OKF Todo is a free, open-source desktop To-Do application for developers and support staff. It keeps tasks, history, images, and attachments together in a local SQLite database. You can manage your work through the desktop interface and use your own AI assistant to turn unstructured information into tasks and plans.
The project is designed for work that often falls between formal systems: production errors, support cases, deployment checks, investigations, ideas, notes, and follow-ups.
A local place to keep the work
A support request rarely arrives as a complete task. It may start with an email thread, a screenshot of an error, a few meeting notes, or an Excel list. You still need to decide what to investigate, what information is missing, and what to do next.
OKF Todo gives that work a place to live. Capture a task, add the source material, and keep the investigation and follow-up together as the work develops.
The desktop application includes:
- Task lists and views for active, ready, urgent, waiting, overdue, and completed work.
- Rich-text and Markdown editing, with images in task bodies.
- Priorities, deadlines, tags, source references, and waiting targets.
- Checklists, file attachments, and relationships between tasks.
- A Timeline combining comments and automatic change history.
- Light and dark themes, database backups, and offline Help.
The application runs on your computer without a hosted account or subscription. You can create, organize, update, and complete tasks entirely through the interface.
Let your own AI assistant work with the tasks
The AI integration starts with the material you already have. Give a compatible assistant an email thread, diagnostic output, a screenshot, meeting notes, or a spreadsheet, and ask it to prepare an investigation plan or propose tasks.
For example:
Read this support thread and propose an investigation task. Include the known facts, open questions, next steps, and information we are waiting for. Do not save anything yet. Show me the proposal for review.
After reviewing the proposal, explicitly approve what should be saved. Then ask the assistant to read the saved result back and check it against your approval. Continue managing the work in OKF Todo.
The assistant interprets the source material; OKF Todo stores and organizes the tasks. Supported input formats depend on the assistant and its tools. The application has no built-in AI model and does not automatically collect material from your inbox.
OKF provides context; MCP provides tools
The Open Knowledge Format (OKF) layer describes the task concepts, database structure, relationships, and rules. It helps an assistant understand what the data means and how to work with it. Context alone does not grant access to the database.
The included MCP server gives a compatible client tools to discover, read, create, and update tasks. It also supports comments, checklists, relationships, attachments, task lists, and lifecycle actions. CLI commands provide another route through the same application services.
Direct SQLite access is also possible when you give a tool access to the database. Those writes bypass the application services, so the caller must preserve data integrity and any required history. The desktop interface, MCP server, and CLI apply the application's validation and lifecycle rules.
For setup, open Help → OKF layer → Copy prompt in the running app. For an MCP client, use Help → MCP server → Copy configuration. The app supplies the paths and configuration for your installation.
Local storage and an external assistant
Tasks are stored locally, and the application has no cloud synchronization or multi-user collaboration. Your external assistant may send supplied material or tool results to its model provider. Review that assistant's data handling before giving it sensitive information.
OKF Todo is free. Your chosen assistant may require a separate account or payment. The recommended process is draft, review, save, verify; follow it explicitly rather than assuming every tool enforces an approval step.
Try the project
Windows users can install from Microsoft Store or use the installer linked in the GitHub README. Windows is the packaged and release-tested platform. macOS and Linux can run from source; packaged installers are still planned.
The project uses .NET, Photino, EF Core, SQLite, and static HTML/CSS/JavaScript. It is built using Codex, but it is not tied to Codex. The source is available under the AGPL-3.0 license.
This 83-second video shows an example of Excel notes becoming reviewed tasks:
The video uses fictional input, replayed planning and approval, and application captures through a browser test host. The demo guide includes the workbook and instructions for trying it.
Explore the project and source on GitHub or see the screenshot tour.
I welcome feedback on the task workflow, assistant setup, and project direction. Would a local, single-user application fit your development and support work, or would optional multi-user functionality make it more useful? Issues and contributions are welcome.
AI disclosure: Codex assisted with drafting and editing this article from the project's README and documentation.
Top comments (0)