Back to delivery paths

Six ways to work with Claude

Every option delivers the same core capability — Claude reading, writing, researching, and reasoning for your team. What changes is how it gets started, what equipment you need, and how much we stay involved after handover.

Six options at a glance

Green — your team runs it. Amber — shared. Blue — we run it for you.

On-Demand

Your team asks Claude to do a task when they need it. Like a smart assistant on call.

On-Demand + Screen Nav

Same as A, but Claude can also click through apps and websites — no API needed.

Scheduled — Mac Mini

Claude runs tasks on a schedule, as long as your Mac Mini is on and the app is open.

Scheduled — Mac Mini (adv.)

More reliable than C, connects to more tools. One-time technical setup required.

Scheduled — Cloud

Claude runs on Anthropic's own servers. Nothing at your end needs to be on.

Event-Triggered — Cloud

Claude reacts the moment something happens — a new lead, a form submission, a Slack message.

Which option fits you?

Answer a few questions to find the right path.

Does this need to run automatically on a schedule — or does your team start it when they need it?

Does it need to open an app or website and click through it?

For tools with no direct connection to Claude — no API available

Do you have a dedicated computer that stays on 24 hours a day?

Like a Mac Mini running in the office

After setup, does your team need to update the instructions themselves — without calling us?

Does it need to react to something happening in real time?

Like a new lead, a form submission, or a Slack message

Recommended option

    Option charts

    Each axis tells you something different about the trade-offs.

    Capability vs. Ownership

    The further right — the more you depend on us. No option is both fully capable and fully self-serve.

    Automation vs. Infrastructure

    Three clusters: on-demand (top-left), Mac Mini required (bottom-right), cloud — nothing needed (top-right).

    What each option can actually do

    A
    B
    C
    D
    E
    F
    Runs on a fixed schedule — no human needed
    Reacts instantly to events — new lead, form, Slack
    Works with HubSpot, Notion, Outlook out of the box
    Works with any tool — including ones with no native connector
    Navigates apps on screen — no technical connection needed
    Keeps running when all your computers are off
    Your team modifies it directly — no need to call us
    One extra step required — a single deploy command we train your team on.
    Full breakdown — for reading in detail

    A — On-Demand

    Your team asks Claude to do a task when they need it — like a smart assistant on call. Nothing runs without a human starting it.

    What we build

    A skills library — 5–10 instruction files tuned to your tools, workflows, and brand voice. Your team runs them via slash commands; they never start from a blank prompt.

    What it can do

    Any text or data task — research, drafts, summaries, analysis. Works with HubSpot, Notion, Outlook, and other tools via native connectors.

    What it cannot do

    Anything that has to happen automatically without someone starting it.

    What you need

    A Claude subscription. That's it.

    When something goes wrong

    Your team fixes it — edit the instruction file, run the task again.

    Ongoing support

    Not needed. Your team owns the library completely.

    B — On-Demand + Screen Navigation

    Same as Option A, but Claude can also click through apps and websites directly — without those apps needing a technical connection to Claude.

    What we build

    Same as Option A — 5–10 instruction files. Some skills can navigate tools on screen, so a missing API is not a blocker.

    Best for

    Tools with no API — legacy CRMs, web portals, anything your team accesses through a browser.

    What you need

    A Claude subscription. Your computer must be on and the screen visible while the task runs.

    Ongoing support

    Not needed.

    ¹ Screen navigation launched in early 2026 and continues to mature. Recommend testing against your specific tools before relying on it for critical workflows.


    C — Scheduled, Mac Mini

    Claude runs tasks automatically on a schedule — every morning at 8am, every Monday — as long as your Mac Mini is on and the Claude app is running.

    What it can do

    Everything in Option A, on a schedule. Daily digests, weekly reports, recurring content tasks. Works with native connectors.

    What it cannot do

    React to real-time events. Guarantee tasks fire if the Mac Mini restarts overnight or the app closes.

    What you need

    A Mac Mini that never sleeps. A Claude subscription. We configure it once.

    Ongoing support

    Optional. Your team can edit instructions directly — no technical steps needed.


    D — Scheduled, Mac Mini (advanced)

    Same idea as Option C — automatic scheduling on your Mac Mini — but more reliable and connects to tools that have no native Claude connector.

    What it can do

    Everything in Option C, plus connections to tools with no native connector. Doesn't require the app to be open — only the Mac Mini to be on.

    How your team updates it

    Edit a plain-text instruction file. To apply the change, run one command we train your team on — or call us.

    What you need

    A Mac Mini that never sleeps. A Claude subscription. Git (pre-installed on Mac). One-time technical setup by us.

    Ongoing support

    Recommended. Logic changes go through us; your team handles content edits independently.


    E — Scheduled, Cloud

    Claude runs tasks on Anthropic's own servers. Nothing at your end needs to be on — not your Mac Mini, not your laptop.

    What it can do

    Full automation — daily digests, weekly reports, multi-step workflows, connections to any tool. Tasks fire, do their work, and deliver results to Slack, email, or Notion — even when everyone is asleep.

    How your team updates it

    The logic lives in plain-English skill files — anyone on your team can read exactly what the system does and why. When something needs to change, you tell us; we handle deployment.

    What you need

    A Claude subscription with a paid plan. We handle the full setup — scheduling, tool connections, instruction files.

    Ongoing support

    Required. We stay on retainer for deployment, monitoring, and new workflows.

    F — Event-Triggered, Cloud

    Claude fires the moment a specific event occurs in your tools — a new lead, a form submission, a Slack message. Same Anthropic infrastructure as Option E, but event-driven not clock-driven.

    What it can do

    Real-time lead enrichment, instant draft generation, live routing, anomaly detection the moment it appears in your data. Most powerful option we offer.

    What you need

    Same as Option E. Your existing tools need to support sending a notification when something happens — most modern tools do.

    How your team updates it

    Same as Option E — plain-English skill files your team can read; changes go through us for deployment.

    Ongoing support

    Required. Most powerful, most dependent on us.

    Working as a team? Every option supports the whole team on the receiving end — everyone gets the outputs. What changes is who can be the Owner — the person who updates instructions and adds new workflows. Options A, B, and C give any team member that role directly. Option D mostly does, with one extra step. Options E and F keep it with us — your team reads the skill files and understands exactly what the system does, but deployment goes through us.

    Ready to pick one?

    The discovery call is where we qualify this together. Free, 30 minutes — you leave with a clear recommendation and a quote.

    Book a free discovery call
    Homepage