Your todo list should have teammates.

A todo app should make it harder to drift.

That sounds obvious, but most todo apps do something much narrower. They remember what you typed. They give you sections, labels, dates, projects, reminders, maybe a nice checkbox animation. If you already know exactly what matters and you already know what to do next and you already have the right amount of pressure in your life, that can be enough.

But that is not the normal failure mode. The normal failure mode is quieter. You have a goal that mattered in January. By March, your list is full of errands, fragments, half-promises, and tasks that were urgent when you wrote them down but have no obvious relationship to the person you are trying to become. You are busy. The app is organized. The goal is still starving.

That is the part we are trying to change with goals.

The latest version of goals. is built around that loop. Goals are no longer just folders for todos. Each goal can have a chat, people, AI agents, linked todos, recurring cadence, and a shared memory of what already happened. It is not a chatbot bolted onto a checkbox list. It is a small team around the thing you said mattered.

The new loop

The product shape is simple:

1. You name a goal.
2. You add people or AI agents who can help with it.
3. You talk in the goal chat about what needs to happen next.
4. Keen or a goal agent turns that conversation into actual todos.
5. Those todos get assigned to you, a teammate, or an agent.
6. Recurring work comes back daily, weekly, or when needed.
7. Agents can follow up in the chat, post progress, add subtodos, or hand work to a paired coding session.
8. The completed work feeds back into the goal, the weekly review, and the next decision.

That loop matters because most goals do not fail from one dramatic wrong decision. They fail because nobody notices the drift early enough. The workout plan gets skipped three times. The launch plan never becomes customer calls. The writing project turns into "research" for six weeks. The savings goal is theoretically important but never gets a weekly money decision attached to it.

Agents are useful when they make those slips visible while they are still small.

Agents are not side chats

In goals., an agent is an addressable AI teammate inside a specific goal. It has a name, a role, a personality, a tool set, and membership in the goal's team. You can mention it in the goal chat. You can assign it a todo. It can reply in the same thread where your human collaborators are talking. It can create or edit todos, add checklist items, schedule follow-ups, mark work done, and suggest the next move.

The important word there is inside. The agent is not floating in a separate AI tab, guessing at your life from a prompt. It is sitting next to the goal, the todo list, the completed work, the chat history, and the cadence of the plan. That changes the quality of the help.

If you ask a generic chatbot, "help me get fit," it can give you a perfectly reasonable workout plan. If you add a Personal Trainer agent to a "Run a half marathon" goal in goals., it can see that this week's tempo run is still open, that the Sleep Coach has been nudging recovery, that your weekly cadence resets Monday, and that you already completed strength work twice. The advice stops being generic. It becomes situated.

The agents you can add

We started with a small catalog of templates because a blank "create agent" box is a little too much like being handed a blank sheet of paper and asked to invent a coworker. The templates are starting points. You can rename them, edit them, or create your own from scratch.

For health and consistency

• Personal Trainer builds and adjusts weekly workouts, tracks overload, mobility, and recovery, and keeps the plan realistic.
• Sleep Coach protects the recovery side of the equation: schedule consistency, light exposure, wind-down routines, and not funding ambition with sleep debt.
• Nutritionist turns health goals into meals and snacks that survive real life, not rigid plans that collapse on day three.
• Meditation Guru keeps mindfulness small enough to repeat: shorter consistent sits, honest streaks, fewer dramatic restarts.
• Drill Sergeant brings no-fluff execution pressure when the goal needs standards more than softness.

For money, work, and building

• CFO owns savings rate, debt payoff, runway, taxes, and the next financial decision worth making.
• Entrepreneur Coach pushes pricing, growth, customer development, and shipping when business goals are getting fuzzy.
• Product Manager forces prioritization, sharpens user stories, and keeps the team building the smallest thing that moves the metric.
• Software Architect pressure-tests technical decisions before they calcify and names complexity while it is still cheap to change.
• Engineer pairs with a codebase through Claude Code, Codex, Cowork, or another MCP client so coding todos can move from the goal into a real development session and back again.

For learning, creative work, and communication

• Teacher designs a learning path, sets weekly study goals, picks resources, quizzes you, and tightens the loop when retention slips.
• Writer pushes on structure, voice, and what is earning its place on the page.
• Designer reviews flows, hierarchy, and copy with a UX eye.
• Marketing Specialist turns fuzzy positioning into sequenced launches, growth experiments, and hypotheses you can learn from.
• AI Specialist helps pick models, sharpen prompts, wire tools together, and spot where AI saves real time versus where it is just novelty.
• Gen Z Intern keeps the internet-native angle honest without turning every plan into a trend chase.

For the human layer

• Impartial Listener gives a calm second perspective on relationships and personal choices.
• Parenting Coach helps turn recurring family friction into practical, age-appropriate next steps.
• Accountability Partner has no domain expertise by design. Its job is to ask, consistently, whether you did the thing.
• Hype Man protects momentum when the work gets heavy.
• Yas Queen brings supportive energy with taste: celebrate the win, name the spiral, pick one confident next move.

The variety is the point. A CFO should not sound like a Meditation Guru. A Designer should not evaluate progress the way a Drill Sergeant does. Different goals need different kinds of pressure, and the app should let you choose the pressure that fits.

