What a task buys and when it comes back
A task pays for one piece of AI work. If the work fails, the task returns to your balance and stays there.
Feature index
What Adamiro does, what it refuses, and when each one shipped.
Every feature, with what it does, why it works that way and what it will not do. Each entry carries the date it shipped.
Latest changes
All changes18 Oct 2026SignalsA new account runs no scheduled briefing until you switch it on.The schedule now decides when a briefing runs. Email the briefing only decides the mail. A new account starts on Manual only.New
17 Oct 2026 · v2.22.0AI VisibilityAI Visibility follows an idea from Voice draft to Reach post to publication, then shows what changed for that question.Shipped.New
16 Oct 2026 · v2.22.0VoiceA view Adamiro learns about you reaches no draft until you confirm it.Shipped.New
16 Oct 2026 · v2.22.0VoiceA draft Reach has published or archived leaves the Voice Drafts list.Shipped.New
How it fits together
Nothing goes out without you.
Reach
A person approves every post.
Reach holds each post in your queue until you approve it, schedule it or press Post now. Each platform gets its own version of the post, and you choose which ones.
Adam
Adam never publishes.
Ask in a sentence and Adam shows a plan card with the screens, the controls and the total task cost. One press of Start approves it. Adam never posts. Talking to it is free.
Reach
Campaign posts arrive as drafts for you to approve.
A campaign writes its posts into your Reach queue as drafts. On the campaign page each post has its own Approve button, and a post is scheduled only after you approve it or give it a time yourself.
Nothing invented reads as fact.
Voice / Reach
Every quotation is checked against the pages Voice read.
Quotes of four or more words are matched to the source pages. One it can't find becomes a fact to verify, with the reason stated.
Since 6 Oct 2026Read entry
Signals
A buyer quote is shown only if it is in the source.
Buyers checks each quote against the text it came from, ignoring capitals, spacing and curly quotes. A quote that is not in the source is dropped and you get a paraphrase in italics instead.
Since 13 Sep 2026Read entry
Voice / Reach
Voice marks a missing figure as a fact to verify instead of guessing.
Where Voice has no figure for a claim, it writes a marked placeholder that says what is missing. The draft lists each open one, and the marker stays until you replace it.
AI Visibility
The AI Presence Grade stays blank when it would mislead.
A report that got answers back for fewer than half of its questions gets no grade. The dashboard keeps your last complete grade and says it is older.
You don't pay for what you didn't get.
Signals / Radar
A briefing that cannot be read is not sent and tasks come back.
If the write-up of a Signals or Radar briefing cannot be read, no briefing email goes out and the run's tasks are returned to you. The articles are offered again on the next run.
Tasks
The task price is printed on the button before you press it.
The Signals, Voice and Mentor start buttons carry a badge with the current price. The number comes from one price list, so the badge and the charge agree.
Tasks / Voice
A Voice draft you throw away within 24 hours gives its tasks back.
Discard a Voice draft unedited within 24 hours of writing it and the tasks return to your balance. Up to three a week for each person, so it covers drafts that missed, not free writing.
Since 20 Sep 2026Read entry
Tasks
On the seat model, tasks never expire, refunds included.
Packs, welcome tasks, discount bonuses, refunds and support grants all stay in your balance until you spend them. There is no monthly allowance that resets.
Since 20 Sep 2026Read entry
You see what it keeps.
Mentor
Mentor's memory is encrypted on disk with a key tied to your account.
The notes Mentor keeps and the text of everything you teach Adamiro are stored encrypted. Only titles and file names stay readable, so your documents can be found and listed.
Mentor
A note Mentor keeps links back to the conversation it came from.
A note from a conversation carries a Where it was said link when its message is recorded. It opens the thread at that message. Without a recorded message, the note links to the thread. Notes also show their source.
Since 25 Sep 2026Read entry
The Tank
A new Tank session starts from your Pitch File, not from earlier conversations.
A new session works from your Pitch File and the names of recent advisers, not the content of earlier sessions. Continue carries the topic forward on purpose.
