NodeTool Cloud is in alpha — run workflows in the browser, on your own keys.NodeTool Cloud is in alpha. Try it →
Create and edit images, video, audio, and text with agents that work alongside you. Let them build and revise workflows, then inspect and edit the results yourself. Your project keeps the brief, assets, and edits together.
Free and open source, AGPL-3.0. macOS, Windows, and Linux.
Route your shots through the best video, image, audio, and language models on the market, or run open weights locally on your own hardware. Switch between them with one click, and pay each provider directly at their published price.
Four jobs that run every week, each a real run against live models: what you end up holding, who it is for, the models the chain calls, and every workflow in it already installed with Studio. Opening a chain and reading every graph in it needs no key and no account. Running it bills the providers you chose, at their list prices.
A vertical product ad, plus the hook lines and thumbnails to test it against.
The picture above is this chain run for real, with 1 of 3 models reached another way — the recipe page names which, and why.
3 provider keys to run the whole chain. None to read it.
One presenter clip, spoken in a second language, with a lip-synced cut, subtitles, and a back-translation to check it.
The picture above is this chain run for real, with 2 of 5 models reached another way — the recipe page names which, and why.
3 provider keys to run the whole chain. None to read it.
One packshot per SKU becomes the whole channel set: cutout, studio scene, seasonal relight, turntable clip, print master, and the listing copy.
The picture above is this chain run for real, with 4 of 6 models reached another way — the recipe page names which, and why.
2 provider keys to run the whole chain. None to read it.
A logline becomes a beat sheet, a numbered shot list, a cut teaser, and a score under it.
The picture above is this chain run for real, with 2 of 4 models reached another way — the recipe page names which, and why.
4 provider keys to run the whole chain. None to read it.
Every recipe is built from the workflows that ship with Studio. All recipes · Single-surface demos
Stop copy-pasting from chat windows. Describe what you need, and NodeTool generates the full project in seconds.
How agents and workflows work together →
Say what you want made and pick its shape — a 30-second spot, a trailer, a mini app — and the agent plans the documents that shape needs. The estimate under the prompt is read off what your own past projects of that shape cost, so you know the order of magnitude before anything runs.
A board created, six keyframes rendered, a contact sheet built — each call is a card you can expand to see the code it ran and what came back. No spinner standing in for the work, and every shot keeps the prompt and the model that made it, so you can re-roll one without touching the other five.
The same six shots, now a board with its stills, a script with the one line still to re-voice, and a cut with its tracks — each card opens the editor that made it. The conversation stays on the left, and the bar along the bottom splits the spend into stills, clips, voice and pipeline. Your keys, your invoice, at provider rates.
Every project keeps what it rendered, what it cost, and where it stands. Open one and its board, script and cut come back as a tab group; drag a loose document onto a card to file it there. Change a shot and re-run — nothing has to be rebuilt from the transcript.
Tell the agent your concept, style, and tone. It drafts the script, casts the voices, and sets the visual direction. Prefer to wire it yourself? It's the same canvas.
The agent storyboards scenes, generates the footage, syncs the audio, and cuts it all together on a multi-track timeline — running on your own accounts at provider prices.
Jump in at any moment. Swap a voice take, re-roll a clip, or trim frames. The agent builds the foundation; you refine the final cut.
Storyboard, script, timeline, sketch, and 3D scene, all on the canvas you generate on. A storyboard becomes a script, the script becomes takes, the takes land on a multi-track timeline. The agent works every one of them through the same tools you click.
Board the film shot by shot. Generate cheap stills first to lock the look, then pay to animate only the shots you approved.
Draft the dialogue, cast a voice per character, and audition alternate line readings. Change the words and the take flags itself stale, so you see exactly what still needs voicing.
Arrange, trim, and layer generated video and audio across multiple tracks, down to the frame and the stem. The agent edits the same document when you ask it to tighten the opening.
Sketch, paint with brushes, and blend hand-drawn elements with AI-generated layers. Bind a layer to a prompt and regenerate that layer alone.
Place primitives and lights in a glTF scene by hand or by tool call. The same operations run headlessly, so a scene is reproducible.
One job, one tool, one screen. Every app here is a shipped workflow with a form on top, so an operator runs it without opening the canvas.
Cut it to the frame the channel wants.
Get the script back out of the footage.
Translate and voice in one pass.
Ship the version people read.
Isolate the product once.
Place it in a described setting.
Change the season without a reshoot.
Give the product page motion.
Upscale only the keeper.
No locked project formats, no markups, no subscription tiers. Bring your own keys and switch providers with one click. You own the project file, the workflow, and the final cut.
Connect directly to OpenAI, Anthropic, Gemini, FAL, KIE, Replicate, and the specialized video models. Your keys stay on your disk in Studio, encrypted in Cloud.
No credit packs, no subscription traps. If an image costs $0.03 at the provider, you pay $0.03 to the provider. NodeTool takes no cut.
Studio and Cloud are built from the same AGPL-3.0 source, with no paywalled features. Read it, fork it, or host it yourself at any time.
The board, the script with its takes, and the multi-track cut are ordinary files on your disk. Export a .nodetool bundle and open it on another machine, or in a fork.
