Frequently asked questions

Cryogram turns a process you describe into a working automation. You explain what you want in a chat, and the AI does the job once for real, in front of you. The routine parts of what it did are turned into ordinary software, which gives the same output for the same input every time, and AI is kept only in the steps that genuinely need judgement. The result is a button anyone on your team can press: the same result every time, in seconds, with every run checked and logged.

A saved prompt is still interpreted by an AI model on every run, so every run takes an AI's thinking time, costs AI usage, and can come out differently from one day to the next. Cryogram extracts the routine steps into actual software, tested against the recorded results of runs that worked. The steps run identically every time, and a value only moves from one step to the next after passing that step's checks.

File pipelines, document intake and extraction, email digests, spreadsheet-to-report flows, CRM updates, web monitoring, outreach merges - most repeatable business tasks that mix fetching, transforming and a few judgement calls. The examples section on the front page shows what early users have built, from CV shortlisting to monthly report packs.

No. A colleague presses Run and watches the progress in plain language. When the automation needs a human decision, it asks - for example, to pick from a list it just fetched, or to approve an email before it is sent - and when it finishes, the results are there to download, with a full history of previous runs.

macOS today. Windows is committed for launch.

The automation stops safely mid-run rather than guessing, and no value moves past a failed check. You get a plain-language account of what stopped and why. The AI then investigates what changed and proposes a repair, and the fix goes live only after passing the automation's tests against the recorded results of past good runs. Anything that touches an external system asks you before acting.

No, and you should be wary of anyone who does. Our claim is that automations are dependable and honest: every step is tested, every run is checked, and when the world changes - a website moves, a file format shifts - the automation stops cleanly and tells you, instead of producing something plausible and wrong. Then it gets repaired, and the repair is tested before it ships.

Anything external - sending an email, writing to a CRM - sits behind an approval gate. For each automation, you choose whether those steps need a human confirmation on every run or are approved to run unattended. The security page describes the full architecture.

Building and running automations on your own machine is free: the full builder, local execution, local scheduling, and everything your team sees when they run an automation, all powered by your own AI key. Cryogram Cloud is the optional paid service: it takes the automations you choose to publish and runs them in the cloud, on schedules and triggers, without your machine being on. Details and pricing are announced at launch.

There is no catch, and the reason is structural: local automations run on your machine and on your AI key, so they cost us almost nothing to support. We charge for the one thing your machine cannot do, which is running your automations around the clock with your laptop closed. Until you want that, there is nothing to pay for.

Yes, an Anthropic API key or subscription. Building and repair sessions run on it, which is part of why the local app can be free and fully private: nothing of yours passes through us. Finished automations barely use the key, because only the steps that genuinely need judgement call a model at all, and those can use small, inexpensive models.

No. Cryogram Cloud is optional and per automation: you choose which automations, if any, to publish. Everything you build and run on your own machine stays there, fully private, and keeps working whether or not you ever use the cloud. Cryogram Cloud exists for the moment you want an automation to run on a schedule, be triggered by another tool, or be used by colleagues who do not have an AI key of their own.

No. Keeping published automations healthy is included in the service. Charging a separate repair fee would amount to a penalty for the world changing, and that is not something we are willing to charge for.

On your machine and in your own systems. Automations process your documents and records at run time and store none of them, and the material from building - your chat, sample files, recorded test evidence - never leaves your machine, even if you later publish to Cryogram Cloud. For published automations, customer data at rest can be hosted in your choice of region, including the EU, encrypted, with documented retention windows. The full picture is on the security page.

No. The AI sees secret names only. The values are stored locally with restrictive permissions and injected at the moment a step runs, then scrubbed from outputs and logs. This is how the product is built, not a policy promise: there is no code path that hands a secret value to a model.

We are opening access in batches, so that everyone who joins gets properly set up and supported. Join the waitlist and we will email you when your access is ready. Telling us what you would like to automate is optional, and helps us send you the right getting-started guidance.

That is a very common starting point: one person who already uses AI well, colleagues who just want the work done, and a stack of recurring processes. That person can build the automations, and the team gets the buttons. But Cryogram works just as well for an individual automating their own recurring tasks, and nothing about it stops larger teams - it is useful anywhere repeatable work meets a few judgement calls.