Platform
Nothing you write or Adamiro generates trains a model without your written consent.
No consent, no training. The code that could keep a sample for training does nothing unless the capture switch is on and your account holds a recorded consent. Consent is off until you give it.
All entries
Signals
Signals / Radar
A briefing's strength badge is checked against the items it read.
Each briefing carries one Strong, Moderate or Weak badge with a count, such as "4 of the 12 items". Adamiro checks that count against the items the briefing read and lowers a badge they cannot carry.
Since 26 Sep 2026Read entry
Signals / Radar
Signals tells the model one source is one voice.
Signals and Radar tell the model that a source claiming "everyone is doing this" is one voice, not a majority, when it states how strongly a finding is backed.
Since 26 Sep 2026Read entry
Signals / Radar
A briefing that cannot be read is not sent and tasks come back.
If the write-up of a Signals or Radar briefing cannot be read, no briefing email goes out and the run's tasks are returned to you. The articles are offered again on the next run.
Signals / Radar
A run that finds nothing new, or nothing relevant, is not charged.
If your sources published nothing since the last briefing, or nothing was relevant to your profile, no briefing is made and the run is not charged. The timeline says "Run not charged".
Since 6 Oct 2026Read entry
Signals
A buyer quote is shown only if it is in the source.
Buyers checks each quote against the text it came from, ignoring capitals, spacing and curly quotes. A quote that is not in the source is dropped and you get a paraphrase in italics instead.
Since 13 Sep 2026Read entry
Signals
Comments on your own posts are read without commenter names.
Buyers reads comments on your own Instagram and TikTok posts without asking for or storing who wrote them. Only the text is kept. Hashtag posts are read as captions, never usernames.
Since 13 Sep 2026Read entry
Signals
You decide which buyer segments are real, with a note of your own.
Each ideal-customer card has Won a deal like this, Lost a deal like this and Not a real segment buttons. Your verdict changes the card, and nothing is deleted.
Since 13 Sep 2026Read entry
Signals
Reviews you paste into Buyers are read at no task cost.
Paste a competitor's reviews on the Buyers tab, one per paragraph, and Signals reads them. It takes up to 50 a paste and 200 a week. Nothing you paste is published.
Since 13 Sep 2026Read entry
Signals
Signals can bring back one old post that did well, once.
A scheduled briefing may suggest one idea called "Revisit: ...", built on a post older than six months that did better than your typical post. Each post comes back once.
Since 8 Oct 2026Read entry
Signals
Questions in a briefing come from sources Signals read.
The briefing's Questions buyers ask section lists up to five questions, with who asked and a link. A question appears only if it ties to a source Signals read. Each can become a Voice idea.
Since 8 Oct 2026Read entry
Signals
A source that can't be reached says so on its row.
Catalogue sources are checked weekly. One that cannot be reached shows "Connection lost" on its row and adds nothing to your briefings. A feed that has moved is followed, and nothing changes for you.
Since 1 Sep 2026Read entry
Signals / Radar
A new account runs no scheduled briefing until you switch it on.
A new account starts on Manual only, so no Signals or Radar briefing runs on its own until you pick a schedule in Settings. Email the briefing only decides whether mail goes out.
Signals
A Signals run costs 1 task per 10 sources ticked, shown first.
The price follows the sources you ticked, rounded up by tens, plus 1 task if you want ideas drafted for Voice. Ideas can be switched off for a plain read.
Since 20 Sep 2026Read entry
Voice
Voice / Reach
Every quotation is checked against the pages Voice read.
Quotes of four or more words are matched to the source pages. One it can't find becomes a fact to verify, with the reason stated.
Since 6 Oct 2026Read entry
Voice / Reach
Voice marks a missing figure as a fact to verify instead of guessing.
Where Voice has no figure for a claim, it writes a marked placeholder that says what is missing. The draft lists each open one, and the marker stays until you replace it.
Voice
Voice settles the headline before it writes a long article.
