Everything In One Place

Max, and yes. Computer is the part of Perplexity that actually does the work, and every module here uses it. Pro gets you in the door but has no monthly credits, so you will run dry fast.

A skill is a folder of instructions that teaches Computer to do one job your way. You install it from a zip inside your Project, or ask Computer to build one from a job you just did well. Module 9 walks you through it.

Give it real material, not descriptions. Your best writing, your offers and prices, a few testimonials, and the export from ChatGPT or Claude in module 5. Foundation turns all of that into your profile.

If the tool has a connector, use that. If it only gives you an API key, that goes in the Credential vault. If it has neither, Comet drives the website for you. Module 4 has a card for each tool.

The weekly work you should not be doing by hand. Email writes your emails in your voice. Content turns your calls into posts. Ads tells you what to kill and what to scale. Sales drafts your follow ups. Module 3 is where you pick which ones you need first.

Doing it yourself is eight to ten hours across a week. Managed is three to four hours of pre work, then one 60 minute session. Installing the bundle costs a few dollars in credits. The first reads and Foundation are where setup money goes. After that it depends on what you run on a schedule. Module 17 has real numbers.

It reads only what you connect. Anything that sends, spends or changes something outside Perplexity stops and asks you first. You decide which jobs run alone. You will see your first approval prompt in module 10 and say no to it on purpose.

One account, one Project per business or client. Nothing inside a Project leaks into another one. If you have a team, each person needs their own account. Shared logins break the connectors.

For this course, Max. Pro has no monthly credits and no Brain, and both are what the agents run on. You can start on Pro to look around, but every module from 7 on assumes Max.

Annual if you are sure, it is about $167 a month instead of $200. Monthly if you want one month to see it work. Switching later is a button in Settings, Account.

How Perplexity counts the work Computer does. 100 credits is one dollar. A quick question is a few credits. Installing the agents is a few hundred to fifteen hundred, and the first reads after that are a few thousand more. Max gives you 10,000 a month plus a one time 35,000 bonus.

Usually, yes, if you do the setup inside the first 30 days. The bonus expires about a month after you get it. The core install, Foundation Light, the mapping session and the first reads land under 35,000 for a solo business. Module 17 has the ranges by business size.

No. Upgrade the one you have, and make sure it is the account you will do the whole course in. Agents, credits, memory and Projects all live in one account. Two accounts is the most common mess I untangle.

It ties your Perplexity account to your Scaling Agents access so your bundle, your updates and your support land in the right place. It takes two minutes. Nothing in the course needs it to work, but the support does.

Then do not. Every step in this course works exactly the same. I say so on the front page and I mean it.

Not on a personal plan. Each person needs their own account, and you share a Project with them in module 6. Shared logins break the connectors, which you will meet in module 4.

Because the agents expect it. Brain is what lets a Tuesday session know what you decided on Monday. Turn it on now and by module 14 it has weeks of your work to learn from.

It will, sooner or later. Perplexity ships changes weekly. The names in the left column have stayed put, so if a switch has moved, type its name in the search box at the top of Settings.

Only long sessions. It does nothing to a short one. Keep one job per session, which this whole course drills, and the extra is small.

Three lines. Who you are, what you sell, how you want it to write to you. In module 5 you replace them with a proper block built from your history. Do not write an essay here today.

Task completion and approvals, on, for email and push. Everything else, your call. An approval you never see is an automation that waits forever. Module 15 depends on this.

Personalization is what you tell it on purpose: your custom instructions, your name, how you like answers. Memory is what it notices and saves on its own. You steer both, in different places.

Only the switches this module names. Devices, Credential vault, Connectors and Skills each get their own module. Leave them for now.

Right. Most of these are permissions, not features. You will see the effect in module 7 when Foundation remembers you, and in module 14 when you open Memory.

Yes. If you wait until you understand the tool, you will design around the tool. Design around the business first. That is why this is on paper and costs nothing.

Because right now it knows nothing about you. It would ask you six questions and write down whatever you said. In module 8 it has read your files, run Foundation and can see your real tools, so it can push back. That is worth waiting five modules for.

Then your business is built on you doing the delivery, and that is worth knowing. Start with operations and finance, where almost nothing needs to be you, and buy back some hours.

