On this page

Account & AI Modes

In short

TensorPM separates two decisions that are often confused. First: where your project data lives and who may see it. Second: which AI service is allowed to work for you.

You set both independently. You can keep a project entirely local and still use AI. You can also run TensorPM without any AI at all, in which case it stays a structured project management tool.

This article covers both sides: the account with its three plans and its credits, and the four ways to connect AI.

When you need this

  • You are just starting and want to know whether you need an account at all.
  • You are wondering why the AI suddenly stopped answering.
  • You want to know what a request costs and where to check it.
  • Your company does not allow cloud AI, and you need a path that runs on your own infrastructure.
  • You received a Pro voucher code and want to redeem it.
  • You are handing back a laptop and want your account cleanly removed from it.

Creating an account and signing in

The account lives at the bottom left of the sidebar. The entry reads Account when you are signed in and Login when you are not. Clicking it opens the account popover. This is the central place for subscription, credits, voucher, passkeys, and workspaces. A workspace is the container your projects live in.

The account popover at the bottom left shows subscription status and credits and leads to workspaces, voucher, passkeys, and log out.
The account popover at the bottom left shows subscription status and credits and leads to workspaces, voucher, passkeys, and log out.

Note that none of this lives in Settings. You open Settings through the gear icon right next to the account entry. Settings is where you configure AI, not the account.

Creating an account

  1. Click Login at the bottom left.
  2. Switch to Create Account in the popover.
  3. Enter your email address and pick a sign-in method: Password or Magic Link. With a magic link you receive an email containing a link and need no password at all.
  4. Confirm your email address.
  5. You know it worked when the popover shows your name with the subscription status below it.

After signing up you are on the Trial plan and have credits for TensorPM AI right away. No credit card is required.

You can also use TensorPM without an account. Local workspaces and local projects work while signed out. Cloud Sync itself is part of every plan including Trial, but it does need an account, and so do TensorPM AI and workspace invitations.

Signing in on a second device

  1. Install TensorPM on the second device.
  2. Click Login at the bottom left and choose TensorPM Login.
  3. Sign in with password, magic link, or passkey.
  4. Check that the popover shows subscription status and credits.

Credits belong to the account, not to the device. Both devices draw from the same pool. Project data is different: it always lives locally first and, in a cloud workspace, syncs through TensorPM's servers on every plan. See Sync & Cloud for the details.

Passkeys

A passkey is a sign-in without a password. Instead of typing a secret you confirm with fingerprint, face recognition, or your device PIN.

  1. Open the account popover and choose Manage Passkeys.
  2. TensorPM opens your system browser on tensorpm.com. Creating, renaming, and removing passkeys happens there, not inside the app.
  3. Sign in on the website if you are not signed in there yet.
  4. Create the passkey. From then on you can use Sign in with passkey in the app.

A passkey requires an existing account. Create the account with password or magic link first, then add the passkey. If you lose the device, password and magic link still get you in.

Logging out on one device

  1. Open the account popover.
  2. Choose Log out at the bottom.

This disconnects only this device from your account. Your subscription, your credits, and your cloud workspaces stay intact. Local projects remain on the device and still open. TensorPM AI stops working after logging out, because it is tied to the account.

Important: Logging out alone is not enough when you hand a device over. Also remove any stored provider keys and check whether local project folders hold data that should not stay on the device. See Uninstall & Data Removal.

The three plans

There are exactly three plans: Trial, Pro, and Business.

What each plan includes

Trial Pro Business
Price free 99 euro per month or 990 euro per year individual, on request
Time limit none while the subscription runs individual
AI credits 2,000,000 one time 10,000,000 per month individual, as a shared pool
Cloud Sync yes yes yes
Own cloud workspaces several (fair use) unlimited unlimited
Members unlimited, free unlimited, free unlimited, free
Active projects 1 active project, archived unlimited 1 active project, archived unlimited individual
Own provider keys no no yes, the only plan with it

The 2,000,000 Trial credits are a one-time grant and they do not expire. Trial itself never runs out of time. You spend the credits at your own pace.

The 10,000,000 Pro credits are a monthly budget. They refill at the end of each billing period. Unused credits do not accumulate.

On Business the organization shares one pool. The account popover and the AI tab in Settings then show a Team pool marker. Every member draws from the same counter.

Important: There are no credit top-ups and no automatic top-up any more. If you need more, you change plan or move to one of the other AI paths described below.