For a blog article, LinkedIn article, newsletter or how-to guide, you pick a headline first. Voice plans the sections to keep what it promises. Headlines are free, and a number you didn't supply is dropped.
Since 16 Oct 2026Read entry
Voice
Voice's second read can't add a number or link the first draft lacked.
After the first draft, Voice edits it once more (not one-page sheets). If that edit adds a figure or link the first draft never had, the edit is thrown away and the first draft is kept.
Since 3 Sep 2026Read entry
Voice
The Editor Check reports what it finds and never rewrites your draft.
After a draft is written, Voice counts what a machine can count: stray dashes, hashtag blocks, announced openers, chopped rhythm. It posts notes on the generation timeline and leaves your text alone.
Voice
Words on your Never use these list are removed from drafts.
After a full draft is written, Voice removes each word on your Never use these list, or swaps it for the word you chose. The same pass restores your brand names to the spelling you set.
Voice
A view Adamiro learns about you reaches no draft until you confirm it.
Adamiro suggests positions from what you said to Mentor and how you edited drafts. Each shows the words that led to it. A suggestion enters a draft only after you accept it.
Since 16 Oct 2026Read entry
Voice
Voice never changes your saved voice without you pressing Use it.
Adamiro watches how you edit drafts and proposes vocabulary and tone changes. They wait in a list with their evidence. Your voice changes only when you press Use it.
Voice
Voice is told not to invent a client, number or quote.
A Case Study starts with a short fact interview. Every field has an OK to invent box, off by default. Voice is told to keep blanks general and to mark any figure you allowed it to invent.
Voice
Voice writes a comment for you to post, and never posts it.
Paste someone else's LinkedIn post and Voice writes a short comment built on your own proof facts or Memory. You copy it and post it yourself. Voice does not store the pasted post.
Since 8 Oct 2026Read entry
Voice
More like this adds ideas, never figures the first idea lacked.
More like this writes up to five ideas on the subject of one idea, each from a different side. An idea that carries a number the starting idea didn't have is dropped.
Since 8 Oct 2026Read entry
Voice
Other hooks drops opening lines with numbers you never supplied.
On a LinkedIn draft, Other hooks suggests three more first lines. A suggestion that carries a number found in neither your post nor your proof facts is dropped. It is free.
Since 8 Oct 2026Read entry
Voice
Technical drafts get a number check that warns and never rewrites.
On technical formats, Voice flags any figure it cannot trace to your topic or sources, including a figure cited to the wrong source. The warnings appear as notes. Your draft is not changed.
Voice
Voice emails you a finished draft only if you tick the box.
Email me a copy when it is written is off until you tick it. The box remembers your last answer, and the copy goes to the address of whoever pressed the button.
Since 10 Sep 2026Read entry
Voice / Reach
A draft Reach has published or archived leaves the Voice Drafts list.
Once a post made from a draft is published or archived in Reach, the draft no longer sits in Voice Drafts. A failed post that isn't archived keeps its draft visible, so you can rework it.
Since 16 Oct 2026Read entry
Voice
A press release invents a quote only when you tick the box.
The Write as dialog shows "OK to invent a quote" when you pick Press release, unticked. Without the tick, Voice tells the writer to quote no one you did not name, and any quote it adds is checked.
Reach
Reach
A person approves every post.
Reach holds each post in your queue until you approve it, schedule it or press Post now. Each platform gets its own version of the post, and you choose which ones.
Reach / Voice
Reach won't approve a draft that has an open fact.
If a post still holds a fact to verify, Reach refuses to approve it, schedule it or post it. The message names what is open, up to five items. A post scheduled before the gap existed is sent as scheduled.
Since 6 Oct 2026Read entry
Reach
Campaign posts arrive as drafts for you to approve.
A campaign writes its posts into your Reach queue as drafts. On the campaign page each post has its own Approve button, and a post is scheduled only after you approve it or give it a time yourself.
Reach
Themes are handed to every campaign post as instructions.
