So it's installed. Now what? Here's the honest answer: you leave a tiny trail every day, a checklist and a timer, and Lifebot reads it and tells you what it actually says. This page shows you what that looks like, how to make Lifebot sound like yours, and what to do when something's off.
You don't need to read it top to bottom. Parts 1 and 2 are the ones to read once. The rest is for when you want it.
Lifebot is three apps. Two small ones record your day. The third one reads what they wrote, plus your notes, calendar and sleep, and answers your questions. That's it. That's the whole system.
You focus and tick things off. They save plain files in Documents/Lifebot.
Reads those files plus your notes, calendar and Oura. Never edits them.
Ask in Telegram, Discord or Hermes Desktop. Get an evening summary every night.
You talk to Lifebot like you'd talk to ChatGPT. The difference is what it knows. It has read your focus sessions, your checklists, your calendar and your notes, so "how did today go?" gets an answer about your day, not a generic one. And your data stays local: your logs, notes and settings live on your computer, not on our servers. Only what's needed to answer a question goes to the AI model you chose, and with a local model, nothing leaves your machine at all.
The Lifebot window is for settings and data sources, not for chatting. Chatting happens in Telegram, Discord or Hermes Desktop. Five tabs on the left:

Is Lifebot running, which AI it uses, how fresh your data is. Restart gateway fixes most things.
Open Pixel Pomodoro or Typewriter, see where they save, find data from before Lifebot.
Your sources. Refresh snapshot now to update one, Current snapshot to see what Lifebot sees.
Lifebot folder, time zone, run at startup, and what it's allowed to remember.
Lifebot runs on Hermes Agent, an open-source AI agent. Launch Hermes Desktop on Overview opens Hermes's own app, and that's one of the two places you chat with Lifebot. Prefer your phone? Connect Telegram or Discord and message your bot like a person. Same Lifebot, same memory, either way. Hermes Desktop also shows every skill, scheduled job and model setting, if you ever want to look under the hood. /new starts a fresh conversation in any of them.

A little always-on-top timer buddy.

A pixel typewriter that turns your checklist into paper.
A checklist in the morning, a timer when you sit down, a tick when you're done. That's the whole trail. First time? Do the ten-minute Try it step in the install guide; it has a demo of each step.
Type three to six things into Typewriter, or ask Lifebot and paste its answer in:
Start a Pomodoro named after the top item. Log it when it rings, tick the item. Pulled away? Ctrl + C, say why. Not sure whether to keep going?
At 9:25 pm the Evening summary lands on Telegram or Discord. Reply to it like a friend:
Like the answer? Make it run by itself every week.
Lifebot ships plain and neutral on purpose. Most of what you'll want to change, you change by telling it. A few things live in the Settings tab. Pick what you want to change.
Just tell it, in the chat, and be specific.
/new afterwards.Want to rewrite the personality itself? It's a plain text file called SOUL.md in your profile folder:
~/.hermes/profiles/%LOCALAPPDATA%\hermes\profiles\
It's a scheduled job, so you change it by asking:
private: or put it under a heading called # private. Lifebot skips those parts.
Use an app Lifebot doesn't know, like a habit tracker, a mood log or a spreadsheet? If that app can export its data to a file, Lifebot can learn to read it. You don't write any code. You describe the file and Lifebot builds the connection itself.

You can also hand it facts that change how it reads your data:

A skill is a saved set of instructions Lifebot runs again by name. Anything you'd otherwise re-explain every week:
/skills pending then /skills approve in Telegram or Discord, or in Hermes Desktop's Skills pane./weekly-review, or just "run my weekly review".If you've ever thought "I wish it would just do this every day", that's a scheduled job. Ask for it in plain language:

