How is headless deployment different from an embedded AI tool?

An embedded tool ships its own interface inside your app. Headless deployment exposes the workflow as a pure API endpoint, so your existing frontend handles the user experience and only the generation runs on Wireflow.

Can I chain multiple AI models in one workflow?

Yes. Connect image, video, audio, and language model nodes in sequence on the canvas. The platform runs them in order and passes data between nodes so one call drives the whole pipeline.

How do I authenticate API calls to my workflows?

Generate an sk- prefixed API key from your dashboard settings and send it as a bearer token in the Authorization header. Keys are shown once at creation and can be rotated from the same page.

Does the platform handle long-running jobs like video?

Yes. Submit the run, then poll for completion or register a webhook callback so your backend is notified when the result is ready instead of holding the request open.

Can I update a workflow without breaking my integration?

Yes. Editing pipeline logic in the canvas does not change the endpoint URL or the input and output schema, so you can swap models or add steps and your application code keeps working.

How many AI models are available?

The live node registry has over 170 nodes, roughly 90 of them model-backed, spanning image, video, audio, and language models such as Recraft V4 and Flux 2 Pro that you wire in as swappable steps.

Is every published workflow also callable by an AI agent?

Yes. Each published workflow is exposed as a hosted MCP tool, so an agent can list your workflows, run one with typed inputs, and receive the generated asset URLs back.

Guide

Headless AI Workflow Platform

Name: Headless AI Workflow Platform
Author: Andrew Adams

Design a multi-model AI pipeline on a visual canvas, then run it entirely through the API.

Every published workflow is a REST endpoint and a hosted MCP tool your own app can call.

View API Docs

Connect to Claude

Free to build · no credit card · Read the agent docs

Andrew Adams · Co-Founder & Operations at Wireflow · Updated October 30, 2026

Headless AI Workflow Platformconnecting to live embed…Open workflow →

Loading interactive canvas…

Build visually, deploy as a headless API

A headless AI workflow platform splits the visual editor from the execution layer. You design a multi-model pipeline by connecting nodes on a drag-and-drop canvas, then run each workflow as a callable REST endpoint instead of through a built-in interface. The demo flow on this page is deliberately small: a Text Prompt node feeding a GPT Image 2.5 Flare image node, published and callable exactly as it stands.

Authentication uses bearer tokens with sk- prefixed keys generated in your dashboard, and the same published workflow is also exposed as a hosted MCP tool for agents. You build on the canvas; your own frontend and backend own the user experience.

Headless platform capabilities

REST endpoint per workflow

Call any published workflow with a POST request and a bearer key.

MCP tool for agents

Agents list and run your workflows as typed MCP tools by name.

Async execution

Submit long jobs, then poll or receive a webhook on completion.

100+ model nodes

Chain image, video, audio, and language models in one graph.

Bearer-token auth

Authenticate with sk- prefixed keys you rotate in settings.

Stable contract

Edit pipeline logic without changing the endpoint or its schema.

When a headless setup is not the fit

Wireflow is the generation and execution layer, not the reasoning brain. It runs the models and orchestrates the graph, but it does not write your application logic, decide strategy, or own your interface. If you want a ready-made UI to click through, an embedded tool is a faster start than a headless endpoint.

There are no offline or local-GPU runs, no custom Python nodes, and no local checkpoints. A headless platform earns its place when your own product needs AI generation behind a stable API and you would rather not build or host model infrastructure yourself.

What one API run costs on Wireflow

Every call to a published workflow bills the credits its model nodes use. Starter is $36 a month for 2,660 credits, and Professional is $59 a month for 4,370 credits.

One GPT Image 2.5 Flare image uses 17 credits, $0.23 on Starter or Professional. That is about 156 images a month on Starter and about 257 on Professional.
One 5 second Seedance 2.5 clip uses 135 credits, $1.83 on Starter or $1.82 on Professional. That is about 19 clips a month on Starter and about 32 on Professional.

Shorter clips can cost a bit more per second; you're always charged for the real length.

Run your workflows from Claude

The same published workflows also run from a Claude chat, with no code. Add Wireflow as a connector once:

In Claude, open Customize, then Connectors. Click + Add and choose Add custom connector.
Name it Wireflow and paste https://www.wireflow.ai/api/mcp as the server URL.
Sign in to Wireflow when Claude asks, and approve the access.

The Claude app setup guide has the full steps. Then ask in plain words, for example:

Example prompt: "Use Wireflow to list my workflows, then run my product image workflow once for each of these 10 product names and send me the image links. Tell me the credit cost and wait for my yes before you run it."

With that last line, Claude looks up the price and waits for your yes before it spends any credits. Runs use the credits on your Wireflow account.

More Than Just Headless AI Workflow Platform

Call it from your backend

Publish the flow once, then POST to it from your own server with a bearer key. Your app sends inputs over the AI workflow API and gets asset URLs back.

Every workflow is an MCP tool

Each published workflow also lists as a hosted MCP tool, so an agent can run it with typed inputs. Wire it into any MCP client and call it by name.

Async runs and webhooks

Submit a long job, then poll for completion or catch a webhook when it finishes. The AI orchestration API sequences execution across nodes.

100+ model nodes, one graph

Chain image, video, audio, and language model nodes on one canvas, from Recraft V4 to Flux 2 Pro. The visual node editor wires them in the browser.

Edit logic, keep the endpoint

Swap a model or add a step in the canvas and the endpoint URL and input schema stay the same. Your AI pipeline API integration keeps working untouched.

Open Platform

Build Any AI Workflow

100+

AI Models Integrated

No Watermarks

Full Commercial License

FAQs

It is a platform where you design AI pipelines visually and run them through an API instead of a built-in interface. Your application sends requests, Wireflow runs the workflow on hosted compute, and returns the results.

More From Wireflow

AI workflow builderDesign multi-model pipelines on the visual canvas before you deploy them.ComfyUI alternativeRun node-based AI generation on hosted compute with no local GPU.n8n alternativeAutomate AI media workflows without wiring model infrastructure yourself.AI asset pipelineTurn one flow into a repeatable pipeline for on-brand assets.best AI orchestration APIs for production appsCompare orchestration API patterns for production workloads.

Written by

Andrew Adams · Co-Founder & Operations at Wireflow

Runs client operations and content strategy at Wireflow. Works directly with creative teams and agencies to build production AI workflows.

Content StrategyClient Operations

Explore the headless workflow API

Design multi-model AI pipelines on the canvas and run them as production endpoints. Read the API docs to see authentication, execution, and polling, then open the live flow.

View API Docs

Connect to Claude

Free to buildNo credit cardNo GPU or installCancel anytime