Each Themes line goes to the planner and to every post's writer as an instruction about what to cover and leave out. Choosing only your themes reads nothing from your Memory.
Since 12 Oct 2026Read entry
Reach
A campaign plans fewer posts rather than invent topics.
If your material supports fewer posts than you asked for, the plan is shorter and the preview says by how many. The price shown is for the posts actually planned.
Reach
Reach picks posting times per weekday and shows what each rests on.
Use recommended times fills a weekly pattern from LinkedIn averages, blended with your own measured posts once you have enough. Each row says whether it rests on the average, a blend or your own posts.
Reach
A time Reach is testing is marked in the plan.
About one campaign post in four goes out at a time you haven't tried, so the next report has something to compare. Those posts say testing a new time. Tick Stick to my best times to turn it off.
Since 31 Aug 2026Read entry
Reach
A figure no post reported reads Not collected, never zero.
The month-by-month card sets one month of LinkedIn posts beside the one before, with every post read at the same age. A figure nobody reported shows Not collected, and a reported zero stays a zero.
Since 7 Oct 2026Read entry
Reach
The follow-up list stores counts, never who commented or saved.
The Follow up tab lists your LinkedIn posts from the last 60 days that gained comments or saves since you last marked them. Adamiro stores numbers only, never the people.
Since 7 Oct 2026Read entry
Reach
Reach tells you before a sign-in runs out under scheduled posts.
Once a day Reach looks for LinkedIn and Instagram sign-ins that run out within two weeks while posts are scheduled past that date. It emails you once, notifies you, and puts a Sign in again note on your Reach pages.
Since 10 Sep 2026Read entry
Reach
Reach says which platforms it posts to and which you paste into.
LinkedIn, Instagram and your own WordPress site publish through a connection. Substack, a blog, LinkedIn Articles, X, Reddit, Hacker News and technical documents are copy and paste. TikTok is read-only.
Reach
Reach asks once whether a pasted post went out.
Reach can't see a copy-and-paste post go live. If it is unconfirmed a day after you copied it, you get one email, and one click on Yes, I published it closes the post.
Reach
Reach checks that a pasted live link belongs to the right site.
When you mark a copy-and-paste post as published, the live link is checked: X, LinkedIn, Reddit and Hacker News links must be on their own site. A link on the wrong site is refused and nothing is saved.
Since 6 Oct 2026Read entry
Reach
Deleting a Reach post deletes its numbers and replies too.
Delete a post and its numbers, stored replies, follow-up mark, image slots, saved wording history and comments go with it. Your template catalogue and media library stay.
Since 7 Oct 2026Read entry
Reach
A version that already went out is not sent again.
If a post has several versions and one already published, a retry sends only the rest, so a version that already went out is not sent again.
Since 4 Sep 2026Read entry
Radar
Radar
Radar keeps briefings out of Voice until you allow sharing.
Sharing starts off. Until you switch on Allow public sharing in Settings, no briefing item or whole-week summary can be sent to Voice. Radar itself publishes nothing. The weekly email still goes to your recipients.
Radar
Competitive and account signals have no send icon in Radar.
Only market moves, technology and regulation items carry a send icon, and only with public sharing on. Competitive signals and Account signals never do. The whole-week button still draws on the headline.
Since 28 Aug 2026Read entry
Radar
Copy from Signals adds terms to your watchlist and removes none.
The button adds the terms Signals watches that Radar does not already have. It matches without regard to case, changes nothing else and costs no tasks.
Since 8 Oct 2026Read entry
Radar
Competitors come from your profile; other organisations are a separate list.
Direct competitors come from your company profile. Partners, suppliers and regulators sit on a separate Radar list, and Adamiro asks the model to read them for what they mean to you. An empty profile falls back to that list.
Since 19 Sep 2026Read entry
Radar
Radar drops job ads whose employer name lacks the company name.
For each account you watch, Radar searches Swedish public job ads. An ad counts only when the employer name contains the company name. A longer name of another employer can still match.
Since 8 Sep 2026Read entry
Radar
Radar skips a company website when robots.txt blocks the whole site.
