ThoughtStream LabSign in

User manual

Everything ThoughtStream Lab does and how to do it, from the first thought to a plan you can cancel.

Ask how to do something

Ask in your own words — “how do I get my dates into my calendar?” — and the answer comes from this manual. With voice control on, you can ask out loud as well.

Sign in to ask

Getting started

ThoughtStream Lab is a private place to notice what you notice. You catch a thought in the moment, it joins your own ledger, and over time the assistant helps you see what keeps coming back and what to do about it.

  1. Make an account from the sign-in page. Capturing thoughts is free, and no plan is needed to start.
  2. On Capture, type what is on your mind, or tap the microphone and say it. A thought is worth keeping before you have decided whether it matters.
  3. Do that whenever something arrives. Two or three a day is usually enough for patterns to start showing.
  4. Open any thought and run Analyze when you want the assistant's read on it: a summary, where it sits on the evidence scale, and what it connects to.
  5. When something keeps returning, turn it into a commitment and put a date on it. The app writes to you before that date, and your weekly review shows how it went.
  • Watch the From Thought to Action walkthrough video, with captions. It is available without signing in.
  • Put the app on your phone's home screen, or in your Mac's Dock, so capture is one tap away.
  • There is nothing to configure. The same account works in a browser, on a phone, and installed.
  • The app is light unless you ask for dark, whatever your device is set to. The moon at the top of every screen switches between the two, and your choice is kept on the device.

What you write stays yours. The assistant adds its own layer beside your words and never rewrites them.

The walkthroughWhat it costs

Asking how to do something

You do not have to find your way through this manual by reading it end to end. Ask the app how to do something, in your own words, and it answers from what is written here.

  • On this page, type your question into the box at the top: “how do I get my dates into my calendar?”, “how do I send a letter?”. The answer appears underneath, and the last few stay on screen for this visit.
  • Or ask out loud. With voice control on, say “how do I send a letter?” — the answer is said as well as shown, and “stop” cuts it short.
  • It answers in two or three sentences with what to tap or say, using the names you see on screen. When the manual does not cover something, it says so rather than guessing, and points you at the Contact page.
  • You need to be signed in to ask, and each question is one AI action, counted with the rest. It only covers using ThoughtStream Lab — for your own thinking, the coach, the analyst and a reflection are the ones to talk to.

If an answer is not good enough, the Contact page reaches a person.

Contact

Capturing a thought

Capture is the screen the app opens on, and the quickest way to get something out of your head.

  • Type in the box and press the arrow, or return. The thought is saved exactly as you wrote it.
  • Or tap the microphone in the box and speak. Your words appear as you talk, and the square stops and saves.
  • The line under the box reads a date out of the sentence as you write it — “by Halloween”, “next Friday”, “in three weeks”, “on the 31st” — and answers with the day of the week and how long is left. It is worked out on your device, from your own words; nothing is sent anywhere for it.
  • Your most recent thoughts are listed underneath, with a count of everything in the ledger. Open one to work on it.
  • The capture bar follows you from screen to screen, so a thought can be written down without leaving the page you are on.
  • Pull down from the top of a page to read the data behind it again.

Correcting a thought later never loses it: the words as you first captured them stay readable beside the edit.

Using it by voice

The floating microphone turns the app into something you can use with your hands full, while walking, or in the car.

  • Tap the microphone to start it listening. The bar along the bottom shows what it heard and answers out loud.
  • Say “take a note”, say the thought, then “save it” — “stop”, “done” and “that's it” end the entry and keep it just the same. Or keep it to one breath: “Note: call Bill tomorrow”.
  • Move around by voice: “open my ledger”, “patterns”, “my commitments”, “coach me”, “weekly review”, “this week”, “my profile”, “go back”.
  • Ask how to do something: “how do I send a letter?” — the answer is said out loud and shown in the bar. “Stop” cuts it short.
  • Ask it to “read the page”, “scroll down”, or “repeat”. Say “help” and it reads the whole list back to you.
  • To write into any box, tap the box first, then say “start dictating” and “stop dictating” when you are done.
  • Say “voice off”, “stop listening”, or just “stop”, to close the microphone.

Voice is transcribed by your own browser, not by us. The app stores the words, never a recording, and keeps nothing else the microphone hears.

Your ledger

Everything you have captured, newest first: the record you are building.

