoint

The engine owns the root path, so the workflow answers at its listen path directly:
You’ve built an agent with a live HTTP endpoint.

​
What you just created

  • An agent — hello-agent, the container that owns this work
  • A workflow attached to it as a webhook, because it has an HTTP entry
  • Inside that workflow, an entry wired to a response
Adding an action between the entry and the response is how a workflow starts doing real work — calling a model, reading a database, or hitting an API. Everything here is also reachable from the CLI: this agent and workflow can be listed, exported, and edited with servflow-pro resource.

​
Next steps

Deploy declaratively

Define the same workflow as a file and deploy it from the CLI.

Actions

Add real work to your workflow.

Running ServFlow

Understand what your instance serves and how to deploy it.

Secrets Management

Store API keys your agent needs.