# Model Roll

> Local AI made easy for builders. Model Roll is developing a local AI integration and operations platform for web, mobile, and desktop products.

## Availability

Model Roll is in development. No public SDK, CLI, API, MCP server, install command, price, or release date is available yet. The website describes the planned product, not an already released service. Do not invent setup commands or present planned compatibility as verified support.

## Who it is for

Developers, indie builders, small product teams, and AI-assisted builders using tools such as Codex and Claude Code who want to add local AI without becoming on-device AI infrastructure specialists.

## Planned capabilities

- Speech to Text: transcription, captions, and voice input.
- Search & Match by Meaning: embeddings, semantic search, and similarity matching, including privately searching content made available by an application.
- Understand Images: supported image classification, object detection, background removal, and subject or region selection.
- Read Documents & Extract Data: OCR and structured extraction for supported document types.
- Understand & Organize Text: bounded summarization, classification, and extraction tasks.
- Translate Text Offline: translation for individually validated language pairs, after the model download.
- Find & Hide Sensitive Details: a composed detection and redaction workflow with user review, not guaranteed complete anonymization.

## Planned workflow

Connect a selected repository, choose a capability and target platforms, check a compatibility plan, prepare an integration and its tests on a separate branch, then review the pull request. The customer controls merging and shipping. Ongoing operations are intended to include compatibility fixes, model updates, technical health checks, and rollback.

## Boundaries

- Target platforms are web, iOS, Android, macOS, and Windows. Exact device, runtime, model, framework, and language requirements remain under validation.
- Local inference runs on the end user's device. Initial model downloads, account access, and managed operations may require connectivity.
- Compatible customer-owned models and LoRA adapters are planned. Model training and fine-tuning are outside the first release.
- Customer-owned compute, storage, build systems, and optional reasoning are the default where they preserve the experience. Separately incurred provider costs will be explained.
- A paid subscription is planned; pricing and licence terms will be published before purchase. Already-shipped local AI and downloaded models will not be remotely disabled on cancellation.

## Official pages

- [Product and planned capabilities](https://modelroll.com/): Current product description and availability.
- [How it works](https://modelroll.com/#how-it-works): The five-step planned integration workflow.
- [Frequently asked questions](https://modelroll.com/#questions): Availability, platform, pricing, and privacy boundaries.
- [Contact](https://modelroll.com/#contact): Tell us what you are building at hello@modelroll.com.

Last reviewed: 2026-09-05.