Keel

What is Keel?

Keel is a local-first desktop app for macOS and Windows that provides an AI assistant whose memory is stored in your own files. The goal is to keep your context in a plain markdown workspace on disk, while letting you choose which LLM provider you want to use for each conversation.

Keel indexes your workspace, assembles relevant context for chats, and writes back captures, decisions, and tasks into markdown. The assistant layer can be pointed at Claude, GPT, OpenRouter, or local models via Ollama, without moving your notes into a remote database.

Key Features

• Local-first markdown workspace: Keel uses a folder of plain markdown (e.g., ~/Keel) as your long-term “memory” so you can edit and back it up with any editor and your existing backup practices.
• Context engine for conversation grounding: It indexes your workspace, selects relevant material for each conversation, and generates structured context for the model to use.
• Capture and write-back to disk: When a chat produces something worth keeping (decisions, facts, tasks), Keel saves it back into your markdown workspace.
• Wiki bases from project folders: You can turn a project directory into a queryable knowledge base by ingesting markdown and PDFs, then refreshing compiled context as files change (via commands such as /create-kb and /refresh-kb).
• Task and dashboard support: Keel surfaces open tasks and reminders, and provides a launch dashboard including the morning brief, reminders, and recent activity.
• Meeting transcription to structured notes: Keel can transcribe locally with Whisper (for recorded/imported audio) and writes decisions, action items, and attendees back into the relevant project.
• Model switching and fallback: In settings you can choose providers such as Claude, GPT, or OpenRouter, and use local models via Ollama; it can fall back automatically when one is down.
• Scheduled jobs for recurring workflows: You can run prompts or workflows on a recurring schedule (e.g., daily digests or weekly reviews) and capture the output back into markdown.
• Desktop notifications for reminders: Tasks and reminders support due dates and time-based reminders delivered as desktop notifications.
• Integrations for importing/exporting: Keel can sync Calendar events into context, read and export Google Docs, sync X bookmarks into a wiki, and publish posts from chat.

How to Use Keel

1. Download Keel for macOS or Windows and install it on your machine.
2. Create or point Keel to a markdown workspace folder on your disk (the default is typically ~/Keel). Keep your project notes, daily logs, wiki material, and task files here.
3. Choose your LLM provider in settings (Claude, GPT, OpenRouter, or a local model via Ollama). If a provider is unavailable, Keel can fall back automatically based on your configuration.
4. Start interacting with Keel in chat. As you work, use capture workflows so decisions, facts, and tasks get written back to markdown.
5. Use the structured features for your routine: generate a morning brief from your workspace, refresh or create wiki bases for a project folder, and record/import meeting audio when you need structured notes.

Use Cases

• Daily logging and summaries: Generate a morning brief from your existing notes, review open tasks and last-day loose ends, then write a structured end-of-day summary back into your daily log.
• Building project knowledge bases: For a specific repository or document folder, create a per-project wiki by ingesting markdown and PDFs, then refresh compiled context as the folder updates.
• Turning meetings into action-oriented notes: Record or import meeting audio; Keel transcribes locally with Whisper and writes meeting outputs such as decisions, action items, and attendees into the relevant project notes.
• Personal knowledge management with queryable context: Maintain a “brain” as plain markdown on your disk and rely on Keel’s indexing so that chats can reference the right parts of your notes.
• Recurring check-ins and digests: Schedule prompts/workflows for periodic outputs (like weekly reviews or daily digests) and capture the results back to markdown.

FAQ

• Is Keel account-based or does it require a server? No. Keel is local-first with no telemetry and no account, and it does not run a server; your workspace is kept on your computer in files you own.

• Where is my context stored? Keel stores your context in a markdown workspace folder on your disk (e.g., ~/Keel). Keel writes captures, decisions, and tasks back into markdown in that folder.

• Can I switch between different AI providers? Yes. Keel supports Claude, GPT, OpenRouter, and local models via Ollama, and you can swap providers in settings; it can also fall back automatically if one provider is down.

• What kinds of input can Keel ingest for wiki bases? The page states that Keel ingests markdown and PDFs for project knowledge bases.

• Does Keel handle audio transcription? Yes. Keel can record or import audio and transcribe it locally with Whisper, then write structured meeting summaries back into your project.

Alternatives

• Local knowledge base + chat tools (markdown-first workflows): Tools that let you manage notes locally and use LLMs for Q&A can serve a similar purpose, though they may not provide the same built-in indexing plus automatic write-back into markdown.
• Desktop “AI assistant” apps with cloud memory: Some assistants store conversation history in a remote system and may require accounts; compared with Keel, they typically trade local ownership of context for convenience.
• LLM dashboards for provider switching: Applications that manage multiple LLM endpoints can be used to switch models, but they may not include Keel’s specific markdown workspace engine, wiki base generation, or structured meeting transcription workflow.
• Note-taking apps with search/RAG integrations: General note apps with AI search can help you surface relevant text, but they usually differ in how they structure outputs (e.g., task reminders, meeting action items) and whether they write structured captures back to markdown automatically.