A little. Add a fourth letter, T, for what the team keeps. But do the honest version first: what does the business look like with no one in it? Then add people back for judgment, taste and energy. That is how I went from eleven people to one.

One line each. What hurts, what happens now, what you want instead. If you are writing paragraphs you are doing module 8’s work with none of module 8’s data. Save it.

Yes, and it is better. Ask each person to fill in the pillars they touch. You will find out that two people describe the same problem differently, which is a problem worth knowing about.

Leave what you want instead blank. That is what module 8 is for. The sketch is about what hurts. The fix comes later.

Only what you want an agent to know. Mindset on the sketch means the thinking and energy side of running the business: decisions you avoid, weeks you crash, who you think out loud with. Not therapy.

Safer than most tools you already use. You see exactly what it may do, you can set every write action to Always ask, and you can remove it in one click.

A connector is a login the tool recognizes. The vault holds a key for tools that do not offer a login. Either way, an agent can then reach the tool by name.

If an agent reads an inbox with client names in it, yes. One line in your onboarding, an AI reads my inbox and calendar, saves an awkward conversation later.

Three options, in order. Check if it has an API key, module 4’s cards show how. If not, Comet in module 16 can drive its website. If neither, export a CSV and put it in your Material folder.

The ones on your sketch. For almost everyone that is Google, Gmail, Calendar and Drive, then your email platform, then your calendar booking tool, then whatever your money runs through. Meta only if ads are on the sketch.

No. Reading, yes. Anything that sends, deletes, posts or spends stays on Always ask for the first month. You can loosen it after you have watched it work. You cannot un-send an email.

Nine times out of ten it is the wrong account on the card. Connectors, Connected, read the email on the card. Remove it, reconnect with the business account.

There is no cost to having them. The cost is in setup time and in a messy Connected list. Connect what the sketch needs now. Add others the week a job calls for them.

No. For native connectors you log in on the tool’s own screen and it hands Perplexity a token it can revoke. For API keys, the key is stored write only in the vault, so even you cannot read it back.

You are about the third person this month. Google wants a passkey on the account before it will finish the connection, and deleting your old passkeys does not stop the ask. Make the passkey when it prompts you, once, then run the connector again. It goes through.

Your own. Connect the account that holds the business you are installing for, not the agency or parent account above it. Client subaccounts come later, one credential each, inside that client’s Project.

Not anymore. Meta is a normal login connector now, you click Connect and approve. It used to want a key, and some older videos still say so. The API key cards below are for email platforms and CRMs, not Meta.

You make it inside the tool, usually under Settings, then API, Developers or Integrations. It goes in Perplexity under Settings, Credential vault, Add credential. If it fails with an auth error, make a fresh key and try again. A few tools, Vercel is one, will not issue a key until there is a card on the account.

If you run client groups in WhatsApp Business, yes, it is worth it. Know that the connector sees the whole account, not just the business side. So tell the agent where to look: only read these groups, and name them. Longer term, move personal chats to a personal number and keep the business number clean. Until then, you can paste one client’s thread into their Project by hand and skip the connector.

Zoom is the one people skip most, because it does not feel like a data tool. It is. Your call recordings and transcripts are the richest thing the Delivery and Sales agents read. Connect it with the account that owns the recordings, then test: list my last five Zoom recordings. One more thing: a transcript can take an hour to show up after a call ends.

Vercel does not give you a read only switch on its tokens. What you choose is reach: your whole account, one team, or one project. Pick the narrowest, the team or project the agent will publish into. Then control the writing on the Perplexity side: anything that deploys or changes a domain goes on Always ask. If you never want it publishing on its own, that setting is the answer, not the token.

Whichever one has your real work in it. ChatGPT remembers more facts but quotes you less exactly. Claude quotes better but reaches back less far. If you have both, run the kit in both and merge with the merge prompt in the kit.

Yes, if you let it. The first prompt tells it to tag every line as stated, observed or inferred, and to write UNKNOWN instead of filling a gap. Then you read it with your own eyes before you trust it. Delete anything inferred that you do not recognize.

Lite is an evening. Full is three to five hours in sittings across a week, mostly reading. Rushing it puts wrong facts into Foundation, and you will be fixing them for a month.

If the agents will write for you, emails, posts, pages, do Full. If they will mostly run operations and read numbers, Lite is enough. You can do Lite now and Full later. Nothing is wasted.

