People ask what it costs before they ask what it does, so here is the whole bill: the one-time setup, what runs every month afterwards, and how it compares with the three things founders actually weigh it against — a support hire, a part-time VA, and a help-desk subscription.
| Package | Price | Who | What you get |
|---|---|---|---|
| Starter | $1,000–$3,000 one-time | Freelancers, solo operators, small business with one inbox | One persona’s context layer, one inbox, one agent (Hermes or a Grok bot) with Claude Code or ChatGPT behind it, one flow — triage + tally + urgent flags with drafts, or review replies, or follow-ups. Persona and memory files, security hardening, runbook, 2 weeks of tuning. Live in 2–3 days. Price rises with personas and flows. |
| Growth | $10,000–$15,000 one-time | Startups with a product, a team, a user database, and a repo | The company context layer — personas for CEO, CTO, engineers, sales, HR/payroll and hiring; inboxes, a sanitised read-only database view, indexed transcripts and the repo as sources — plus inbox + store reviews → research tally → Claude Code or ChatGPT drafts and builds → “you asked, it’s live”. Sales outreach tuning and the X/Reddit/Instagram/TikTok content radar. Written security review, per-persona training, 30 days of tuning. Live in 2 weeks. Price rises with personas, sources and flows. |
There is no subscription and no per-seat fee. The agents are open, self-hosted tools you own; we are paid to install, harden, tune and hand them over. Full package details are on the AI Employee Setup page.
Want a quote for your inbox volume?
Tell us roughly how many emails and reviews you get a month and we will tell you which package and what it costs — same day.
A full-time support person in the US costs somewhere around $45,000–$65,000 a year fully loaded, more in expensive cities, and takes weeks to hire and train. A Growth setup is a quarter of one year of that, once. But be precise about what it replaces: the reading, the counting, the first draft, the follow-through on “it shipped.” It does not replace the person who handles an angry customer or a lost-data ticket — at Pocket Clear that is still the founder, and should be.
A part-time virtual assistant runs roughly $1,000–$2,000 a month and needs the same runbook the agent does, plus management. The VA is better at judgement calls; the agent is better at never missing a message, never forgetting a follow-up, and counting. Many small businesses end up with a Starter setup doing the reading and a human — often the owner — doing the 20% of replies that need a person.
Intercom, Zendesk and their peers are $100–$300+ a month for a small team and rising with seats, and they organise the queue rather than read it. Nothing in them tells you that “privacy-first” is 3% of what your users talk about. The tally does — that is the whole point of the pipeline. Some teams keep the help desk as the mailbox and point the agent at it; that works fine.
Because the work is repeatable. We have done this setup enough times — on our own products and for other founders — that the scope is known: install, connect, persona, memory, flows, hardening, runbook, tuning window. Hourly billing would make you pay for our learning curve; a package makes us pay for it. It also keeps the Starter honest: at $1,000 it has to be a checklist we can run in days, which is exactly what a small business should be buying first.
What the context layer actually contains is in The context problem, the stack in Hermes or Grok with Claude Code or ChatGPT, and what the hardening actually involves in Securing an AI employee on your inbox.
A Starter setup — one persona’s context layer, one inbox, one agent, one flow, hardened, with a runbook — is $1,000–$3,000 one-time depending on personas and flows. Ongoing costs are your own model API usage (tens of dollars a month at small-business volume) and a machine that stays on.
The Growth package — the company context layer for up to six personas, a sanitised read-only database view, indexed transcripts, inbox and reviews read into a research tally, Claude Code or ChatGPT drafting and building, sales outreach tuning, the content radar, a written security review and 30 days of tuning — is $10,000–$15,000 one-time, live in about two weeks.
No. The agents are open, self-hosted tools you own. You pay a one-time setup fee to HyperNest Labs and your model provider directly for usage. There are no per-seat fees and no retainer.