What that comes to in a month, model by model, is on the pricing page.
Chain models, editors, and code into a node graph you can re-run on new inputs, ship as a mini-app, or drive headless from the CLI. Each part has its own page.
A hosted platform will make the film, and it keeps the model list, the billing, and the project. Here is the same table with the studio open.
| NodeTool | Hosted platforms | |
|---|---|---|
| Models | Every major provider, switched in one click | The list they picked |
| When a better model ships | Add it the day it ships | Wait for them to add it |
| What you pay | Provider list prices, on your own keys | Their credits |
| What you keep | The board, the takes, and the multi-track cut as an editable project | An exported video. The project stays in their app. |
| Source | Open, AGPL-3.0 | Closed |
| Where it runs | Desktop app and browser, self-host any time | Their servers only |
Comparing against a specific product? See NodeTool and ComfyUI and NodeTool and Figma Weave.
Take Seedance, one of today's best video models. It is sold by FAL, Replicate, and KIE at different prices, and NodeTool lets you pick the cheapest of the three. When the next Veo or Kling arrives, you switch to it in one click.
That is what holding the keys buys you: the best model at the best price each week, and nothing to lose if your favorite tool gets bought, repriced, or shut down.
Same workflows, same building blocks, same providers. Pick the one that suits how you work, and switch whenever you like. Both are AGPL-3.0 open source: Cloud is simply our hosting of the same code you can run yourself.
Local-first desktop app
Best for: artists with a good graphics card or an Apple Silicon Mac, big local model collections, offline work, and anyone who wants everything kept on their own disk.
Hosted in the browser · Alpha preview
Best for: studios and solo artists who want to start in seconds, work from any device, and skip the hardware setup, while still using their own keys with every provider.
One recommendation, not two: use Studio for work you are being paid for — it is the finished edition. Cloud is worth a look if you want to see the workspace without installing anything, and it is in alpha until further notice. Your workflows move freely between the two.
Whether you use the desktop app or the browser version, you get the exact same features, tools, and capabilities: the same agent, the same workflows, the same models.
Runs directly on your computer. Keeps your files and settings private on your hard drive, with optional offline support.
Go to StudioRuns in your web browser with zero installation needed. The fastest way to jump in and start creating. Currently in alpha.
Go to CloudConnect custom scripts, automate tasks, and integrate your own code into the visual canvas, through the SDK, the CLI, or an MCP server.
Go to DevelopersTurn a brief into a production workflow you run again for every product, market, and variant.
Go to MarketingNodeTool is an open-source creative AI workspace. Create and edit images, video, audio, and text with agents that work alongside you. Let them build and revise workflows, then inspect and edit the results yourself. Your project keeps the brief, assets, and edits together. Use supported local models or connect cloud providers with your own keys.
Read more →Yes. The whole project is AGPL-3.0 on GitHub. Studio and Cloud are built from the same source, and nothing is held back for a paid tier. You can host it yourself at any time.
Read more →NodeTool Studio, the desktop app, is free and open source. NodeTool Cloud is free while it is in alpha. At full release a subscription covers hosting, at a price announced before anyone is charged. In both editions you bring your own API keys and pay each provider their list price directly, so what a run costs is whatever the provider charges for it. Local models through Ollama, MLX, or llama.cpp cost nothing per call.
Read more →NodeTool Studio runs on your own machine: the workspace, your workflows, your files, and your API keys stay there, and open-weight models can run locally through MLX, Ollama, llama.cpp, vLLM, and LM Studio. When you call a hosted model, that request goes from your machine straight to the provider you chose, on your key — it does not pass through NodeTool servers. Cloud is the same open-source app hosted in the browser, and your keys still call the providers directly.
Read more →The leading models, including Flux, Seedance, Wan, Veo, Kling, Hailuo, Qwen Image, Whisper, ElevenLabs, and Suno, reached through providers such as FAL, KIE, OpenAI, Anthropic, Gemini, Replicate, Together, Groq, Mistral, OpenRouter, and HuggingFace. Models can also run on your own computer through MLX, Ollama, llama.cpp, vLLM, and LM Studio.
Read more →Every editor in NodeTool — the node canvas, sketch pad, storyboard, video timeline, script editor, 3D scene, and app builder — is exposed to agents as tools, around 120 in all. An agent doesn't describe what you could do; it builds the workflow, runs it, and repairs what fails, on the same surfaces you use. The same tools are exposed over MCP, so outside agents such as Claude Desktop and Claude Code can drive NodeTool too.
Read more →No. NodeTool works entirely with cloud providers — bring your own API key, and the model runs on the provider's hardware with nothing to install locally. A GPU only matters if you choose to run open-weight models yourself through Ollama, MLX, or llama.cpp: it speeds those up, but many smaller local models run fine on CPU too, just slower.
Read more →NodeTool is open source under AGPL-3.0. Star the project on GitHub, join us on Discord, and share workflows with the artists, motion designers, and studios already using it on real jobs.
Questions, bug reports, feature requests — we answer.
Download the open-source studio for macOS, Windows, and Linux. Free, AGPL-3.0, and built alongside working filmmakers, designers, and video editors. Or try Cloud in your browser, in alpha, with nothing to install.