Which plan fits whom

  • Trial fits when you want to try TensorPM properly on a real project. Cloud Sync and free colleagues are already included. It also works as a permanent setup, as long as 2,000,000 credits and a handful of cloud workspaces of your own are enough.
  • Pro fits when you work with the project agent daily and need a bigger monthly credit budget, or an unlimited number of cloud workspaces of your own. Monthly if you want flexibility, yearly if the project runs longer.
  • Business fits when an organization is behind it: several projects at once, a shared AI budget, your own contracts with AI providers, or a policy that AI requests must run through your own provider accounts.

Members are free and unlimited on every plan, and joining somebody else's workspace costs nothing. AI requests made inside a workspace are charged to that workspace owner's credit budget. See Workspaces & Teams.

Current prices and commercial details are on https://tensorpm.com/pricing. What the app shows for your account is always authoritative.

Understanding credits

What a credit is

A credit is the unit of measure for AI usage through TensorPM. Think of it as a meter: every AI request turns the meter a bit further. How far depends on how much text the AI had to read and how much it wrote.

Two things matter:

  • Credits are only spent on requests that run through TensorPM. If you use your own Claude or ChatGPT subscription, your own server, or your own provider keys, not a single credit is consumed. The provider bills you directly instead.
  • Reading is cheap, writing is expensive. A long analysis with a short answer costs less than a short question that produces a multi-page document.

The exact formula

For those who want the precise rule: AI breaks text into small pieces called tokens. A token is roughly a word fragment, about four characters. Billing counts three kinds separately:

credits = input tokens x 1
        + cached input x 0.2
        + output tokens x 4
  • Input tokens are everything the AI reads: your question plus the project context TensorPM sends along.
  • Cached input is the part the provider still holds from an earlier request. It costs only a fifth. That is exactly why continuing a chat is cheaper than sending the same context five times in five fresh chats.
  • Output tokens are what the AI writes. They cost four times as much, because the model produces them one piece at a time.

You can see the same formula in the app: the AI tab in Settings shows an info icon next to AI Credits that explains it.

A worked example

A typical question inside a running chat, for example "what is blocking the critical path right now?":

Part Tokens Factor Credits
new input (question plus fresh context) 4,000 1 4,000
cached input (context from the running chat) 8,000 0.2 1,600
output (the answer) 900 4 3,600
Total 9,200

For scale: the 2,000,000 Trial credits cover roughly 200 such requests. The 10,000,000 credits in the Pro monthly budget cover roughly 1,000 per month. Short follow-up questions cost far less, a generated document or a large project analysis far more.

Where you can see your usage

There are three places, and they show different levels of detail:

  1. The account popover at the bottom left. On Trial the status line reads {{amount}} credits left or Exhausted. On a paid plan it shows the subscription status and the next billing date. The small refresh button next to it pulls a fresh state from the server.
  2. Settings -> AI. With TensorPM selected as preferred AI provider and while signed in, an AI Credits block appears showing the amount used, the total budget, and the date the budget resets. On Business it also shows Team pool.
  3. AI Activity History. Open it through the activity icon in the top bar. It lists every AI action of the project with status and duration. The token usage panel breaks the consumption down by category: chat, project creation, action items and project structure, guidance and project analyses, documents and files, automatic project support, AI agents. This is the only place where you can see which area is eating your budget.

How to reduce usage

  • Start with AI Activity History. Usually more than half of the cost sits in a single category.
  • Turn off automatic features you do not need. Settings -> General -> AI holds Automatic AI Prioritization and Daily Project Status Review. Both run in the background and spend credits.
  • Review proactive project check-ins. Each scheduled check-in makes one AI request per active project. A lower frequency cuts usage immediately.
  • Stay in the running chat instead of opening a new one for every follow-up. Cached context costs only a fifth.
  • Attach only the documents that matter for the question. Every attached page is input.
  • Ask precisely. "Summarize the three biggest risks in two sentences each" costs a fraction of "analyze the project".
  • Pick a smaller model for simple work. The model selector sits in the top bar and under Settings -> AI -> Preferred Model.
  • Move steady load onto one of the other AI paths. Your own Claude or ChatGPT subscription spends no credits at all.

The four ways to use AI

TensorPM can reach AI in four ways. All four are configured under Settings -> AI. The active one is whatever is selected under Preferred AI Provider.

The AI tab in Settings with the groups AI Provider Settings, Subscriptions & Agents, AI Keys, and Web Search.
The AI tab in Settings with the groups AI Provider Settings, Subscriptions & Agents, AI Keys, and Web Search.

Path 1: TensorPM AI through your account

Requirement: a TensorPM account, signed in. Nothing else.

Cost: spends credits according to the formula above. No separate bill, no provider accounts.