No, and here is why. You do not write the way you talk. Use them for stories, frameworks and how you explain things, which is what the Full kit’s Phase 1 and the story prompts pull from them. Written voice comes from things you typed.

Common. The AI remembers summaries of your chats, not the exact words. That is what the data export is for. Request it on day one, and when it arrives, mine it inside Perplexity for your own messages. Forty real passages beats fifteen remembered ones.

Three places. Block B goes into Settings, Personalization, Custom instructions. Block C goes into your Project instructions in module 6. Block A gets pasted into memory ten lines at a time. The Context File itself goes into your Material folder for Foundation to read.

Some of Phase 4 is. Health, family, money. The kit excludes those from Block A by default. You decide what goes into the Project. Nothing moves until you have read the audit.

No. Same move, different door. Run the kit prompts in whichever tool has your real work in it, export what you can, and put every file in the Material folder. The Project reads the folder the same no matter where the files came from. Code projects stay where they are; bring the notes and decisions, not the repo.

One per business. The agents live inside it, each in its own session. Split them across Projects and they cannot read each other’s work, which defeats the point of Foundation.

Settings follows you everywhere, so keep it short and about how you like to work. Project instructions are about this business and only apply inside it. Module 5 gives you a ready block for each.

Yes. Click the session, open its menu, rename it or move it into a Project. Names like Operations Agent or Weekly Review are what keep the Project readable a month from now.

Sources are permanent. Every session in the Project can read them. A file attached to a session is temporary and only that session sees it. Anything Foundation or the agents need goes in Sources.

Files added from a connector are one time copies. Add the file again, or ask the agent to fetch the latest version through the Drive connector.

Yes. In a Computer session, choose Convert session to project. It suggests a name and instructions and carries the files over. Useful if you started working before you read this module.

No limit. But the right number for most people is one, and for an agency one per client. More than that and you spend your time wondering where things are.

Ten thousand files or five gigabytes. Most people never get close. If you do, move raw recordings to Drive and keep the transcripts in the Project. The agents need the words, not the video.

If they will run agents, yes, as Can edit. They connect their own Google login. If they only need to read the output, Can view. Either way, turn Use session creator’s memory off so you both work from the same Brain.

Yes. One Project per client, each with its own files, connectors and instructions. The agents and skills are shared, they just read whichever Project they are standing in. Start with your next new client, not by moving everyone at once.

Yes, and most people should. Light today, Regular or Full when you are ready. Every answer carries over and it only asks what is new. Your agents pick up the new profile on their own.

Just answer it. It reads your files first, but a fact buried in a long document can get missed. Thirty seconds of typing beats a wrong profile.

The interview. It asks how you talk, what you never say, what you say too much. That is thinner than real samples, so if you have any emails at all, drop them in the folder before you start.

Because a form of forty questions gets skimmed and a conversation gets answered. It also lets it ask for clarity the moment an answer is unclear, instead of saving a guess.

On purpose. Foundation gathers. Scoring your pillars and naming the constraint is the Strategy Agent’s job, and designing the offer is the Offer Agent’s. Keeping them apart keeps the profile honest.

It is saved in your words, in your human context and the PDF only. It never goes into marketing copy or the story vault unless you say, in so many words, that it may be used and where.

For the business parts, with you nearby for the facts, and never for the voice part. The voice has to come from the person whose name goes on the writing, and the mailbox it reads must be yours.

Run Foundation once for the business. In the voice part, decide whose voice the agents write in, or run the voice part twice and name the two packs. Everything else is shared.

Open the Foundation session and say: fix my foundation. It asks what is off, saves the fix, and rebuilds the PDF. Every agent reads the new version next time it runs.

The foundation folder in your Project files, and a copy in a Drive folder if you connected Drive when it asked. Perplexity is the main store. Drive is your copy.

If you did not save it as a skill, it went into memory, and memory is not enough. Memory keeps notes. A skill is read in full before the agent writes a word. Say: save this calibrated voice into my voice skill. Next time you adjust it, say: update the existing voice skill, so you are not stacking copies.

No. The ones on your sketch. If marketing is on it and Meta is not connected, it will say not connected and use your sketch for that pillar. You can rerun the session later and it updates the map.

Good. That is the point. Look at what it found, then decide. Most people find one pillar is smaller than they thought and one is bigger. Change the order if that happens.

