Mikan

Mikan is a minimalist daily to-do companion that proactively helps you get things done. Add a to-do task, and Mikan will JUMP (Just Understand, Make Progress) start the work, saving you time and keeping you focused. We all procrastinate, but Mikan is here to inspire you and guide you into the flow state, making you feel motivated and focused to get things done.

How does it work?

Think of Mikan as your loyal productivity companion, much like our real-life furry friends. You can set Mikan up to use your emails, files, documents, websites, integrations, and MCP connections as context to work on your to-do tasks. For example, if you entered "Cook dinner":

1. Mikan will search for relevant content from your personal knowledge base, then grab some popular weeknight recipes from the web.
2. It'll then present to you its findings, such as a list of recipes, with follow-up questions like "What kind of food are you in the mood for?" or "What are some ingredients you want to use?"
3. You can type or speak your responses, or take a picture of some ingredients to expand Mikan's knowledge about the task.
4. It'll then find more recipes with the updated context for you to choose from.
5. Let's say you found something you like. Mikan will try to provide helpful content, like a shopping or ingredients list with instructions to cook dinner.

Mikan works with you to get things done, iterating on tasks in a flow state. Here are a few more ways Mikan makes you more productive:

• "Reply to Jenny" => Let Mikan create a draft using context from the email thread, meeting notes, documents, and more to create a reply in your writing style with relevant information.
• "Meet with Scott & Stephen" => Gather relevant content and information for you to reference along with some talking points to use during the meeting.
• "Clean up the garage" => Find helpful DIY organization ideas and suggestions based on what Mikan knows about you, such as your hobbies (woodworking, snowboarding, etc.).

What is a Mikan?

Mikan is the product mascot and logo. Mikan is an orange colored Husky/Akita mixed dog with playful and friendly energy. We use Mikan, an anthropomorphised dog, to help convey complex concepts, tedious tasks, and features in a friendly and easy-to-understand way. Here are a few examples:

Train Mikan

Just like teaching your dog how to sit or roll over. You can train Mikan to do tricks by adding integrations, files, and tools that it will use.

1. Similar in architecture to the project knowledge base used in Claude Projects, except the focus is on the best user experience, and the knowledge base can be populated by the AI agent using content from files, multimodal inputs, MCP, tools, other AI agents, and integrations. [Claude Projects | Claude Projects Announcement]
2. Mikan can use MCP and other integrations to create and manage context

Feed Mikan

A fun way to frame how a user provides context for Mikan to process and use. The user experience should aim to invoke the joy of throwing your dog (Mikan) a treat to eat, except in this case, you're feeding it multimodal input, files, MCP, and other integrations.

1. Capture context, quickly and without friction
2. Quickly capture/input content like links, documents, images, videos, audio, and text
3. Uses sources like a quick capture UI, MCP, files, folders, screenshots, meetings, integrations, tweets, emails, messages, and calendars

Vibe with Mikan

Effortlessly iterate on the content created by Mikan. Mikan is here to inspire you and guide you into the flow state, making you feel inspired, motivated, and focused to get things done.

1. Automatically associates relevant context with a to-do task
2. Enrich to-do tasks with organized context
3. Users can manually add context to a to-do task

This README is a WIP of raw content. It'll be cleaned up in later commits once the project is out of early concept and ideation.

Project Structure

/
├── apps
│   ├── desktop
│   ├── docs
│   ├── mobile
│   └── web
├── docs
├── packages
│   ├── eslint-config
│   ├── tailwind-config
│   ├── typescript-config
│   └── ui
└── scripts

Tech Stack

• Frontend: Next.js, TypeScript, Tailwind CSS, ShadCN, MagicUI
• Backend: FastAPI, Python, Vercel & CF Functions, Server Actions
• Database: PostgreSQL (with Drizzle ORM)
• AI: TBD
• Deployment: Vercel (Frontend), TBD (Backend)
• Analytics: TBD (PostHog? VA? ??)

Key Features

1. Capture context, quickly and without friction be able to:
   1. Quickly capture/input content like links, documents, images, videos, audio, and text
   2. Uses sources like a quick capture UI, MCP servers, files, folders, screenshots, meetings, integrations, tweets, emails, messages, and calendars
2. Uses AI to create your personal productivity context:
   1. Automatically associates relevant content to a to do task
   2. Organizes associated and unassociated content
   3. Enrich to do tasks with content data
   4. Allow for manually adding content to a to do task
3. Your AI productivity companion:
   1. Drafts: proactively encourages you to start a task by getting it started for you. A draft is the AI's attempt at completing a task. It helps get the ball rolling, starting a task is often the hardest part.
      1. When it can, it will draft email or message replies, summaries, outlines, research, web search, RAG search, meeting invites, and other content for your review and iteration.
      2. Draft creation have two modes: autonomous and on-demand modes.
      3. There is UI for user review and iteration of a draft through prompting and adding context with suggested prompt quick action buttons.
      4. Drafts have version history that can be rolled back to like checkpoints.
      5. It uses your writing style and tone based off recommended and available style/personalizations.
      6. It does not nag and it is not annoying. It's your companion to getting things done.
   2. Organizes your content and to do tasks so you don't have to. Make productivity apps productive and not time wasted tagging, sorting, assigning, and organizing your tasks.

What's inside?

This Turborepo includes the following packages/apps:

Apps

• desktop: a Electron.js app with Tailwind CSS
• docs: Help and support guides/docs Next.js app with Tailwind CSS
• mobile: a React Native app with Expo
• web: a Next.js app with Tailwind CSS

Packages

• @mikan/ui: a stub React component library with Tailwind CSS shared by both web and electron applications
• @mikan/tailwind-config: shared tailwind configs for packages & apps using @mikan/ui
• @mikan/eslint-config: eslint configurations (includes eslint-config-next and eslint-config-prettier)
• @mikan/typescript-config: tsconfig.jsons used throughout the monorepo
• @mikan/emails - create this email templates and sending package
• @mikan/api - shared fetchers or api calling between nextjs + RN + electron-vite
• @mikan/drizzle? or prisma?
• @mikan/testing - shared configs and react-testing-library setups

Building Packages

• Uses git-cliff for versioning and releases
• cliff.toml for git-cliff configuration
• .github/workflows/changelog.yml for GitHub release action
• @mikan/ui build details

pnpm run version-update
pnpm run release

Utilities

This Turborepo has some additional tools already setup for you:

• Tailwind CSS for styles
• TypeScript for static type checking
• ESLint for code linting
• Prettier for code formatting

Handle a team of AI agents

The strongest developers are learning how to break down problems into smaller chunks for multiple agent developers to work on, and then tying together the solutions they get back as a result. “No one’s an expert at this yet, but one of the key skills we’ll see going forward is spending time on markdown files and providing additional instruction and context to LLMs, almost like you would help an intern ramp up on a problem,” says Marcel. “This looks like: ‘Please think about this, consider this part of the problem, and look at this part of our solution space that we’ve already developed.’”