Want the thinking to stay on your computer too? Hermes can run a model locally: its own Local Models option, Ollama or LM Studio. The demo shows the whole thing with Ollama.
gemma4:31b: about 20 GB of disk and 24 GB of memory.http://127.0.0.1:11434/v1 for Ollama or http://127.0.0.1:1234/v1 for LM Studio, no API key, Test, Save.Then ask Lifebot What did I complete today? If it answers with the sessions from your Pomodoro log, you're set. If it says it can't read files, or the numbers don't match your log, the model isn't using tools; switch back to a cloud one. Scheduled jobs run on the same model, so expect them to be slower. We haven't tested local models with Lifebot ourselves yet; the cloud options in the setup wizard are the ones we know work well. Hermes's local setup guide has the details, including the 64,000 context length Ollama needs.
Wondering what to actually do with it? Pick the situation closest to your week. Each one has prompts you can copy with one click; change the details to fit you.
Typewriter holds today's writing tasks, one scene or section per line. Each Pomodoro is a writing session, and the note is where the word count goes: 1,200 words, chapter 3. Your drafts stay in your notes app, where Lifebot can read them if you ticked those folders.
Name each Pomodoro after the client or project, like acme: proposal, and keep one Typewriter plan per day. Lifebot can then split your focus time by project, which no timer does on its own.
With your calendar connected, Lifebot sees your events seven days back and ahead, next to your focus log. That's enough to show what meetings cost you and where the gaps are.
If you connected an Oura ring, Lifebot has your sleep, readiness and activity scores next to your focus log. It won't give health advice. It shows what the two logs say together, once it has 14 nights of data.
Paste a syllabus or reading list into Typewriter and it becomes a checklist with a section per week. Name each Pomodoro after the subject, and Lifebot can tell you where your hours really went.
This one needs Telegram or Discord, because that's where you can send a photo, and an AI model that can see images (the ChatGPT, Anthropic and OpenAI options can).
Estimates, not measurements. Photo calorie counts can be way off, especially for sauces and portions. Treat them as a rough trend, and a caption like "half portion" helps.
Lifebot comes with its source code. The fourth download, lifebot-kit-source.zip, has the Lifebot app, Pixel Pomodoro and Typewriter as small projects you can change with a coding agent like Claude Code. You don't need to read code. You describe the change, the agent makes it, rebuilds the app and installs it.
If you just want Lifebot to talk differently, or your summary at another time, you don't need any of this. See Make Lifebot yours.
The app you open is a built copy, so every change ends with the agent rebuilding and installing it. That takes a minute or two, and sometimes a couple of tries. Normal. Ask it to run the app's tests before it builds.
y to already installed. Overwrite it? Your version is replaced with the original.| It can | Examples |
|---|---|
| Read | Pomodoro sessions and notes · Typewriter plans and tick times · the Obsidian, Joplin, Apple Notes and Notion folders you chose · calendar events 7 days back and ahead · Oura sleep, readiness and activity · photos you send on Telegram or Discord · the web, when you ask |
| Analyze | Focus time per day and week · completed versus abandoned sessions · your most productive hours · what keeps getting postponed · calendar load versus focus · sleep versus focus (after 14 nights) · themes in your notes |
| Schedule | Recurring summaries and check-ins · one-time reminders · refreshing your sources every 3 hours |
| Remember | Your preferences and facts you tell it · skills it creates for you |
| It can't | Why |
|---|---|
| Change your sources | It only reads your notes, calendar and health data. It never edits or deletes them. |
| See what you haven't connected | No access to other apps, email, or folders you didn't choose. |
| Run while your computer sleeps | Scheduled jobs need your computer awake and Lifebot's service running. |
| Know exact calories or sleep | Photo estimates are rough. Oura scores are Oura's own 0 to 100 scores, not medical measurements. |
Click the one that sounds like your problem. Whatever it is, Restart gateway on Overview is the first thing to try: it takes a few seconds and changes nothing else.
/new and ask again, or ask Lifebot:/new in Telegram or Discord, or start a new chat in Hermes Desktop. New preferences, memories and settings apply to new conversations.Lifebot can read its own logs and settings, so tell it what you expected and what you got:
/approve.YOUR_PROFILE replaced by your profile name from the bottom left of the Lifebot window, and send us the output with what you were trying to do.Want to start the personality over? Settings → Reset persona to defaults. Keys, memories, skills and jobs stay.