Yes. Open the same Project, new session, paste the prompt again. It reads the old map and asks what changed. Do it at the day 30 review in module 19, and any time your business shifts.

Sixty to ninety minutes if your connectors are in. It reads first, which takes a few minutes, then one question at a time. Do it in one sitting if you can. The order decision at the end needs the whole picture in your head.

Say so. It asks before it marks anything H, and your answer wins. The H is a suggestion based on the cost of a mistake, not a rule. Some people hand over their whole inbox. Some never will. Both are right.

For the operations and delivery pillars, yes. They know how things actually run. For the order at the end, it is your call, because it is your business and your hours.

Nothing bad. The installer checks its own record, updates what changed and leaves the rest alone.

No. The bundle is built for Perplexity Computer: Project skills, Project files, connectors and automations in one place. That is the reason this course lives here.

Do the job in a plain session with Foundation behind it. If you do it well twice, ask Computer to turn it into a skill. That is how most of these started.

You can, and you will still get generic work, because a prompt does not know your business, your voice, your method or the approval rules. An agent reads Foundation, follows a method, asks before it acts and gets better when you correct it. A prompt does none of that.

Perplexity’s skills teach Computer how to use a tool. These teach it how to run a part of your business: the method, the order, what good looks like, when to ask. They were built on real businesses over a decade and they read your Foundation first.

With the full 2.0 bundle they all install in one go, each with its own workspace. You still only wake up the core six first. An installed agent that has never run costs nothing.

Core are the six every business needs on day one: Foundation, Offer, Funnel, Email, Content and Image. Add ons are for a pillar that is on your map: Sales, Ads, Operations, Finance, Intimacy, Outreach and the rest. The directory below marks each one.

When an agent gets better, you get the new pack. Project settings, Skills, Add skill, and the new version replaces the old one. Your corrections and rules in that agent’s workspace stay.

Yes. One Project per client, the bundle installed in each, Foundation run for each. Nothing crosses between Projects. That is what agencies in the program do.

Download the zip again from the link in your welcome email and check the version in the file name. Attach it fresh in a new session and run the install again. If the same pack fails twice, send us the installer output. Module 18 has it in the fix table.

Install the bundle once, all of it, then stop looking at the directory. Open your map from module 8. The top line names the agent. That is the only one you run this week. The directory is for later, when the first one is earning its keep.

Open Connectors, then Connected, and look at the card for that tool. The email on the card is the account it reads. Remove it, connect the right one, and ask the agent to read again.

Tell it: that is in the Foundation profile, read it first. If it still asks, the profile is missing that fact. Add one line in the Foundation session and move on.

The numbers match what you see in the tool. List size, open rate, ad spend, call count. If they look strange, it is on the wrong account or the wrong date range. Ask it which.

No. One or two a day across the week is better. You will correct each one more carefully when you are not rushing, and the corrections are the point.

Say so. Tell it: there is no list yet, start from Foundation. It will build from your profile and your map, and read the real data when it exists. New businesses run this way for the first month.

No. Drafting costs a few credits and sends nothing. Sending, spending and publishing are what it asks about. A draft you did not ask for is a sign it understood the job, and you can delete it.

Either that tool’s write action is set to Allow in the connector, or the instruction did not say to ask. Open Connectors, open the tool, set every write action to Always ask. Then try the test again.

Do not. They will share a history, cost more on every message, and start answering each other’s jobs. One agent, one session, named for the job. It is the cheapest habit in this course.

Usually the third run. First read, first correction, then it is doing the job the way you would. If it is still generic on run five, Foundation is thin on voice. Add five of your emails to Material and tell Foundation to update.

No. Describing the job works too. Names are just faster, and they stop two agents loading at once.

Ask the first one to hand off. Say: hand this to the email agent. Offer writes a spec Funnel reads, Content hands to Image and Email. Module 12 is all about this.

Follow the pillar. Marketing goes to Offer, Funnel, Email, Content, Image, Ads. Sales to Sales and Proof. Delivery to Delivery and QA. Operations to Operations. Sales after the sale to Intimacy, and before the conversation to Outreach. Finance to Finance. Not sure? Ask in the Project with no agent named and it will route.

No. I built it so I could run every agent on a business and break things without a client paying for it. Maya does not exist. Eight of the thirteen scenarios, and every screenshot, are what the agents actually produced in that demo Project. The other five are how those agents work, not a run.

