Start free.
Scale when ready.

A credit is a tenth of a dollar of model usage, and we pass model costs through at what they cost us. A full agent run — reading a set of sources, reasoning over them, writing something up — usually lands between 4 and 20 credits depending on how much there is to read and how many steps it takes. So Solo's 80 credits is about ten runs a month and Pro's 450 is about a hundred, which is three Programs working every day. The app shows what each run actually cost, so you never have to guess.

A daily scan of procurement portals runs about 220 credits a month; an overnight inbox review about 140; asking your books questions and closing the month about 40. So Pro's 450 covers all three with room left over, and Solo's 80 covers one Program run a couple of times a week, or a month of asking a room questions. If you outgrow a tier you can top up or point that work at your own API key — you are never stopped mid-run.

No, on either paid tier. Credits reset at the start of each billing period and unused ones do not carry forward. We would rather say that here than have you find out at the end of a quiet month.

You are warned before it happens. When they are gone you can top up 40 credits for $19, or add your own Anthropic, OpenAI or Google key and keep working at provider rates. A run already in progress is never cut off mid-way.

Free, with the $5 of starter credits — it costs nothing and it is enough to watch one Program finish something real. Take Solo when you want that Program on a schedule without setting up an API key. Move to Pro when you have two or three Programs running every day and topping up has started to feel like a chore.

You paste in an API key from Anthropic, OpenAI or Google, and Pencel talks to them directly — you pay the provider, we take nothing. It is how the Free plan works once your starter credits are spent, and it stays available on Pro for a model we do not manage or for volume beyond your included credits.

The paid third-party services your agents reach for during a run — AI web search, page fetching, and the procurement data sources the RFP Program uses — are covered by your subscription on Solo and Pro. There is a fair-use ceiling on search volume; you will only meet it by running something unusual.

Yes, and without a card. Verify your email and you get $5 of managed credits — enough for roughly half a dozen real agent runs on your own data. After that the Free plan keeps working with your own API key, indefinitely.

No. Pencel runs on your machine, so scheduled work happens while the app is open and the computer is awake; anything it missed is caught up the next time you open it. There is no cloud execution today — we would rather be straight about that than sell you an uptime promise we cannot keep.

Pencel is a single-user desktop app today. Shared workspaces, pooled approvals and SSO are what we are building with our first team customers, so if that is you, talk to us — you will get pilot pricing and a say in how it works, not a waiting list.

Yes. No contracts and no cancellation fees. You keep access through the end of the period you have paid for, then move to the Free plan, where your workspace and every document you made stay on your machine exactly as they were. Annual plans can be cancelled for a prorated refund.

Your workspace — rooms, documents, knowledge, memory and run history — is stored on your own machine on every plan. When you use managed credits the prompt for that request is routed through Pencel's proxy to the model provider over an encrypted connection, and we do not log prompt content. Prefer nothing in between? Use your own key and prompts go straight from your machine to the provider.

Not today. Pencel is a desktop app for macOS, Windows and Linux. Mobile is on the roadmap and is not part of this release.