Deployment overview

Genkit flows are plain functions. You choose how to expose them over HTTP — either with the built-in flow server or with a framework adapter like Express, Flask, or FastAPI. Once they are wrapped in an HTTP handler they can run anywhere: Cloud Run, Firebase Cloud Functions, Fly.io, AWS, bare-metal servers, or any container platform.

Two deployment patterns

Flow server

Zero-config: Genkit automatically wraps every registered flow as a POST /<flowName> endpoint. Best when you want to expose all flows quickly.

Framework adapter

Mount individual flows inside an existing Express, Flask, FastAPI, or net/http app. Best when you need full control over routes, middleware, and auth.

Flow server

The flow server is the fastest path to a running HTTP API. Each language has a slightly different API.

Node.js
Python
Go

Use startFlowServer from @genkit-ai/express to start an Express server that exposes each flow at POST /<flowName>.

import { genkit } from 'genkit';
import { googleAI } from '@genkit-ai/google-genai';
import { startFlowServer } from '@genkit-ai/express';

const ai = genkit({ plugins: [googleAI()] });

const menuFlow = ai.defineFlow('menuSuggestion', async (theme: string) => {
  const { text } = await ai.generate(`Suggest a menu for a ${theme} restaurant.`);
  return text;
});

startFlowServer({
  flows: [menuFlow],
  port: 8080,
});

The server reads the PORT environment variable if port is not set, and defaults to 3400.

Use genkit_flask_handler from genkit-flask or genkit_fastapi_handler from genkit-fastapi to mount individual flows. For a full app, use the Flask or FastAPI framework directly:

from flask import Flask
from genkit import Genkit
from genkit.plugins.flask import genkit_flask_handler
from genkit.plugins.google_genai import GoogleAI

ai = Genkit(plugins=[GoogleAI()], model='googleai/gemini-2.0-flash')
app = Flask(__name__)

@app.post('/menuSuggestion')
@genkit_flask_handler(ai)
@ai.flow()
async def menu_suggestion(theme: str) -> str:
    response = await ai.generate(prompt=f'Suggest a menu for a {theme} restaurant.')
    return response.text

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=8080)

Alternatively, use FastAPI:

import uvicorn
from fastapi import FastAPI
from genkit import Genkit
from genkit.plugins.fastapi import genkit_fastapi_handler
from genkit.plugins.google_genai import GoogleAI

ai = Genkit(plugins=[GoogleAI()], model='googleai/gemini-2.0-flash')
app = FastAPI()

@app.post('/menuSuggestion', response_model=None)
@genkit_fastapi_handler(ai)
@ai.flow()
async def menu_suggestion(theme: str) -> str:
    response = await ai.generate(prompt=f'Suggest a menu for a {theme} restaurant.')
    return response.text

if __name__ == '__main__':
    uvicorn.run(app, host='0.0.0.0', port=8080)

Use genkit.Handler and net/http directly. There is no separate flow server package — you register handlers on a http.ServeMux and call server.Start.

package main

import (
    "context"
    "fmt"
    "net/http"

    "github.com/firebase/genkit/go/genkit"
    "github.com/firebase/genkit/go/plugins/googlegenai"
    "github.com/firebase/genkit/go/plugins/server"
)

func main() {
    ctx := context.Background()
    g := genkit.Init(ctx, genkit.WithPlugins(&googlegenai.GoogleAI{}))

    menuFlow := genkit.DefineFlow(g, "menuSuggestion",
        func(ctx context.Context, theme string) (string, error) {
            resp, err := genkit.Generate(ctx, g,
                ai.WithPrompt(fmt.Sprintf("Suggest a menu for a %s restaurant.", theme)),
            )
            if err != nil {
                return "", err
            }
            return resp.Text(), nil
        },
    )

    mux := http.NewServeMux()
    mux.HandleFunc("POST /menuSuggestion", genkit.Handler(menuFlow))

    server.Start(ctx, ":8080", mux)
}

Framework adapter pattern

For more control — custom auth middleware, additional routes, or integration with an existing service — mount flows individually inside a framework you already use.

Node.js (Express)
Python (Flask)
Go

import { expressHandler } from '@genkit-ai/express';
import express from 'express';
import { UserFacingError } from 'genkit';
import type { ContextProvider, RequestData } from 'genkit/context';

const authProvider: ContextProvider<{ user: string }> = (req: RequestData) => {
  const token = req.headers['authorization'];
  if (!token) throw new UserFacingError('UNAUTHENTICATED', 'Missing auth token');
  return { user: verifyToken(token) };
};

const app = express();
app.use(express.json());

app.post('/menuSuggestion', expressHandler(menuFlow, { contextProvider: authProvider }));

app.listen(8080);

from genkit._core._context import RequestData
from genkit.plugins.flask import genkit_flask_handler

async def auth_provider(request: RequestData) -> dict:
    token = request.headers.get('authorization')
    if not token:
        raise PermissionError('Missing auth token')
    return {'user': verify_token(token)}

@app.post('/menuSuggestion')
@genkit_flask_handler(ai, context_provider=auth_provider)
@ai.flow()
async def menu_suggestion(theme: str) -> str:
    ...

import "github.com/firebase/genkit/go/genkit"

mux.HandleFunc("POST /menuSuggestion",
    genkit.Handler(menuFlow,
        genkit.WithContextProviders(
            func(ctx context.Context, req core.RequestData) (api.ActionContext, error) {
                token := req.Headers["authorization"]
                if token == "" {
                    return nil, core.NewPublicError(core.UNAUTHENTICATED, "missing token", nil)
                }
                return api.ActionContext{"user": verifyToken(token)}, nil
            },
        ),
    ),
)

Disabling the dev reflection server in production

In development, Genkit starts a local reflection API server on port 4000 (used by the Dev UI). This server must not run in production.

Never run the Genkit reflection server (GENKIT_ENV=dev) in a production deployment. It exposes an unauthenticated API for inspecting and invoking all registered flows and actions.

Set the environment variable before starting your server:

export GENKIT_ENV=production

In Python and Go the reflection server is only started automatically when GENKIT_ENV=dev; removing that variable is sufficient.

Wire format

Every flow — regardless of language — accepts and returns a consistent JSON envelope:

// Request
{ "data": <your flow input> }

// Response
{ "result": <your flow output> }

Streaming responses use server-sent events:

data: {"message": <chunk>}
data: {"message": <chunk>}
data: {"result": <final output>}

Next steps

Firebase Cloud Functions

Deploy flows as HTTPS callable Firebase Functions.

Cloud Run

Run containerised Genkit apps on Google Cloud Run.

Observability

Traces, metrics, and logs for production Genkit apps.

Flows

Learn how flows work and how to define them.

Get Started

Core Concepts

Guides

Plugins

Deployment

Deployment overview

Two deployment patterns

Flow server

Framework adapter

Flow server

Framework adapter pattern

Disabling the dev reflection server in production

Wire format

Next steps

Firebase Cloud Functions

Cloud Run

Observability

Flows

Build docs developers (and LLMs) love

Get Started

Core Concepts

Guides

Plugins

Deployment

Documentation Index

​Two deployment patterns

Flow server

Framework adapter

​Flow server

​Framework adapter pattern

​Disabling the dev reflection server in production

​Wire format

​Next steps

Firebase Cloud Functions

Cloud Run

Observability

Flows

Build docs developers (and LLMs) love

Two deployment patterns

Flow server

Framework adapter pattern

Disabling the dev reflection server in production

Wire format

Next steps