Most of them land between 300 and 800 credits, which is $3 to $8. The first run of any agent costs more, because it sets itself up at the same time. The Email Agent’s first job in the demo was about 1,300. The Website Agent build is the expensive one, a few thousand. Module 17 has every number.

Foundation is thin, or the agent did not read it. Ask: did you read the Foundation profile and the voice file. If no, say: read them and redo. If yes, add five of your real emails to Material and tell Foundation to update voice.

Yes. The map is the order, not a fence. But if you find yourself giving agents a lot of jobs that are not on it, go back to module 8 and update the map. The map should match what you actually do.

Two or three. The one on your top line every working day, and one or two more. Add one a week after that. Twelve agents in week one is how people burn credits and give up.

Name the skill in the ask, yes. Once you have more than a couple of skills it will not guess which one you meant, and the lighter models guess least. Write in my voice, use the voice skill. Ten extra words. The agents in the bundle do this for you, a plain session does not.

No. Once Meta is connected you do not go into the Ads Manager screen again. You do need a Business Manager with you as admin on the ad account, which takes twenty minutes if you have only ever boosted posts. Campaigns upload paused. You look, you approve, then they run.

Yes, and you should. Open it and type: what do you do and what do you need from me. It answers in a few lines. Cheaper than guessing from the name.

They can, and you should not. They share history, cost more on every message, and start answering each other’s jobs. One agent, one session. Hand off through files.

The first agent saved it to the session instead of Project files, or named it something odd. Go back to the first session and say: save the handoff for the [agent] to Project files. Then try again.

A little. Writing the handoff file is a few hundred credits at most. Far less than the second agent asking you twenty questions it could have read.

The first agent will save the handoff and tell you the agent is not installed. Install it from your bundle through Project settings, Skills, Add skill, then open it. The handoff is waiting.

Through the shared files. Anything saved to Project files is visible to every agent. Anything left in a session is not. If you want other agents to know something, make sure it was saved.

Each client Project has its own shared files and its own agents. They do not see each other. Agents that apply to every client sit at the account level and read whichever Project you open them in. That is the privacy wall.

The new pack replaces the instructions. The workspace folder, with your rules and notes, stays. You do not lose your corrections on an update.

Ask the Strategy Agent or the Operations Agent. Both are built to read the shared files and summarize. Or open Project files and read the folders yourself. Everything is there in plain text.

Leave it on High and let Computer pick. Go to Ultra for Foundation, offers and anything in your voice. Drop to Light or Standard for scheduled jobs. Open Custom only when you have a reason.

Plan, region and rollout. Models come and go monthly, and not every plan or region sees the same list on the same day. The three tiers stay the same even when the names change.

Yes. A heavy model costs about five times a light one for the same work. A long session at Ultra on Fable is the most expensive thing you can do here. A short one at Light on Sonnet is close to free.

Because most jobs do not need it and the bill is real. Triage at Ultra is the same piles as triage at Standard for five times the price. Save Ultra for judgment work: offers, Foundation, decisions, builds.

Because anything in your voice will come back sounding like a marketer, and you will spend the saved credits rewriting it. Light is for work nobody else reads.

By default, Computer does, inside the effort level you set. The agents were tested on that. You override by hand in a session, or in the automation’s settings. The one place to be deliberate is automations.

Sessions fall back to the preset. Automations do not, they sit. After a model retirement, open every automation and set the model again. Module 18 has it in the fix table.

Only when you are sitting there waiting. It costs more credits for the same answer. Never on an automation, which runs while you sleep anyway.

No. For decisions that cost more than $20 to get wrong and where you want to see models disagree. For writing, pick one model and run it at High.

No, and that is the point. It runs the models from the other companies, Claude, GPT, Gemini and the rest, and swaps them in as they get better or cheaper. You pick per task, or let Computer pick. You are never stuck with one lab’s bad month.

Memory is small facts about you, everywhere. Brain is the connected picture across sessions, everywhere, and per Project. Foundation is your full written profile, inside the Project. Project files are the raw material. Fix a wrong fact where it lives.

Yes, in that Project’s Memory settings, if you own it. You can also turn it off for your whole account under Settings, Memory. And incognito sessions never touch it.

