AI-Flow
AI-Flow is a visual, drag-and-drop canvas for chaining AI models from OpenAI, Anthropic, Google and Replicate into repeatable image, video, text and document pipelines. Pay with platform credits or plug in your own provider keys, with no subscription.
What is AI-Flow?
AI-Flow is a visual platform for assembling AI workflows. A flow is a directed graph in which every node does one job: taking an input, calling a model, transforming data, controlling execution, reaching an outside service or presenting a result. You drag nodes onto a canvas, connect an output handle to a compatible input handle, configure prompts, files and model settings, then run a single node while testing or use Run All for the whole graph. Conditional branches, loops and reusable subflows let a pipeline grow without collapsing into one unreadable diagram.
Its defining trait is provider neutrality. OpenAI, Anthropic, Google Gemini, Replicate, fal.ai, Stability AI, ElevenLabs, DeepSeek, xAI and OpenRouter all appear as nodes in the same canvas, so a single run can pass through Claude Sonnet 4.6 for reasoning, GPT Image 2 or Nano Banana 2 for stills, Seedance 2, Veo 3.1 or Kling for motion, and Music 1.5 or Speech 2.6 for audio. Compute is paid for in one of two ways: platform credits, which need no vendor account at all, or bring your own key, where a credential pasted once into an encrypted Secure Store is picked up automatically by every compatible node and the model call is billed by the provider instead. AI-Flow argues that hosted tools add a three to fivefold markup on provider pricing, and the cost of each node is shown before you run it.
The same flow can be used four ways, which is what separates AI-Flow from a prompt playground: as a canvas you keep editing, as an App where a colleague fills in a short form and hits Run, as a subflow reused inside a larger automation, or as a REST endpoint called from your own backend through the API Builder. Notion, Airtable, Zapier and Make have dedicated integration nodes, an HTTP Request node covers everything else, and MCP servers let assistants such as Claude Code and Cursor run your saved flows.
A hosted version runs at app.ai-flow.net. The core is also open source under the MIT licence, self-hostable through Docker Compose, a local install or a Windows desktop build, which makes AI-Flow unusually easy to leave.
What it does
- Chain several AI models from different providers inside one visual workflow
- Generate and edit images, video, text, audio and 3D assets in the same canvas
- Connect your own provider API keys and pay each vendor at its published rate
- Turn any workflow into a form-based App a colleague can run from a shared link
- Expose a workflow as a REST endpoint or trigger it through a webhook
- Extract, summarise and reformat content from uploaded documents and web services
- Run the open-source build on your own machine with Docker or the Windows executable
When to use AI-Flow / When not to
A quick filter to help you decide if AI-Flow is the right fit.
When to use AI-Flow
- Content and video teams producing AI assets at volume, who need the same pipeline to run again on every new brief
- Creators and small teams who already hold OpenAI, Anthropic or Replicate accounts and want to pay providers at their published rates
- E-commerce merchants and merchandisers turning a single product shot into photography, mockups and ad visuals
- Developers who want to expose a finished workflow as a REST endpoint or a webhook inside their own product
- Teams with non-technical colleagues, who can be handed a workflow as a form-based App without ever opening the canvas
When not to use AI-Flow
- Organisations bound by HIPAA, FISMA or GLBA, since the terms of service exclude those uses outright
- Anyone under 18, who the terms of service bar from registering for the service
- Buyers who need a signed DPA, a published subprocessor list or a SOC 2 or ISO 27001 certificate, as none of these is published
- Teams with a data residency requirement, because no hosting country or region is declared anywhere on the site
- Mobile-first users, as there is no iOS or Android app: the product is a web application plus a Windows desktop build
How to use AI-Flow
A typical end-to-end flow, from setup to results.
