Overview

LocalChat is an on-device AI workspace aimed at people whose work legally cannot go to the cloud — lawyers, therapists, doctors, accountants, and consultants. Rather than pasting snippets, you hand it the entire file: a 200-page deposition, a full patient chart, a tax packet, a contract. It reads PDFs, markdown, CSV, XLSX, PNG, and TXT locally and answers with citations back to the source. Each project is its own workspace with isolated files, conversations, and embedded memory, and a toggle constrains the model to answer only from that project's documents. Point it at an Obsidian vault and it scans notes, follows links, and turns them into searchable memory. Models come from a curated catalog of 25+ options tested on Apple Silicon — reasoning, vision, OCR, and embedding models — plus any GGUF model from Hugging Face via URL or drag-and-drop, and you can swap models mid-thread without losing the conversation. There is no account to create, no usage tracking, and after setup no internet connection is required.

Pricing: Single: $49.50 launch price (reg. $99) | Family: $199.50 for 5 devices (reg. $399) | Team: custom — one-time purchase including 1 year of updates, 30-day money-back guarantee

Minimum macOS: 11.0 (Big Sur)

Architecture: Apple Silicon

Key Features

Reads whole files on-device — PDF, markdown, CSV, XLSX, PNG, and TXT
Answers cite the source, down to page level
Project workspaces with isolated files, conversations, and embedded memory
'Answer only from this project's files' toggle to keep the model inside the matter
Obsidian vault integration — scans notes, follows links, builds searchable memory
Curated catalog of 25+ models tested on Apple Silicon: reasoning, vision, OCR, and embeddings
Any GGUF model from Hugging Face via pasted URL or dropped file
Swap models mid-thread without losing the conversation
No account, no login, no usage tracking
Fully offline after setup — internet only needed for install and model downloads
One-time license rather than an AI subscription