locala

Important

This repository is archived. Locala has moved into web-ai-sdk as the local-first agent playground. Development continues there; see web-ai-sdk#145 for the migration.

Local chats in your browser. On-device chat using the Web Built-in AI APIs, with zero round-trips to a backend.

• All inference runs on-device via navigator.LanguageModel (Chrome / Edge Built-in AI).
• A small set of personas (Concise, Explorer, Coder) you can pick per chat.
• Chats are persisted to localStorage. Nothing leaves the browser.
• External agents (Chrome agent, Cursor, Claude, etc.) can drive the app via WebMCP tools (list_chats, new_chat, send_message, set_mode, ...).

Built on web-ai-sdk

locala is a consumer of @web-ai-sdk, a set of headless wrappers for the browser's Built-in AI APIs:

• @web-ai-sdk/prompt powers the chat (system prompts, streaming, abort, warm-session reuse).
• @web-ai-sdk/webmcp exposes the in-app tools to external agents.

See web-ai-sdk.dev for the docs.

Browser requirements

The Web Built-in AI APIs are still behind flags. The compatibility matrix moves; treat this as a starting point and verify against web-ai-sdk.dev.

• Prompt API: Chrome 138+ or Edge 138+ with the Prompt API flag enabled.
• WebMCP (optional, only needed for external agent drive-through): Chrome 146+ / Edge 146+ with the WebMCP testing flag.

If the Prompt API isn't available, the UI shows a banner with the flag URL.

Develop

Requirements: Node 22+ (pnpm 11 needs it), pnpm.

pnpm install
pnpm dev

Open http://localhost:5173. Type-check and production build:

pnpm typecheck
pnpm build

Deploy

The app deploys to GitHub Pages automatically when main is updated. The workflow lives at .github/workflows/deploy.yml.

License

MIT © Beto Muniz