  • Search your own words, or filter the list by evidence state.
  • Open a thought to edit it, run the assistant over it, change its evidence state, turn it into a commitment, or write a letter from it.
  • Select several thoughts at once to act on them together.
  • Export takes everything you have — thoughts, patterns, commitments and reviews — out as a file of your own, whenever you want it.

A thought can be deleted, one at a time or in bulk. Deleting is the one thing here that cannot be undone.

The evidence scale

A thought is not a fact. Every entry sits somewhere on an evidence scale, and how far it has moved is part of what it means.

  • The assistant proposes a state and can say why, but it never applies one. Its reading stays in its own layer, marked as the assistant's.
  • Every move is kept in the thought's history, with who made it and why, so the path back is always readable.
StateWhat it means
CapturedA raw thought just recorded. No judgement yet.
ObservedSomething you noticed happening, stated as a fact.
RecurringThis keeps coming back across multiple thoughts.
InferredA hypothesis or interpretation — not yet confirmed.
ConfirmedBacked by enough evidence to treat as true.
RejectedRuled out by contradicting evidence.
ArchivedSet aside — no longer active, but kept for the record.

Rejected is not failure — something that contradicts a thought is evidence too. Archived sets a thought aside without losing it.

What the assistant does with a thought

Open a thought and there are five things to ask for. All of them are written beside your words, never over them.

  • Analyze — a short summary of the thought, where it sits on the evidence scale, how confident the reading is, and what it connects to among things you wrote earlier.
  • Research — what is known about the subject, from the web. It is kept in its own field and marked as coming from outside.
  • Proofread — your own words with spelling, grammar and punctuation put right. The meaning and your voice are left alone.
  • Draft — a short first draft written from the thought: something you could send, publish or keep.
  • Letter — a finished letter, written onto your own letterhead and signed off with your closing block.

Each is one AI action, whatever the length of the thought, and any of them can be run again after you edit. If a request never reaches the model, the action is handed straight back and does not come out of your allowance.

Discover: what keeps coming back

Discover reads across your whole ledger and proposes what repeats, what connects, and what contradicts what.

  • Detect starts a run. It looks for themes, recurring topics, connections between thoughts, contradictions, and opportunities.
  • Each pattern shows the thoughts behind it, and how strong the evidence is, so it can be judged rather than taken on faith.
  • Save the ones that ring true. A saved pattern stays beside the thoughts it came from.
  • When one is worth acting on, open the thoughts behind it and turn them into a commitment.

A run reads across the ledger, so it sharpens as the ledger grows. One run is one AI action, however many thoughts it reads.

Act: what you will do about it

Act is where an insight becomes something you do. A commitment is one thing you have decided to do, with the thinking behind it.

  • Create one with New, or from a thought with Commit. A commitment takes a title, optional detail, and an optional date.
  • Each one sits in Pending, Active, Done or Abandoned. Moving the status along is how the list stays honest.
  • Anything you date goes onto your own calendar as you save it, once a calendar is connected.
  • Coach, at the top of the screen, takes them one at a time: what you said you would do, what actually happened, and what you learned.

Nothing here is chased or scored. The app writes to you once about a date you set, and the weekly review is where the rest is looked at.

Being reminded about a date

Put a date on a commitment and the app will write to you about it — as early as you want to hear about it.

  • Choose how early in Settings, under Your assistant: the morning it comes due, or 1, 2, 3, 5, 7 or 14 days before.
  • The email lists what is coming due, the day each one falls on, and how long is left, with a link straight to your commitments.
  • Each commitment is mentioned once, on the first morning it falls inside the window you chose. Nothing is repeated day after day.
  • Only commitments with a date trigger a reminder. One with no date never does.

The reminder goes to the address on your account, and there is nothing in it but your own commitments.

Getting dates into your calendar

Two ways to get the dates you set into the calendar you already use. Neither needs a plan.

  • A private subscription link: the app gives you a web address your calendar app can subscribe to. Add it once in Google Calendar, Apple Calendar or Outlook, and the commitments you date appear there.
  • Google Calendar: connect your own Google account, and a dated commitment is added to your calendar as you save it.
  • Replace link mints a new address and stops the old one working. Worth doing if a link has travelled further than you meant.

A subscription calendar refreshes on the calendar app's own schedule, which can be a few hours. Adding a dated commitment through Google Calendar puts it there immediately.