Radar reads the home page on every briefing run, and pricing and careers pages at most weekly. If robots.txt blocks Radar or all readers from the whole site, it fetches nothing. Page-level rules are not read yet.
Since 8 Sep 2026Read entry
Radar
Radar reports a watched page only when the start of its text changed.
The first time Radar sees a page it records it and says nothing. Later it reports a page only if the first part of its readable text differs from the last visit, and names the page.
Since 8 Sep 2026Read entry
Mentor
Mentor
Documents marked Only Mentor can't be read by anything that writes or publishes.
Hand over a business plan, budget or contract as Only Mentor. Mentor reads it and names it. Voice, Reach and Campaigns can't see it, and the separation is enforced in storage, not on screen.
Mentor
Mentor's memory is encrypted on disk with a key tied to your account.
The notes Mentor keeps and the text of everything you teach Adamiro are stored encrypted. Only titles and file names stay readable, so your documents can be found and listed.
Mentor
A note Mentor keeps links back to the conversation it came from.
A note from a conversation carries a Where it was said link when its message is recorded. It opens the thread at that message. Without a recorded message, the note links to the thread. Notes also show their source.
Since 25 Sep 2026Read entry
Mentor
A business fact from your chats is kept only if you said it.
When Mentor saves a thread to memory, a fact about your business needs a short quote from something you wrote. A line Mentor wrote doesn't count. Decisions Mentor files itself are in its own words.
Since 28 Sep 2026Read entry
Mentor
Mentor is given the Knowledge Library's real count and framework names.
Mentor is handed the Knowledge Library's real count by type and the names of its frameworks, up to 60. It is told to answer from that list and not to count Adamiro's own modules.
Since 31 Aug 2026Read entry
Mentor
Mentor charges per half hour of conversation, not per thread or message.
The first message after a quiet spell starts a half hour. Everything you send in the next thirty minutes, in any thread, is part of it. A retried send never charges twice.
Since 25 Sep 2026Read entry
Mentor
Mentor replies in your browser's language until you pick another.
Mentor doesn't guess the language from what you type. It uses your browser's language, or the one you set under Settings, Reply language, and keeps to it.
Mentor
A correction you give Mentor applies in every later conversation.
Say that something is wrong and Mentor keeps the correction as a standing rule, up to 30 at once. The list shows under Standing corrections. Forget one and it stops applying from the next reply.
Since 23 Sep 2026Read entry
Mentor
Mentor names both notes when two of yours disagree.
When it saves new notes, Mentor compares them with older ones and records where they disagree. When two notes disagree and it bears on your question, Mentor names both with their dates and asks which holds.
Mentor
The free Sounds Like Them test stores nothing about either site.
Two homepages in, three interchangeable sentences out. No URL, text, sentence pair or diagnosis is written to options, tables or logs. Your browser holds the text between steps.
Since 15 Sep 2026Read entry
The Tank
The Tank
A new Tank session starts from your Pitch File, not from earlier conversations.
A new session works from your Pitch File and the names of recent advisers, not the content of earlier sessions. Continue carries the topic forward on purpose.
The Tank
The Tank convenes five advisers for each round.
Each round the Chairperson picks five advisers. If it cannot pick five, it asks what you want to put to the board instead of running a smaller one.
The Tank
Each Tank adviser is told to hold its position unless given a reason to move.
Every adviser is instructed to keep its position unless someone gives it a reason to change, and to name the flaw in its earlier reasoning if it does.
The Tank
The Tank counts the board's verdict and reports a split as a split.
On a real decision the Chairperson names the options first, each adviser backs one or declines, and a dealbreaker shows even when it loses. An option without enough weight returns as a split.
The Tank
The Tank tells you when an adviser did not weigh in.
If an adviser does not respond in time, the Chairperson says how many did not. You never get a smaller board presented as five voices.
The Tank
A Tank session costs a flat 3 tasks.
One price buys the session and its block of adviser time. It does not grow with the number of rounds you run inside that block.