The most common wrong memory. Delete it in Manage memories. Next time you test with made up data, open with: this is a test, do not remember any of it.

Running it on a Project with Update automatically on uses the Project owner’s credits. Update now uses yours. Tasks that use Brain usually cost less overall because Computer reads less to get its bearings.

It reviews your recent activity in the background: sessions, connected tools, files you made, corrections. Your activity only, never anyone else’s. Incognito is excluded. It does not act or message you on its own.

Perplexity keeps a log of cleared memories for up to 30 days to stop that from happening. If it did come back, something in a session restated it. Find the session, correct it there, delete the memory again.

Not today. Skills that improve themselves are a future feature on top of Brain, and anything that would change how Computer works is shown to you first. Your corrections to an agent land in that agent’s own rules, which you can read in Project files.

The Project’s Brain is shared. Personal memory is not. If Use session creator’s memory is on, each of you also draws on your own memory, so answers differ. Turn it off for one picture everyone shares.

Into Memory, through your own messages, ten lines at a time. Perplexity decides what it keeps, and not every line lands on the first pass. Check Manage memories after and paste again what is missing.

Only what you allow. Reading and reporting runs alone. Sending, spending, posting or editing stops and asks, unless you set that tool to Allow. Leave write tools on Always ask for the first month.

Every few weeks, copy the instruction, delete the automation, and create it again. Or just ask it to start a fresh thread from the next run. A short thread is the biggest cost saver in this whole system.

Open it and check two things: the schedule, and the model. If the model it was built on has been retired, the job just sits. Pick a new one.

A daily briefing from the Operations Agent, and a weekly review that reads what every agent produced. Those two make the whole system visible. Add the rest after week two, when your agents start asking for them.

Thirty minutes before you sit down. If you start at 8, run it at 7:30. It needs a few minutes, and you want it waiting for you, not the other way around.

In the automation’s own thread, and as a notification if you turned those on in module 2. Ask it to also email you the result, or post it to Slack, if that is where you live.

It can, but it needs the machine on and you there for logins and codes, so it is a poor overnight job. Keep scheduled work to things a connector can reach. Module 16 has the rule.

A daily briefing on a light model is 100 to 300 credits a run, so 2,000 to 6,000 a month. A weekly review on High is 300 to 800 a run, so about 1,200 to 3,200. At the low end both fit inside your monthly grant. At the top end they use most of it, so watch the first two weeks.

Open the automation’s thread. The approval is waiting there. Approve or decline. Then check Settings, Notifications so the next one reaches your phone.

Yes. Open the Project, right column, Automations, and pause it. It keeps its instruction and history. Unpause when you want it back.

Only if something on your map lives on your machine: local files, a notes app, bank exports. If every tool you use has a connector, skip it for now.

It is a browser, so your passwords live in it the same way they live in Chrome. What is new is the assistant can act in a logged in tab. It stops at payments and passwords and hands back to you, and you can watch everything it does.

You can have both, but install the Perplexity app first. It is the one that carries approvals and two factor prompts from your desktop.

Plain words. On my computer. In Comet. Use the connector. If you say nothing, it uses the cloud and the connectors, which is right most of the time. If it picks wrong, say: no, do that in Comet.

Comet tasks and desktop tasks need the machine on and you reachable for logins and codes, so they are not good overnight jobs. Put anything that can run through a connector on the cloud schedule. Save Comet and the desktop for jobs you kick off while you are there.

It is built to stop at payment and hand you the button. It also stops at passwords and two factor codes. If you want it to go further than that, you cannot, and that is the point.

Cloud through a connector is cheapest, because it reads data directly. Comet is dearer, because it reads screens and clicks, step by step. The desktop app sits in between. Another reason to use a connector whenever one exists.

Connector. Always. It is faster, cheaper, and it does not break when the website changes its buttons. Comet is for the gaps.

Make Comet your default browser while you work in Perplexity. Mac: System Settings, Desktop and Dock, Default web browser. Safari is the one browser I stopped using with this, it drops sessions and Comet does not.

Because the honest number depends on your map. A coach running two light automations and a weekly email is under $200. An agency running briefings for five clients is over $600. Pick your row in the table.

With auto refill on, it buys more in the step you set and the job keeps going, until you hit your monthly spending cap. With it off, running jobs pause, your work is saved, and they pick up on their own once credits are back. Turn auto refill on and set the amount low. Running out halfway through Foundation is a bad afternoon.