The weekly review

Once a week, look back at what you noticed, what kept returning, and what you did about it.

  • Generate writes up your week from your own entries: what you captured, what kept coming back, what held up, and what you did about it, with your commitments alongside.
  • Reviews are kept, so you can read back over months and see what changed.
  • This week, in the bottom bar as Week, is the same seven days at a glance: what you captured, what is due, and what is done.
  • Guided reflection, at the top of your reviews, is the conversation version: one question at a time, written up as this week's review.

A review is one AI action. Capturing through the week is what makes it worth reading.

Letters, drafts and letterhead

ThoughtStream Lab can write from a thought onto your own paper, in your own voice.

  • Add a letterhead in Letterhead & templates: upload the file and write the closing block you sign off with. Letterhead files are kept in private storage, reached only through your account.
  • Save more than one signature if you sign letters differently, and choose which one a letter goes out with.
  • Write a letter from a thought. It goes onto your letterhead, in the voice you described in your settings, and signed off with your closing block.
  • Rewrite it as often as you like: every earlier version is kept, so nothing written for someone is lost.
  • Proofread and Draft are the shorter forms of the same idea, on the same thought.
  • Send a letter or a draft by email from the thought itself: put in the address and a subject. Sending is part of a paid plan, on every address including your own — free accounts keep everything else in the app — and there is a daily limit on how many go out. Otherwise, copy or download it and send it from wherever you like.

Letters go out from the app's own address, not from your own mail account, so a reply comes back to the app rather than to your inbox.

Four conversations with the assistant

Each one is reached from the screen it belongs to, and each works only from what is already yours.

  • Coach — from Act. One commitment at a time: what you said you would do, what actually happened, and a record kept behind the thought it came from.
  • Guided reflection — from your reviews. One question at a time through the week, then written up as this week's review.
  • The analyst — from Discover. What keeps coming back in your notes, and what you will do about it: a pattern saved once you recognize it, a plan saved once you agree to it.
  • The editor — from Letterhead & templates. Drafts, edits and tightens a letter or a first draft with you, in your voice, on the letterhead you chose and closed with your own sign-off: it works from your own words, asks for anything it does not have rather than inventing it, and hands you the text.
  • Every reply is typed onto the screen and then said out loud, in a voice your own device provides. Just above the conversation the speaking can be turned off, and the voice chosen.
  • A conversation is ended deliberately. Finish and review reads it back and says what came out of it, and nothing is written into your records until you confirm it. A finished conversation stays readable, and a new one starts beside it.

These are assistant actions like any other, and what they write stays on the screen it belongs to rather than being added to your ledger.

If you need help now

ThoughtStream Lab is a place to write things down, and the assistant is a writing aid. Neither is a crisis service, and neither can help with something that needs a person. If you are in danger, or thinking about ending your life or hurting yourself, please reach someone who can.

  • In the United States, call or text 988 for the Suicide & Crisis Lifeline. It is free, and someone answers at any hour.
  • Or text HOME to 741741 to reach the Crisis Text Line — free as well, and a conversation by message rather than by voice.
  • If you or anyone else is in immediate danger, call 911.
  • Anywhere else in the world, findahelpline.com lists the lines that answer where you are.

If you write something like this in a conversation, the assistant stops what it was doing and says this rather than carrying on. Writing it down here does not replace talking to someone, and whatever you write stays yours.

Contact us

Your assistant settings

Under Assistant, at the foot of any screen: how the assistant writes for you, which account it writes with, and how early you hear about a date.

  • Written in your voice: a note about how you write, in your own words. The assistant reads it, along with a sample of your recent thoughts, before it writes anything for you. Proofreading always corrects your own words rather than replacing them with someone else's way of saying it.
  • Your own AI key: on a paid plan you can save a key from OpenAI, Anthropic or Google Gemini. Analysis, patterns, reviews, proofreads, drafts and letters then run on your provider account and do not use your monthly allowance. Web research always does, because it needs the app's own search-grounded model. The key is scrambled at rest and is never shown again.
  • How early you hear about a date: the reminder lead time, from the morning it is due to fourteen days before.
  • Danger zone: delete my account, and everything in it.

Plans, AI actions and billing

Capturing, and reading back what you captured, is free and stays free. A plan pays for the assistant.