Since 20 Sep 2026Read entry
The Tank
If a Tank session cannot be saved after payment, the tasks are refunded.
If the session record cannot be saved after the tasks were taken, the tasks go back and the message says nothing was charged. This covers starting a session and extending one.
Since 29 Aug 2026Read entry
The Tank
The first three Continues in a Tank conversation cost nothing.
Continue carries your topic and living summary forward at no charge, up to three times. The fourth is refused with a clear message.
Since 20 Sep 2026Read entry
The Tank
The Pitch File is a free-text profile you can read and edit.
It is the one thing that carries between sessions. You change it when you like. In a Pitch Session the advisers are told to ask questions and hold back advice.
The Tank
The First Round does not save your idea.
Your offer, the site text it reads and the board’s answer are not saved to a table, option, transient or log. The lead record holds your email, language and consent.
The Tank
The First Round names what would prove its verdict wrong.
The verdict closes with one checkable sentence, such as a renewal count or a date. When the board names nothing, that line is left out.
The Tank
A First Round that fails on our side does not count against you.
The round is given back to your allowance when it fails. The allowance is a small number of rounds per week per browser, plus an hourly limit per connection.
AI Visibility
AI Visibility
The AI Presence Grade stays blank when it would mislead.
A report that got answers back for fewer than half of its questions gets no grade. The dashboard keeps your last complete grade and says it is older.
AI Visibility
The free AI Visibility Checker won't invent questions about a site it can't read.
If the site returns a bot check, has no readable text, or never says plainly what you sell, the check stops and tells you which. It does not make up five buyer questions.
Since 1 Sep 2026Read entry
AI Visibility
The free AI Visibility Checker reports a count and never a grade.
It asks five questions of one AI assistant and says how many named you. The word grade never appears, so it can't be mistaken for the report.
AI Visibility
The free AI Visibility Checker returns the same result for a domain checked again.
A domain checked twice within a month gets the stored result. No second run is made.
AI Visibility
AI Visibility shows Not measured, not zero, when no question left your name out.
Named unprompted counts only questions that don't contain your name. With none to count, the card says Not measured. A measured zero stays a zero.
Since 2 Sep 2026Read entry
AI Visibility
AI Visibility records which tracked competitors each answer named.
Answers are matched against your company and the competitors you track. When an answer picks one brand, that pick is recorded, including a brand you don't track.
AI Visibility
Crawler checks on your pages run only after you prove the domain.
The crawler check makes our servers fetch your pages, so the domain needs proof first: a tag in your head, a DNS record or a small file. Presence and the free checker read a home page without it.
AI Visibility
Scheduled AI Visibility reports run weekly at most.
The schedule offers weekly, monthly or manual. You can also start a report by hand at any time, one at a time, and each costs the same task.
AI Visibility
Each seat's weekly AI Visibility read includes ten prompts.
Prompts past the first ten cost 1 task per extra ten, rounded up, plus an engine surcharge if you add Claude as an extra engine. Writing a question set, editing competitors and reading any screen cost nothing.
Since 20 Sep 2026Read entry
AI Visibility
Asking Adam about your visibility opens the AI Visibility dashboard and spends no tasks.
Three questions about your visibility each get a card that opens the AI Visibility dashboard at Presence. You read the numbers there. The card costs no tasks.
Since 10 Sep 2026Read entry
AI Visibility
AI Visibility follows an idea from Voice draft to Reach post to publication, then shows what changed for that question.
Send an idea to Voice and AI Visibility notes how that question looked. Once the piece is published through Reach and a later report has run, the Impact panel shows what changed since, not what caused it.
Since 17 Oct 2026Read entry
Adam
Adam
Adam never publishes.
Ask in a sentence and Adam shows a plan card with the screens, the controls and the total task cost. One press of Start approves it. Adam never posts. Talking to it is free.
Adam
Adam answers how-it-works questions from the Adamiro guide.
Ask how a feature works and Adam tries to answer from the user guide, with a link to the module. When the guide doesn't cover your question, it says so and points you to the full Guide.
