# InvokeAI — card for AI agents

Free open-source, self-hosted AI image generation app with a layer-based canvas and node-based workflows.

Category: AI image generators · maker InvokeAI Contributors · checked 2026-11-10

## Price
- Price: Free
- Free tier: Free and open source, self-hosted under Apache 2.0

## License
Apache License 2.0 — https://github.com/invoke-ai/InvokeAI/blob/HEAD/LICENSE

- commercial: **yes** (Commercial use OK)
  > Free to use under a commercially-friendly license
  (https://github.com/invoke-ai/InvokeAI)
- attribution: **no** (No credit needed)
  > You must give any other recipients of the Work or Derivative Works a copy of this License
  (https://github.com/invoke-ai/InvokeAI/blob/HEAD/LICENSE)
- in app: **yes** (Can ship inside your app)
  > perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work
  (https://github.com/invoke-ai/InvokeAI/blob/HEAD/LICENSE)

## Rules to follow
- Keep a copy of the Apache 2.0 license and any NOTICE file when you distribute the software or derivatives.
- Mark modified files with a prominent notice that you changed them.
- Do not use the Invoke name or trademarks beyond describing the origin of the work.
- Check the separate license of each model (Flux, SDXL, etc.) before commercial use; this license covers only the software.

## How to connect
- Download only
- Repository: https://github.com/invoke-ai/InvokeAI
- Website: https://invoke.ai

## Notes
Added automatically. License read from the official pages.

## Numbers (AgentGid data)
- Gid Score: 45
- GitHub stars: 29,121

## Lines to paste into your prompt or AGENTS.md

```
- InvokeAI (AI image generators). Free open-source, self-hosted AI image generation app with a layer-based canvas and node-based workflows. Price: free; commercial use OK; no credit needed.
  Connect: download from https://invoke.ai (no API).
  Rule: Keep a copy of the Apache 2.0 license and any NOTICE file when you distribute the software or derivatives.
  Rule: Mark modified files with a prominent notice that you changed them.
  Rule: Do not use the Invoke name or trademarks beyond describing the origin of the work.
  Rule: Check the separate license of each model (Flux, SDXL, etc.) before commercial use; this license covers only the software.
  License: https://github.com/invoke-ai/InvokeAI/blob/HEAD/LICENSE (checked 2026-11-10; details: https://agentgid.com/assets/invokeai/)
```

## Links
- AgentGid page: https://agentgid.com/assets/invokeai/
- Is it free for commercial use? https://agentgid.com/assets/invokeai/commercial-use/
- Source: https://invoke-ai.github.io/InvokeAI
- Source: https://github.com/invoke-ai/InvokeAI/blob/HEAD/LICENSE
- Source: https://github.com/invoke-ai/InvokeAI


## InvokeAI: free image-making app on your own computer

InvokeAI is a free app you install on your own computer to make and edit pictures with AI image models. It opens in your web browser with a canvas for painting, layers and fixes, plus visual workflows you can save and reuse. The app itself is free for paid work; each image model you load has its own license.

### Where to use it
- Fix or extend a picture: Paint over a part of an image to redraw it, or widen the frame so the scene continues. It works on AI images, sketches, photos and renders.
- One look for a whole image set: Build a step-by-step workflow once, such as model, style, size and upscaling, then reuse it for every new batch of product or blog images.
- Private images that stay at home: Prompts and pictures stay on your machine when you use local models, which suits client drafts or unreleased products.
- Images from your own script: Your AI assistant or a small script sends a job to the running app and picks up the finished image, for example to fill a gallery overnight.
- A shared studio for a small team: Run one copy on a strong computer and give each teammate their own sign-in and boards, with an admin who manages models.
- Hosted models without a big graphics card: Pick GPT Image or Gemini inside the same app when your computer is too weak for local models. Those images are made by the provider and billed to your account there.

### How to start
- Check that your computer can run the models you want (graphics card memory and RAM per model family).
- Download the Invoke Launcher for Windows, macOS or Linux and click Install; it installs and updates Invoke for you.
- On a Linux server with an Nvidia card you can start it with Docker instead, then open port 9090 in a browser.

```
docker run --runtime=nvidia --gpus=all --publish 9090:9090 ghcr.io/invoke-ai/invokeai
```


### License
The app is free under Apache 2.0, also for paid work; keep the license and notices if you pass it on. Each image model has its own license, and some limit commercial use.

### Good to know
- The app license covers only the app. Check each model's own terms before selling or publishing its images.
- You need a capable computer. SDXL at 1024 by 1024 wants an Nvidia 20xx card or newer with 8 GB of video memory and 16 GB of RAM.
- Macs need Apple Silicon (M-series); Intel Macs are not supported, and 16 GB of memory or more is recommended.
- Hosted models such as GPT Image or Gemini cost money on the provider's account and need that provider's key.
- Docker on a Mac cannot use the graphics chip, so images come out slowly; the docs point Mac users to the launcher.

Checked 2026-11-10 against: https://invoke-ai.github.io/InvokeAI/, https://invoke.ai/, https://invoke.ai/sitemap-0.xml, https://invoke.ai/start-here/system-requirements/, https://invoke.ai/start-here/installation/, https://invoke.ai/start-here/manual/, https://invoke.ai/configuration/docker/, https://invoke.ai/configuration/low-vram-mode/, https://invoke.ai/concepts/models/, https://invoke.ai/features/external-models/, https://invoke.ai/features/multi-user-mode/api-guide/, https://invoke.ai/development/guides/workflow-api/, https://invoke.ai/troubleshooting/faq/, https://invoke.ai/download/, https://github.com/invoke-ai/InvokeAI/blob/main/README.md, https://github.com/invoke-ai/InvokeAI/blob/main/LICENSE, https://pypi.org/pypi/invokeai/json

### Mistakes to avoid
- Don't: Fetching a saved workflow from /api/v1/workflows/ and expecting it to run → Do: Build an executable graph and POST it to /api/v1/queue/default/enqueue_batch, then poll the queue item (The queue accepts graphs, not saved workflow records.)
- Don't: Installing into the system Python 3.13 → Do: uv venv --relocatable --prompt invoke --python 3.12 --python-preference only-managed .venv (Invoke supports Python 3.11 and 3.12 only.)
- Don't: Committing api_keys.yaml to git or sharing the Invoke folder with it inside → Do: Keep api_keys.yaml out of version control; the person pastes the provider keys in themselves (The file holds the provider secrets that bill their accounts.)
- Don't: Calling the API of a multi-user install without a token → Do: Log in with POST /api/v1/auth/login and send Authorization Bearer <token> on each request (In multi-user mode most endpoints require a JWT bearer token.)
- Don't: Running the Docker image on a Mac for real work → Do: Use the Invoke Launcher on Apple Silicon Macs (Docker cannot reach the Mac's graphics chip.)

### Questions
- Is InvokeAI free for commercial work? The app is under Apache 2.0, which allows commercial use; the models you run in it have their own licenses.
- Do my prompts and images leave my computer? Not with local models; Invoke runs on your own hardware. Hosted external models send the request to that provider.
- Which Python version does a manual install need? Python 3.11 or 3.12; the launcher installs Python for you.
- Can a script run a saved workflow by its name? Not directly. The queue takes a graph, so a client reads the workflow and sends a graph to enqueue_batch.
- Which models does it support? The README lists SD 1.5, SDXL, SD 3.5, several FLUX versions, Z-Image and Qwen Image, among others, plus a few hosted models.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
