Keel is a local‑first AI assistant for Mac and Windows that stores all data as plain markdown files on the user’s disk, ensuring full control over memory and context. Users can switch between models such as Claude, OpenAI’s GPT, or locally‑run Ollama models without losing their history, and the open‑source MIT‑licensed app includes no telemetry or account requirements.
Funding
Funding not disclosed
Founders
Product
Problem
Users of AI assistants often rely on cloud‑based services that store conversation history on remote servers, creating privacy concerns and limiting data portability. Switching between different language models can also cause loss of context, forcing users to restart interactions or migrate data manually.
Solution
Keel provides a local‑first AI assistant for macOS and Windows that records every interaction as plain markdown files stored directly on the user's device. By keeping the conversation history locally, Keel ensures that data remains private, portable, and under the user's control. The app supports seamless switching among Claude, OpenAI GPT, Ollama, and any OpenRouter‑compatible model without losing context, allowing users to choose the best model for each task. Keel’s open‑source MIT license guarantees transparency and eliminates telemetry or account requirements, while built‑in features for capture, organization, scheduling, and action execution enable a complete productivity workflow on the desktop.
Target Audience
Keel is aimed at privacy‑conscious professionals, developers, and knowledge workers who need a reliable AI assistant on their desktop without relying on cloud storage or vendor lock‑in.
Features
- Stores all chats as plain markdown files on the local filesystem for easy backup and portability
- Model‑agnostic architecture lets users swap between Claude, GPT, Ollama, or other compatible models without losing context
- Local‑first design with no telemetry, no accounts, and full data ownership
- Integrated tools for capturing notes, organizing information, scheduling tasks, and executing actions within the same app
- Cross‑platform support for both macOS (Apple Silicon and Intel) and Windows x64
- Open‑source MIT‑licensed codebase, allowing community contributions and transparent security auditing