Adam
Asking Adam a question costs no tasks.
Asking where something is, or how a feature works, is free. You pay only when you press a button that runs a generation yourself. A fair-use daily limit caps how many requests you can send.
Adam
Adam only offers steps from a fixed list of things it knows.
Every step Adam names comes from a fixed catalog. A request outside it, such as publishing for you, is refused. If only part of your request is outside it, Adam says which part and answers the rest.
Adam
Adam points questions about your own business to Mentor.
Ask Adam whether to raise your prices and it says that is a question for Mentor. A Take this to Mentor button opens Mentor with your question in it. Nothing is sent until you press send there.
Since 2 Sep 2026Read entry
Adam
Adam turns several requests into one ordered list of steps.
Ask for a rewrite, an image and a send to Reach in one sentence and Adam returns numbered steps in the order they have to happen, each with its own instruction. A plan holds at most four steps.
Adam
Adam asks you a question when it would have to guess.
When a request lacks something a step needs, Adam asks instead of picking a value. No step is offered until the missing piece is supplied.
Adam
Adam only offers steps your plan and seat allow.
Before Adam lists a step, it checks that your plan includes the feature behind it. A post can't be scheduled through Adam on a plan without scheduling, and images need graphics on the plan.
Adam
Adam only takes you to your own dashboard screens.
When Adam moves you to a screen, it only follows addresses for your own Voice, Reach and AI Visibility pages. A link to another site or another account is blocked, and a saved plan is dropped if the account changed.
Adam
Adam's problem report lists everything it sends before you send it.
The Report a problem form in Adam's panel lists the page, address, last request, plan progress and browser it will attach. Your note is cut to 1,000 characters when it is saved. No AI reads the report.
Tasks
Tasks
The task price is printed on the button before you press it.
The Signals, Voice and Mentor start buttons carry a badge with the current price. The number comes from one price list, so the badge and the charge agree.
Tasks / Voice
A Voice draft you throw away within 24 hours gives its tasks back.
Discard a Voice draft unedited within 24 hours of writing it and the tasks return to your balance. Up to three a week for each person, so it covers drafts that missed, not free writing.
Since 20 Sep 2026Read entry
Tasks
A run that fails returns its tasks to your balance automatically.
When a run ends as failed, including one that stalls and is swept, the tasks it used are refunded without you asking. The refund happens once per run.
Tasks
On the seat model, tasks never expire, refunds included.
Packs, welcome tasks, discount bonuses, refunds and support grants all stay in your balance until you spend them. There is no monthly allowance that resets.
Since 20 Sep 2026Read entry
Tasks
Every module spends from the same task balance.
Signals, Radar, Voice, Mentor, The Tank, AI Visibility and Reach's AI media take tasks from one balance. Adam's campaign plan is free. There are no per-module budgets to top up.
Tasks
Only the account owner can buy tasks.
On Team, a Co-Editor has no buy button. When the balance runs out they read that the account owner adds tasks, because the owner's card is the one charged.
Since 2 Oct 2026Read entry
Tasks
The task ledger lists who pressed the button for every line.
On Plan and billing, every task spent or returned is listed with when, module, what and who. The owner reads every line. A Co-Editor reads only their own.
Since 6 Oct 2026Read entry
Tasks
The trial is 15 tasks with no card, and it ends when they are gone.
A trial starts with 15 tasks and no payment card. It does not end on a date. When the tasks run out the dashboard stays readable and everything you built stays.
Since 20 Sep 2026Read entry
Tasks
Your first seat payment adds a one-time welcome bundle of tasks.
The first cleared payment for a seat adds 20 tasks on Solo or 50 on Team, once per account. A renewal does not add it again, and these tasks do not expire.
Since 20 Sep 2026Read entry
Tasks
Auto top-up is off until the owner turns it on.
The owner picks a threshold and a pack. Adamiro then buys that pack when the balance falls to the threshold. It never fires during a trial, and a failed charge pauses it.