Privacy consequence: the request travels through TensorPM's servers to the model provider behind them. Project content you send leaves your device. That is the trade-off for the simplest path. Before using it with sensitive projects, read the privacy information on https://tensorpm.com/privacy.

Setup:

  1. Sign in at the bottom left.
  2. Open Settings -> AI.
  3. Select TensorPM under Preferred AI Provider.
  4. Pick a model under Preferred Model.
  5. You know it worked when the AI Credits block appears below and the chat answers.

Path 2: A Claude or ChatGPT subscription through the local runtime

Requirement: an eligible Claude subscription (Pro or Max) or an eligible ChatGPT subscription. On top of that, the provider's official local application must be installed on the machine: Claude Code for Claude, Codex for ChatGPT. If it is missing, TensorPM shows the status Not installed.

Cost: no TensorPM credits are spent. You only pay your existing subscription with the provider.

Privacy consequence: the credentials stay inside the provider's local application, TensorPM never sees them. Project content still goes to the provider, because that is where the model runs. Their terms apply, not TensorPM's.

Setup:

  1. Install the official local application if it is missing.
  2. Open Settings -> AI -> Subscriptions & Agents.
  3. In the Claude and ChatGPT subscriptions section, click Sign in next to the provider you want.
  4. Complete the sign-in in the browser window that opens. It is the provider's official sign-in page.
  5. Return to TensorPM. The status switches to Connected.
  6. Select the provider under Preferred AI Provider.

If the connection expires, the button reads Reconnect. Which models are available is decided by your subscription and can change over time.

Path 3: Local AI through a compatible server

Requirement: a running model server you can reach. Anything speaking the common OpenAI-compatible interface works: Ollama, LM Studio, vLLM, or an internal gateway run by your IT. TensorPM recommends models with at least a 128,000-token context window and solid tool-calling ability, otherwise the project agent cannot use its tools properly.

Cost: no credits are spent. You pay with your own hardware or server budget.

Privacy consequence: this is the path with the smallest data outflow, provided the server really sits with you. Be careful though: local does not automatically mean private. A gateway that forwards requests to a cloud provider looks exactly like a machine under your desk from inside TensorPM. Clarify that first.

Setup:

  1. Open Settings -> AI and expand Local AI Server.
  2. Click Auto-Detect Servers or type the Endpoint URL yourself.
  3. Add an API key (optional) if your endpoint is protected. It is stored through the operating system's secure storage.
  4. Click Test Connection. On success TensorPM reports how many models it found.
  5. Select a detected model or type the Model Name manually.
  6. Switch the preferred provider to the local server and ask a harmless question first, before you allow write actions.

Path 4: Your own provider keys on Business

Requirement: a Business account. Own provider keys are the one capability reserved for this plan alone. Supported providers are OpenAI, Anthropic, Google, and Mistral.

Cost: no TensorPM credits are spent. The provider bills the account that owns the key directly, so the cost risk sits with your organization.

Privacy consequence: the request goes straight from the device to the provider, without passing through TensorPM. The contract between your organization and the provider applies, including their rules on logging and data retention.

Setup:

  1. Open Settings -> AI -> AI Keys.
  2. Enter the key in the field of the matching provider. Keys stay local on the device, are stored encrypted there, and are never transmitted to TensorPM servers.
  3. Select the provider under Preferred AI Provider and a model below it.
  4. Optional: set Requests/min and Tokens/min under Rate Limits so TensorPM does not overrun your provider quota. Detect reads those values from the provider's responses.

Important: never paste a key into chat. It belongs only in the field in Settings. Everything written in chat goes to the model.

Independently of the plan, the AI tab also holds the Web Search group. Enter a Brave Search API key there if the project agent should be able to search the web. As soon as a valid key is stored, search is used automatically.

Comparing the four paths

Path Requirement Credits Data goes to
TensorPM AI account, signed in yes TensorPM and its model provider
Claude or ChatGPT subscription subscription plus local application no the subscription provider
Local AI your own compatible server no your server, nobody else
Own provider keys Business plan no the provider directly

You can switch at any time. The active path applies to every project on this device.

Redeeming a Pro voucher

A Pro voucher unlocks Pro for a fixed period, without Stripe and without a credit card. The entry only appears on Trial accounts.

  1. Open the account popover at the bottom left.
  2. Choose Redeem voucher.
  3. Enter the code in the Pro voucher code field.
  4. Click Redeem.
  5. On success the app reports the date Pro is active until. The subscription status in the popover changes.

