The library of everything your AI tools make.

Archibal is the library of everything your AI tools make, with how each file was made and whether you're allowed to use it. It does three things. It keeps a library of every generation, searchable by prompt, model and project. It fills that library automatically: connect your tools before you generate and each file arrives with its model and prompt. And it shows the usage terms of the model behind each file - commercial use, training on your work, who owns the output - plus the rules for the markets a project will run in. When the job is done, your client gets the final files with one link showing how each was made.

Connect your AI tools first, before you generate: Claude (which drives Higgsfield, Kling, Runway, ElevenLabs, Magnific and more), ComfyUI with the Archibal node, or the Premiere Pro plugin. It is a one-time setup. Then generate as usual - each generation lands in your project with its model and prompt, with nothing to log by hand. Claude logs because the connector asks it to, so glance at the capture feed now and then.

It depends on the source. Through Claude: the tool, model, prompt and job id of each job Claude logs, plus the file when Claude passes its link. Through the ComfyUI node: the full workflow, seed and models of each render. The Premiere Pro plugin sends an export with the clips in its sequence and links the ones already in your library. An uploaded file is read only for what it still carries inside it, and most exports carry nothing - so capture happens while you generate, not after.

Yes, as the deliverable: add it on top of the clips you captured while generating, and the record shows how those clips were made. On its own, a finished edit shows no AI, because editing software writes a new file and drops the history of the clips inside it. That does not mean no AI was used, only that the file cannot say. Archibal never guesses from pixels; it records each clip as it is made.

The final files and one link showing how each was made and with which models, a verify link anyone can open later, and a PDF to file. The record is fixed at delivery; later changes become a new delivery. By default the client view leaves out your prompts and workflow - they appear only if you choose the full archive view. Delivered files also carry signed C2PA Content Credentials, a label inside the file saying how it was made; for now Archibal signs with a test certificate while its production certificate is pending.

Each file in your library shows the usage terms of the model behind it: whether commercial use is allowed, whether the provider trains on your prompts or outputs, and who owns the output. The terms are taken from the providers' own documents, per model and per plan or access route, with a link to the clause where there is one. The model terms index is public, so you can search it without an account.

Rules such as EU AI Act Article 50 put the marking duty on AI model providers and the audience-disclosure duty on whoever publishes the content. An Archibal record does not discharge either duty. It documents how a deliverable was made, which is what brands and platforms increasingly ask their studios for, and each project lists the rules for the markets it will run in. The details live at our compliance hub.