Since 20 Sep 2026Read entry
Platform
Platform
Nothing you write or Adamiro generates trains a model without your written consent.
No consent, no training. The code that could keep a sample for training does nothing unless the capture switch is on and your account holds a recorded consent. Consent is off until you give it.
Platform
Adamiro's main text model runs in an EU region.
The model that reads and writes most of your text is pinned to an EU region, and a weekly check watches the pin. Some features send text outside the EU. The data-protection page lists them.
Platform / Mentor / Radar
What you teach Adamiro is stored encrypted under a key for your account.
Mentor memory, Radar content and the text of everything on your Memory page are encrypted on disk. Titles and file names stay readable so your documents can be listed.
Platform
Two-step sign-in is on for every account, and you can't turn it off.
After your password you enter a code. An emailed code works from day one and an authenticator app is optional. You can't switch it off yourself. A device you trust skips the code for 45 days.
Since 30 Sep 2026Read entry
Platform
A recovery code works once, and a new set cancels the old one.
You get 10 recovery codes to use in place of an email or app code. Each works one time. Making a new set cancels every unused old code, and Adamiro emails you when that happens.
Platform
The owner can cap how many tasks a Co-Editor spends each cycle.
A Co-Editor spends from the owner's shared task pool. Set a limit per billing cycle on a person and a run that would go past it is refused. Leave it blank and they draw on the whole pool.
Platform
The owner chooses which modules each Co-Editor can use.
Every module starts open to a new Co-Editor. The owner can switch any of them off for one person. A switched-off module stays visible with an Ask button that emails the owner.
Since 22 Sep 2026Read entry
Platform
An ownership transfer changes nothing until the new owner accepts and adds a card.
You name the new owner and confirm. They get a link that works for 7 days. Nothing moves until they accept and add their own card, and you can cancel any time before that.
Since 24 Sep 2026Read entry
Platform
Each agency client gets its own workspace account.
A client workspace is a separate account with its own owner account, set up for the agency owner. The agency pays for each workspace as a line on its own subscription.
Since 1 Oct 2026Read entry
Platform
A client can cap how many tasks each agency person spends.
When a client pays for the work itself, it sets a monthly cap on each agency person. A run past that cap is refused, and one person's spending doesn't count against a colleague's.
Since 1 Oct 2026Read entry
Platform
An agency can hand a client workspace to the client, and the client must accept.
The agency sends the client an emailed link that works for 7 days. Nothing changes until the client accepts. The workspace keeps its tasks and work, and the client pays from then on.
Since 1 Oct 2026Read entry
Platform
The public API is read-only and returns only content already public.
It serves pricing, the guide, documentation pages and the feature index as JSON, plus an OpenAPI description. It needs no login, reads no account data and runs no AI.
Since 9 Oct 2026Read entry
Platform
The public MCP server is anonymous, read-only and holds no session.
An assistant can ask it about Adamiro, its prices and its documentation. It reads no account data, runs no AI and keeps no session. Requests above 64 KB are refused.
Since 9 Oct 2026Read entry
Platform
Optional cookies stay off until you choose to allow them.
Analytics storage starts denied on every page. It switches on only if you accept analytics in the cookie banner. Your choice is stored in a first-party cookie and logged with a hashed address.
Platform
Closing an account erases nothing at once, and a closed account can be reopened.
Closing stops the plan and the card, and the account then waits about a month before it may be erased. In that time it can be reopened, though the plan does not come back.
Since 11 Oct 2026Read entry
Platform / Mentor / The Tank
You choose how long Mentor and The Tank transcripts are kept.
Keep them until you delete them, or have them cleared after 30, 90, 180 or 365 days. One button wipes every transcript now. Notes you saved to Memories stay.
Platform
When a Co-Editor is removed, what they made stays with the account.
Their work in the account moves to the owner, unless they also have a separate account of their own in it. Their other accounts and their login are left as they are.
Since 1 Oct 2026Read entry
No entry matches. Clear the filter or the search.