- Open the hosted application and create an account; no credit card is required to start
- Follow the five-minute Getting Started guide, which walks through an App, a template, the canvas and a first run
- Pick a starting point: a ready-made App for a quick result, a template to adapt, or a blank flow to design from scratch
- Add nodes from the library and connect an output handle to a compatible input handle
- Configure the prompts, files and model settings each node needs
- Optionally open Settings, then the Secure Store tab, paste a provider key and select Validate so every compatible node uses it
- Check the credit cost shown on a node, then run that node alone to test it before using Run All on the whole graph
- Define clean interfaces with Flow Input and Flow Output nodes so the workflow can be reused
- Publish the result: share it as a form-based App, reuse it as a subflow, or expose it as a REST endpoint through the API Builder
- To self-host instead, download the Windows executable, run docker-compose up -d from the docker directory, or install locally with npm and poetry
Pros & Cons
Pros
- Genuine bring-your-own-key support: model usage is billed by the provider at its published rate, with no platform markup
- Several providers can be mixed inside one workflow instead of being limited to an imposed catalogue
- No subscription at all; credits are bought when needed and the platform repeats this across the site
- A permanent free tier with 25 welcome credits and 20 free runs a day, and no credit card required
- The core is open source under the MIT licence and can be self-hosted with Docker, a local install or a Windows build
- One workflow can be reused as an App, a subflow or an API endpoint without being rebuilt
- Credit cost is displayed per node before execution, and public documentation covers the product in detail
Cons
- No pricing page on the marketing site: prices appear only inside the application and on the product photography page
- No postal address and no full legal entity name are published anywhere on the site
- A single email address serves support, legal and GDPR requests alike
- No DPA, no subprocessor list and no security certification such as SOC 2 or ISO 27001
- No hosting country or region is declared, and generated media is kept only seven days on the free tier
- Pay-as-you-go credits expire after a year without a new purchase, and all purchases are non-refundable
- No mobile app, an interface available only in English and French, and legal documents dating from 2023 and 2024
Pricing & Plans
AI-Flow offers a permanent free tier: 25 welcome credits on sign-up and 20 free runs per day on bring-your-own-key nodes, inputs and tools, with no credit card required. The terms of service also mention a 20-credit free trial for new registrants. Beyond that there is no subscription; access is bought as credit packs. The lowest publicly listed entry point is 4.99 USD for 200 credits on the product photography page, while the cheapest pack in the application itself is 9.99 USD for 500 credits. When your own provider key is connected, model usage is billed by that provider and AI-Flow charges only a small amount of credits for the resources around the run. All payments are stated to be in US dollars.
- 25 welcome credits and 20 free runs per day on BYOK nodes
- inputs and tools
- with no credit card required
- 4.99 USD for 200 credits
- covering 3 photoshoots with high-resolution output and commercial use
- 9.99 USD
- with 2 nodes running in parallel
- 19.99 USD
- with 4 nodes running in parallel
- 29.99 USD
- with 6 nodes in parallel and HTTP trigger
- flagged as the most popular pack
- 99.99 USD
- with 10 nodes in parallel
- HTTP trigger and early access to new features
Data, GDPR & hosting
A consolidated view of how AI-Flow handles your data.
GDPR overview
The privacy policy carries a dedicated section titled Your Data Protection Rights Under General Data Protection Regulation (GDPR). It tells EU and EEA residents that AI-Flow will take reasonable steps to let them correct, amend, delete or limit the use of their personal data, that they can ask what data is held and request its removal by writing to support@ai-flow.net, that identity may be verified first, and that they retain the right to complain to their national supervisory authority. The privacy policy names France as the governing country and the terms of service apply French law. Beyond that section, implementation is thin: no Article 27 representative and no data protection officer is named, no data processing agreement is offered, no subprocessor list is published, and the site never states in so many words that it is GDPR compliant.
Who owns the data?
The privacy policy of 26 January 2024 states that AI-Flow collects your email address and usage data such as IP address, browser, pages visited and time spent, and that third-party Service Providers may process it on the company's behalf. The terms of service claim ownership of the platform itself, its source code, databases and marks, and go further on one point: any Submission you send in, meaning a question, comment, suggestion, idea or piece of feedback about the Services, is assigned to the company along with all its intellectual property rights. On generated content the position is favourable to the user: the Product Shots page confirms that uploaded photos stay private to your account and that images you generate are yours to use commercially.
Reuse rights
For the outputs you create, the site is explicit that reuse needs no further permission: photos generated through the product photography workflow are yours to use for your store, your ads, your social channels and your marketing materials, and the same page confirms that uploaded images are neither shared publicly nor used to train models. The terms grant you a non-exclusive, non-transferable and revocable licence to access the Services and to download or print portions of AI-Flow's own content, but only for personal use or an internal business purpose. Everything beyond that, including republishing, redistributing or reselling the platform's content and marks, requires written permission. The reverse direction is stricter: feedback and suggestions you send to the company become its property. No clause in the terms claims ownership of the workflows you build or of the model outputs they produce.