Yes. Pro has no monthly credits and no Brain. The 10,000 credits alone are $100 of the $167, and Brain is what makes month two cheaper than month one.

Setup. Installing the agents, Foundation, the mapping session and every first read happen once. Your 35,000 bonus exists for exactly this, and it expires after about 30 days, so do not sit on it. From month two you are only paying for running and the odd build.

A daily job left in a long thread. It cost me $50 to $70 a day. New session, same job, a few dollars. Second is an automation left on Ultra. Third is Deep Research used for things a plain question would answer.

Open the session. The right panel shows its usage. Write the number next to the job on your map in week one. Settings, Analytics shows the month by model, which is how you catch an automation set too high.

Agents that build ask for a ceiling and warn you at half. For everything else, auto refill with a small amount is your limit: it tops up in steps you chose, and you see each one. There is also a monthly spending cap on every Max account. It defaults to $200 and you can set it anywhere up to $5,000 in your credit settings. When you hit it, running jobs pause and nothing more is charged until you raise it.

No. Comet reads screens and clicks step by step. A connector reads the data directly. Same job, the connector is usually a third of the price. Use Comet only where no connector exists.

No. Turn on auto refill and let it top up as you go. Buying a big block up front just makes a wrong setting cost more before you notice it.

Probably not. Setup week looks like that. Foundation, the context load and the first reads run heavy models on long jobs, and a $200 grant can go in a day while that is happening. What is not normal is week three looking the same. Check Usage, set the spending cap, and move routine jobs to a lighter model.

The fix table above. Find the line that sounds like your problem, do the fix in order, then test in a fresh session. Nine times out of ten you are done in five minutes.

Each row says who. You means the fix is a setting or a reconnect on your side. Us means a bug in an agent’s files, and you send us the output. Nobody means the platform is having a bad hour, and you wait.

Ask the agent in its session: did you read the Foundation profile, which connectors did you use, what was the last error. A no, a wrong account or a connector error is you. The same error in a fresh session with Foundation read and the right connectors is us.

The exact error text, the session link, and one sentence on what you expected. With those three we usually ship a fixed pack the same day. Without them we spend the day asking.

Almost never. It is Foundation being thin on voice, or the agent not reading it. Ask: did you read the Foundation profile and the voice file. The fix table has the full line.

Stop. Check what went out and clean up by hand first. Then Connectors, open the tool, set every write action to Always ask, and add ask me before you send to the automation. Then tell us, so we can see how it slipped.

Settings, Analytics, look at the model split and the skills that ran. Then open your automations and check two things on each: the effort, and how old the thread is. A long thread on Ultra is almost always the answer.

For an automation, yes, that is the fix for cost and for stalls. For an agent, no. Reinstalling the same pack changes nothing. Ask it the three questions first.

Yes, Perplexity has one. If everything is slow or odd at once, check it before you touch anything. Your Project is saved. Nothing is lost by waiting an hour.

Start week one again from today, with just the top line of your rhythm. Do not try to catch up on everything at once. The month is the same length no matter when it starts.

Then do not add anything. Run the review anyway, because the cost and memory checks still matter. One thing done well every day beats ten things done badly.

You stopped doing the thing on the top line. That is the only measure. Hours back, or a job happening that never used to happen. If neither is true at day 30, the map was wrong, not you. Redo module 8 with what you know now.

Week one, twenty to forty minutes a day, mostly correcting. Week four, ten minutes to read a briefing and approve a few things. If it is taking longer than that in week four, something is set up wrong, and the fix table in module 18 is where to look.

The same review, shorter. One new line from the map if the last one is working. Strategy Agent scores the pillars. Foundation gets one update if anything changed. Put the day 60 and day 90 reviews on your calendar before you close this module.

One, if a line on your map calls for it. Run its first read and give it two weeks. Waking three at once in month two is the same mistake as waking twelve in week one.

Turn its automation off and leave the agent installed. Nothing is lost. If the line it was on still hurts, the fix is usually Foundation or the first read, not the agent. Check those before you give up on it.

At day 30 if your constraint moved, and any time your business changes shape: a new offer, a new team member, a new channel. Open the Project, new session, paste the mapping prompt from module 8. It reads the old map and asks what changed.