Cadence is the magic

The easy demo of AI agents is "ask an agent to make todos." That is useful, but it is not the interesting part.

The interesting part is cadence.

goals. supports one-off todos, daily todos, weekly todos, and reusable "when I need it" todos. Agent-assigned recurring todos can run on schedule and post back into the goal chat when they are due. That means the Personal Trainer can check in every week on the training plan, the Accountability Partner can ask every morning whether you did the habit, the CFO can bring the money decision back every Friday, and the Writer can keep a draft alive without you needing to remember to restart the conversation.

This is not about turning AI into a nagging machine. The cadence is explicit. You choose the goal, the agent, the todo, and the rhythm. The agent's job is to make the commitment reappear at the moment it has a chance to change behavior.

The best accountability is not loud. It is timely.

A reminder that says "write" at 9pm when you are exhausted is noise. A Writer agent that knows the draft goal, sees the last three checklist items, and nudges you Monday morning with "finish the intro before you open research again" is different. It carries context. It has a point of view. It keeps the goal from dissolving into the day.

A concrete example

Say your goal is: "Launch a paid beta by June 1."

In a normal todo app, you might create a project and start typing tasks:

• Landing page
• Pricing
• Beta users
• Email list

Not bad. Also not enough. Those tasks are containers for decisions you have not made yet.

In goals., you could add a Product Manager, Designer, Marketing Specialist, CFO, Engineer, and Accountability Partner to the goal chat. Then the work gets more specific:

• The Product Manager asks what metric defines a successful beta and turns that into a "choose beta success metric" todo.
• The CFO asks whether the price is testing willingness to pay or maximizing revenue and schedules a pricing decision for Friday.
• The Marketing Specialist turns "beta users" into a checklist: write outreach list, draft email, send 20 invites, follow up 48 hours later.
• The Designer reviews the onboarding flow and adds a "remove one step before paywall" todo.
• The Engineer takes the implementation todo, pairs through the Claude Code bridge, works in the repo, and posts back a branch or commit when it lands.
• The Accountability Partner checks in every weekday morning until the outreach actually happens.

That is the difference between a task list and a working system. The tasks are not just stored. They are argued into shape, assigned, followed up, and connected back to the outcome.

The same pattern works beyond solo productivity. A shared goal can hold a collaborative plan for a family vacation, wedding, house move, fundraiser, product launch, or retreat: people, todos, chat, and AI help in one place instead of scattered across group messages and memory.

How this differs from every other todo app

Most todo apps are optimized for capture and organization. That is valuable. But goals usually fail after capture.

The product category starts to feel different when agents can participate in the same primitives as everyone else. A todo is not only something you own. It can be assigned to a friend, a teammate, Keen, a CFO agent, or a paired coding agent. A chat is not just comments. It is where strategy becomes work. A recurring todo is not just a repeating checkbox. It can be a standing appointment with the version of accountability the goal needs.

The coding-agent bridge matters

For builders, this gets especially interesting with MCP. We built a Claude Code bridge so Claude Code, Codex, Cowork, or another MCP client can pair into a goal as an agent. Once paired, the coding agent can read the goal context, see open todos and checklist items, use the repo path attached to the goal, do the work in the codebase, then write back to the same chat as the agent.

This is not "copy this task into an AI tool and hope you remember to update the tracker." The tracker is the source of truth. The AI session joins it, does the work, and returns with receipts.

That loop is small today, but it points at a much larger product philosophy: the apps where your goals live should be able to hand work to the tools where the work actually happens. Writing, coding, planning, research, design review, outreach, budgeting - different tools, one goal thread.

Keen is still the center

The agents are powerful, but Keen still plays a different role.

Keen is the in-app goal companion that sees across your goals. It can parse messy voice capture into structured todos and goals. It can answer "what should I do first today?" from your actual open work. It can write weekly briefings from completed todos, voice reflections, calendar time, Health signals, and optional spending data. It can propose memories when you say something important about your preferences, routines, or constraints.

Think of Keen as the cross-goal guide, and agents as specialized teammates inside a specific goal. Keen helps you orient. Agents help you pursue.

The trust boundary

There is an obvious bad version of this product: AI agents spraying tasks everywhere, inventing urgency, and making your life noisier.

We do not want that.

Agents in goals. are owned by a human. They live on goal teams you choose. You can assign them work, mention them, rename them, edit them, or remove them. Cadence is attached to todos, not hidden in some mysterious agent brain. Tool actions show up in chat as visible chips. Weekly review and memory features are designed around user-approved context, not secret profile-building. The whole thing should feel like chosen accountability, not surveillance.

The line we keep coming back to internally is: the user keeps agency; the agents provide momentum.

The bigger argument

The future of todo apps is not prettier checkboxes. It is better follow-through.

That means the app has to understand the difference between "I captured a task" and "I moved the goal forward." It has to help create the right next step when the list is vague. It has to bring the right kind of pressure at the right moment. It has to let other people and agents own parts of the work. It has to remember enough context that every nudge is not starting from zero. And it has to close the loop every week so you can see whether your time, effort, and habits matched what mattered.

That is what we are building with goals.

Not a dashboard. Not a chatbot. Not another place for your tasks to quietly age.

A todo list with teammates.