Data retention & training
Hosting summary
AI-Flow says only that flow data is saved in the cloud. No hosting country, region or provider is declared anywhere on the site, and no security certification such as SOC 2 or ISO 27001 is published. The privacy policy of 26 January 2024 states that information is processed at the company's operating offices and in any other place where the parties involved in the processing are located, so data may be transferred outside your own jurisdiction; your consent to the policy is treated as agreement to that transfer, and the company undertakes to keep adequate controls in place. France is named as the governing country and the terms apply French law, but neither statement says where servers sit. A purely technical observation, not a vendor declaration: the domain resolves to an Amazon CloudFront edge node geolocated in Zurich, which tells you about content delivery rather than about where your workflows and generated media are stored. Anyone with a data residency requirement should ask the vendor directly, or self-host the open-source build.
Things to keep in mind
Risks and trade-offs to weigh before adopting AI-Flow.
- Generated media is deleted after 7 days on the free tier and 30 days above it: export anything you need to keep before it expires
- Pay-as-you-go credits expire one year after purchase if no new credits are bought, and all purchases are non-refundable
- With your own key connected, model spend lands on your provider account, so an unattended loop or batch can run up a bill outside AI-Flow's view
- No DPA, no subprocessor list and no declared hosting country: run your own diligence before putting personal data through a workflow
- The terms exclude use under HIPAA, FISMA and GLBA, and a single email address handles support, legal and GDPR requests
- Chaining models makes it easy to publish at volume without ever reviewing an output; keep a human check on anything customer-facing
- A pipeline that produces acceptable results can quietly deskill the judgement it replaced, from prompt writing to image selection
Setup & Integrations
Technical difficulty
Low for the hosted version. You sign up without a credit card and the vendor's own guide puts a first result five minutes away; building a flow is drag-and-drop and needs no programming. Adding your own provider key is a one-off paste into Settings, then Secure Store, then Validate. Colleagues can run a published App from a link with no setup at all. Difficulty rises at the edges: calling a workflow through the API Builder assumes you can consume a REST endpoint or a webhook, and self-hosting means Docker Compose or a local install with npm and poetry.
Deployment
Integrations
Supported languages
Behind AI-Flow
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement AI-Flow.
Frequently asked questions
Do I need a subscription to use AI-Flow?
Can I use my own API keys?
Is there a free tier?
How much does the first paid pack cost?
Where is my data stored, and for how long?
Are my files used to train AI models?
Can I run AI-Flow on my own machine?
Can a workflow be called from my own application?
Which third-party services does it integrate with?
Is there an age limit or a mobile app?
Should you pick AI-Flow?
AI-Flow answers a narrow but real complaint about hosted AI tools: the markup and the closed catalogue. Its canvas puts ten providers side by side, lets a credential pasted once into an encrypted store carry the cost of every compatible node, and shows what a node will spend before you spend it. The product is at its strongest when a workflow stops being an experiment: the same graph becomes a form-based App for a colleague, a subflow inside a bigger automation, or a REST endpoint in your own backend, without being rebuilt. That the core is open source under the MIT licence and runs under Docker or as a Windows executable matters as much as any feature, because it makes the tool easy to leave.
The reservations are about the company rather than the software. No postal address, no full legal entity name and a single email address covering support, legal and privacy requests make the vendor hard to assess. There is no data processing agreement, no subprocessor list, no security certification and no declared hosting country, and the privacy policy and terms date from January 2024 and October 2023, well behind the product they describe. Pricing is oddly hidden for a tool that makes cost transparency its argument: no pricing page exists on the marketing site, and the packs surface only inside the application. Generated media is deleted after seven days on the free tier, credits expire after a year, and purchases are non-refundable.
For creators, small content and video teams, e-commerce merchants and developers who already hold provider accounts, AI-Flow is a serious and unusually portable option. For a regulated buyer with procurement and data residency requirements, the paperwork is not there yet.
- Choosing a selection results in a full page refresh.
- Opens in a new window.