  • Free on every plan: capture by text or voice, your ledger, editing, evidence states and their history, commitments, your calendar, and exporting everything you have.
  • Counted against a plan: analyzing a thought, web research, a run of pattern discovery, generating a review, a proofread, a draft, a letter, and each message you send to one of the assistants. Each of those is one action.
  • Talking to the assistants always counts, even if you run the assistant on your own AI key, because the assistants use the app's own model.
  • Writing to somebody outside the app is part of a paid plan: Plus can post to 30 addresses a day, Pro to 100.
  • Free accounts get 5 AI actions a month. Plus and Pro raise that to 200 and 1,000. The current prices and what each tier includes are on the plan page and in Limits.
  • Pro runs the three deep readings — analyzing a thought, pattern discovery, and the weekly review — on the strongest model.
  • The count is per account and per calendar month, shared across every kind of action, and it does not carry over. The app shows how many are left and pauses when they run out; nothing you have already written is affected.
  • Cancel whenever you like. Cancelling stops the renewal and keeps your access until the end of the period you have paid for. A lifetime plan has nothing to cancel.
  • Cancel inside 7 days of a payment and that payment comes back in full. On the way out the app offers 33% extra time instead — 10 days on a monthly plan, 120 on a yearly one. Taking it or leaving it changes nothing about your right to cancel.
  • Unlimited time can be put on an account from the outside, for example while something is being tried out with you. Nothing is counted or capped while it runs.

What you have written is never withheld over a plan. If a plan ends, your ledger stays readable and exportable; the assistant is what pauses.

The plansLimits and what counts as an action

Gifts and invitations

Two ways to bring someone else in, or to give ThoughtStream Lab to someone.

  • Invitations: your profile holds your own code and a link. When a friend joins on it and makes their first payment, you both get free time. The reward waits for their payment, so nobody collects for accounts that never start.
  • A gift: buy a plan for someone else and the code arrives by email, with any note you write. They put the code in on the gift page and it is theirs.

A gift code is claimed once, and the plan it carries runs for the months it was bought for.

Privacy and your data

ThoughtStream Lab is built so that what you write is yours and stays yours.

  • Everything you write is scrambled at rest, and readable only through your own account, with a key derived for you alone.
  • Records are isolated per account. Nothing in the app, including the person who runs it, can read your thoughts from anywhere else.
  • Voice is transcribed by your browser, and no recording is kept.
  • There is no advertising in the app, and nothing you write is sold or used to advertise to you.
  • Export everything from your ledger whenever you want it: thoughts, patterns, commitments and reviews.
  • Delete your account, and everything in it, from Settings. It is permanent.

The privacy policyTerms of use

On your phone or in the Dock

ThoughtStream Lab runs in the browser, and installed it behaves like anything else on the device.

  • iPhone or iPad: open the app in Safari, tap Share, then “Add to Home Screen”.
  • Mac: in Safari, open the File menu and choose “Add to Dock”.
  • Android or another browser: open the browser's menu and choose “Install app” or “Add to home screen”.
  • On a phone or a tablet that has not installed it yet, a card sits at the top of every screen: on Android it runs the browser's own install prompt, and on an iPhone or an iPad it spells out the two taps — Share, then “Add to Home Screen”. Waving the card away keeps it away.
  • The Install link at the foot of the app offers the same from anywhere, a Mac's Dock included.

Installed, it opens full screen, stays signed in, and is otherwise identical to the browser version.

When something isn't working

Most of what goes wrong has a plain explanation.

  • Running out of assistant actions: the allowance starts again on the first of the month, upgrading raises it immediately, and a saved provider key takes writing and analysis off the count entirely.
  • The assistant did nothing: pattern detection needs a ledger with something in it to read across, and a letter needs a letterhead saved first. A request that never reaches the model is not counted.
  • No reminder arrived: a reminder needs a date on the commitment, an email address on your account, and the day to have come round. The app writes once, on the first morning the date is inside the window you chose.
  • A calendar is not showing a date: a subscription calendar refreshes on its own schedule, which can take a few hours. Adding a dated commitment through Google Calendar puts it there immediately.
  • Signed out and cannot get back in: use Forgot password on the sign-in page, and the reset link goes to your email address.
  • A page looks stale: pull down from the top of it to read its data again, which clears most of what looks stuck.

If none of that fits, the Contact page reaches a person, and the walkthrough video shows how to move from thought to action.

Contact usThe walkthrough