Skip to main content
ServFlow is an agent builder: you define AI agents and the workflows they run, and a Go engine serves them. This guide installs the servflow-pro binary, starts an instance, and walks through the first-run account setup.
Setting up a local environment with tracing and debugging? See Local Environment Setup.

Prerequisites

  • A terminal
  • One of: Docker, Homebrew, npm, or Go — depending on the method you pick

Quick start

Then open http://localhost:8080/dashboard and create your account.
One port serves everything. The engine owns the root path, and --dashboard mounts the builder UI under /dashboard on that same port. Earlier versions used a separate port for the dashboard; they no longer do.

Installation methods

Homebrew (macOS / Linux)

npm

Docker

Run with default configuration:
Run with a custom configuration and persistent storage:
This exposes port 8080 for the engine, the API, and the dashboard, and mounts your configuration and data directories so agents survive restarts.

Binary downloads

Download pre-built binaries from the Releases page. Available for Linux (x86_64, arm64) and macOS (x86_64, arm64).

From source

Building from source requires Go 1.21 or later.

Create a configuration file

ServFlow reads a TOML file. Only server.config_folder is required:
Never commit master_key to version control. Set it with the SERVFLOW_SQLITE_MASTER_KEY environment variable in production.
SQLite stores agents, workflow configs, secrets, integrations, and user accounts. Running without it falls back to file-based storage and disables secrets management, OAuth integrations, and the account system. See the Configuration Reference for every option.

Start the server

Without --dashboard, only the engine and API run — see Running ServFlow.

First run: create your account

The dashboard is gated by default (authentication.mode = "local"). On a fresh instance with no accounts yet:
1

Open the dashboard

Navigate to http://localhost:8080/dashboard.
2

Complete setup

A first-run setup wizard collects instance settings — your workspaces directory and whether to enable tracing. These are written to your config.toml.
Setup settings are read once at startup, so ServFlow asks you to restart the server after saving them.
3

Create the first account

Register with an email and a password of at least 8 characters. The first account is created without authentication to bootstrap the instance; afterwards, creating more accounts requires being signed in.
4

Sign in

Use those credentials on the sign-in screen. Sessions are opaque tokens stored in the database, so they survive restarts.
Setup and first-account registration are only available while no account exists. Once one does, both are closed off.
To disable the gate entirely on a trusted local machine, set mode = "none":

Verify the installation

With the server running, create an agent and its first workflow in the dashboard, or check the CLI can reach your store:

Next steps

Quickstart

Deploy your first API with the dashboard or declaratively.

Configuration Reference

Every TOML option and environment variable.

Running ServFlow

What the server serves, and how to deploy it.

Secrets Management

Store API keys and credentials securely.