Possible messages and what they mean:

Message Meaning
This voucher code doesn't exist. A typo or the wrong code. Check your input.
This voucher has reached its redemption limit. The code has been used its maximum number of times.
This voucher has been disabled. The code was withdrawn. Ask whoever issued it.
This voucher has expired. The code is past its validity.
You can't redeem a voucher while a subscription is active. You already have a running Stripe subscription.
You already redeemed this voucher. Not an error, the period was already credited once.

If your Pro already comes from a voucher, another voucher can extend it.

When credits run out

The account popover then reads Exhausted, and the chat reports that your AI credits are used up.

What still works:

  • All project data stays intact and open.
  • You can keep working: action items, dates, budget, files, trail, everything without AI.
  • Cloud Sync keeps running as long as your subscription is active.

What stops working: everything that triggers an AI request through TensorPM. That means chat, analyses, guidance, distillation, automatic prioritization, and project check-ins.

Your options:

  • Trial: upgrade with Upgrade to Pro in the account popover, redeem a Pro voucher, or move to one of the other three AI paths.
  • Pro: wait for the monthly budget to reset. The date is shown under Settings -> AI in the AI Credits block. A Claude or ChatGPT subscription or a local server bridges the gap.
  • Business: the pool is shared, so it may have been spent elsewhere in the organization. Talk to whoever manages the contract. Business can additionally fall back to its own provider keys.

Topping up individual credits is no longer possible, and there is no automatic top-up either.

Why the model list changes

TensorPM only lists models that are actually usable for your account, your chosen path, and this device. That is why the list changes:

  • With a subscription, models are read from the connected local application.
  • With Local AI, they come from your server.
  • With TensorPM AI and your own keys, the catalog changes with new app versions.

The selector in the app is therefore always the current truth. If it shows only AI with no model, no usable path is connected yet.

Security checklist

  • Use only official sign-in windows and official provider applications.
  • Never type passwords or provider keys into chat.
  • Connect only providers approved for this project's data.
  • With local servers, check whether the server itself forwards data.
  • Test a new setup on a low-risk local project first.
  • Read proposed write actions and generated files before approving them.
  • Log out and remove stored keys when a device changes hands.

Frequently asked questions

Do I need an account to use TensorPM? No. Local workspaces and local projects run while signed out. Cloud Sync is included on every plan including Trial, but it does require an account, as does TensorPM AI.

Does Trial expire after 14 or 30 days? No. Trial has no time limit. It ends only when the 2,000,000 credits are spent, and even then all project data stays usable.

Does a Claude subscription still spend credits? No. Only requests through TensorPM cost credits. Subscription, local server, and your own keys all bypass the credit counter entirely.

Can I run several projects in parallel on Pro? Pro allows one active project. Archived projects are kept without limit and can be reactivated. Anyone who permanently needs several in parallel needs Business.

What happens to my data if I cancel? The projects are local and stay local. What happens to cloud workspaces is covered in Workspaces & Teams.

Can I buy extra credits? No. Credit top-ups and automatic top-up were removed. What remains is changing plan or using another AI path.

Does a failed AI attempt cost credits too? Yes, once the model has read the request and started answering. Cancelling after a few seconds is therefore cheaper than a full answer, but not free.

Where do I see which area consumes my budget? In AI Activity History, through the token usage panel broken down by category and model.

Can I use different paths for different projects? No. The preferred AI provider applies device-wide to every project. You can switch at any time.

When something does not work

The chat does not answer and reports that AI is unavailable. TensorPM names the right next step inside the message itself. It is usually one of four cases: not signed in, credits exhausted, a missing Business key, or a half-configured Local AI. The button in the message takes you straight to the right place.

The subscription status in the popover is stale. Click the refresh button on the right of the status line. It pulls subscription and credits from the server again. Without an internet connection the last known state stays on screen.

Sign in for the Claude or ChatGPT subscription does nothing. First check whether it says Not installed. Without the provider's official local application TensorPM cannot start the sign-in. If the app reports that sign-in took too long or that no subscription was detected, start it again and finish the browser window completely this time.

The connection to the local server fails. Check the endpoint URL, whether the server is running, and whether a firewall sits in between. If TensorPM finds the server but no models, type the model name manually.

The voucher code is rejected. Look up the message in the table in the voucher section. The most common case is an active Stripe subscription blocking redemption.

A provider key is not accepted. Own keys only work on the Business plan. Also check that the key belongs to the right provider and that your operating system offers secure storage. On Linux, GNOME Keyring or KDE Wallet must be installed and unlocked.

The model selector shows only AI. No usable model is connected yet. Walk through the setup of your chosen path once more.

More symptoms and fixes are collected in Troubleshooting.

Next steps

Previous
Workspaces & teams
Next
Encryption & security
Was this page helpful?