Questions buyers ask us frequently.

Anything with an API. Most-shipped: Stripe, HubSpot, Salesforce, Slack, Gmail, Notion, Linear, Zendesk, Intercom, Webflow, Airtable. Three integration levels:

• API, webhook, MCP or equivalent: we wire it directly.
• No open API? We integrate through our own Python or JavaScript scripts.
• Locked-down tool? As a last resort, workarounds (Puppeteer, solving certain captchas, IP rotation...), but then you have to weigh the risk to long-term reliability.

Claude Routines is a feature. We ship a system. Five concrete differences:

• Done-for-you vs DIY: Routines, you operate it. We build it, we monitor it, we fix it. You use it.
• Trigger: Routines runs on cron. Our workflows also react to a webhook, an inbound email, a new row in Stripe, a Slack message...
• Model: Routines means Claude, full stop. With us, the model is agnostic. We swap to OpenAI, Mistral, or a private model in an afternoon.
• Infra: Routines is hosted by Anthropic. With us, you pick: public, private cloud, or on-premise.
• Support: when an automation breaks, a human steps in. Claude doesn't answer the phone.

The two can live together. If Routines is enough, keep it. If you need cross-tool orchestration, that is us.

For sure you could! AI gets you most of the way on almost anything. The internet is full of tutorials on every subject: fixing a leaky toilet, building your own fully secured SaaS in just one hour... depending on where you start, it all stays doable. Having done it more than once ourselves, and having picked up projects clients already started, the path is usually less linear than you imagine and paved with surprises where you least expect them (hello API rate limits, error handling and retry logic, the API that returns a cheerful 200 with an empty payload...). That gap, between the template you found online and a system you can trust, is exactly the thing. That’s also how you learn, you’ll tell us? Let’s say we spare you the learning curve.

Yes, and it's governed, not just promised. What we put in place:

• Hosting in Switzerland or the EU, never under US jurisdiction.
• Encryption, audit logs and access control on every deployment.
• Private AI as an option: the model runs on infra you control, and no client byte leaves your perimeter.
• Governed tool access: per-tool permissions, an allowlist, call rate limits, OAuth scoped to the strict minimum, every call logged.
• A human in the loop at the steps you set. An agent only touches what you allow.

Your data stays yours, and everything is traceable. In writing, not on trust.

It depends on the deployment type and the model you pick:

• On public: it comes down to the model you pick and the provider’s terms of use at any given moment.
• On private: the model runs on infra you control. The question stops being a question.

We help you see this choice clearly, and we settle the direction together. Your data leaves your perimeter only if you sign off, and we log it when it does.

You’re covered on two levels: the resilience built into the automation, and our support when it isn’t enough.

• Built into the automation: automatic error handling and fallbacks (retry, backup model, alternate path). Most incidents resolve without intervention.
• SLA: support handled per the terms of your contract.
• Included: automated monitoring, immediate alerts on failure, fixes for errors tied to the automation platform.
• Excluded: new feature requests, logic changes, after-hours support.
• Limits: we don’t cover outages caused by your third-party API accounts being suspended (e.g. OpenAI quota exceeded) or third-party maintenance windows (Airtable, Gmail...).

Yes. You define:

• Permissions (what each agent is allowed to read, write, execute).
• Approval steps (where a human must sign off before the agent continues).
• Source boundaries (which tools, which databases, which accounts the agent can touch).
• Escalation rules (to whom, in which case, at what urgency).

Nothing runs without the controls you set.

It asks clarifying questions or escalates to a human, based on your rules. There is always a human in the loop. Agents don’t guess or act blindly.

Sometimes yes, and always upfront. Two possible cost families:

• Third-party tools (APIs, connectors, external SaaS): we flag them ahead of time, and we lean as much as possible on tools that are free or already included in what you use today.
• AI features: some models or token volumes can generate extra costs.

Everything is laid out ahead of time, never as a surprise on the invoice.

We planned for it. Three reasons:

• Every workflow is model-agnostic by design.
• We swap the model behind your prompts in an afternoon.
• Most clients run a primary model and a fallback by month two.

You’re paying for time you don’t get back. The eight automations run 24/7: invoices filed, follow-ups sent, quotes and contracts generated, social posts published. If that frees up a few hours a week, you’re already ahead, never mind the missed follow-ups that were costing you far more. And it all stays maintained: when a tool changes its rules, that’s our problem, not yours.

You do. They’re your tools, your accounts, your subscriptions, exactly as today. Our job is to make them work together automatically. You don’t pay twice: your subscription with us covers the orchestration and the maintenance, not the licenses.

The Starter Pack includes one custom field per automation, to fit the way you work. Beyond that, we talk: if your needs outgrow the pack, it’s time to move to a custom plan.