I searched 'marketing automation freelance' looking for someone to hire. Is Clone a freelancer?+
No. Clone is a tool you run on your Mac. The first page of Google for this keyword is mostly marketplaces (Upwork, Toptal, Fiverr, Truelancer) selling you a human at $40 to $90/hr whose job is to wire up Zapier, HubSpot, or ActiveCampaign for your business. This guide is written for the other reading of the same keyword: the freelance marketer who already does marketing for clients and wants to stop hand-running their own inbound, their own LinkedIn, and their own pipeline review every week. If you are looking for a body to hire, the SERP marketplaces are the right answer. If you are the freelance marketer, Clone replaces the three weekly chores that never get done.
What is memory/rituals/own-growth/ exactly?+
A folder on your Mac with three markdown files, each about 23 lines. inbound-reply.md triggers on a Gmail label and drafts a reply under the inbound thread, using your voice.md and pricing.md plus the prospect's site and LinkedIn. weekly-linkedin.md runs Wednesday 07:00, scans your client deliverables folder, and drafts a LinkedIn post anchored in actual work you did that week. pipeline-review.md runs Friday 16:00 and writes a 5-bullet summary into a Notion page. The folder is not a SaaS, not a dashboard, not a workspace. It is three text files under your user directory.
How is this different from Apollo, Lemlist, or Instantly?+
Those are outbound cadence tools. They send cold email sequences to prospect lists you imported. They have nothing to do with inbound replies. inbound-reply.md only runs on threads where the prospect emailed you first. It never sends cold. It never builds a list. It reads one email at a time, the one you received, and drafts one reply under the thread. The difference between an outbound sequencer and this ritual is roughly the difference between a megaphone and a handwritten note on your own letterhead.
How is this different from Flowrite, Lavender, or Gmail Smart Reply?+
Those are reply-suggestion sidebars. They propose three canned phrases based on the email content. inbound-reply.md reads four sources before drafting one line: the prospect's website, their LinkedIn, your voice.md, and your pricing.md. The architecture layer described in src/components/architecture.tsx lines 18-22 (Computer Agent: reads the screen, clicks, types, scrolls) opens the prospect's site in the same Chrome profile you use personally, reads the /about page and the top nav, then drafts. A suggestion sidebar never leaves Gmail. This ritual leaves Gmail, reads the prospect, comes back, and writes under the thread.
How does Clone know my voice without training a model on my data?+
It reads memory/voice.md at the top of every draft. The file is plain text of your last 10 or so sent emails, pasted in. Architecture principle in src/components/architecture.tsx lines 51-55 is verbatim: Clone observes how you draft emails, formats proposals, and closes engagements, then mirrors that style. It is your working habits scaled, not a generic template library. There is no fine-tuning step, no model upload, no embedding training job. You paste text into a file on your disk. Every run reads the file. When you update the file, the next run uses the update.
Does anything actually send without my approval?+
No. The review_before_send flag defaults to true on both inbound-reply.md and weekly-linkedin.md. Replies sit as Gmail drafts under the prospect's thread. LinkedIn posts sit in the scheduled-posts queue. You approve from your phone. Architecture principle 4 (architecture.tsx lines 61-63) is explicit: every action is logged and reversible, drafts are previewed before they send, you can roll back an entire morning of work with one click. pipeline-review.md has review_before_send: false because it only writes to your own Notion page, which you own and can edit at any time.
I already use Pipedrive / HubSpot / a spreadsheet. Do I have to switch CRMs?+
No. pipeline-review.md has one line naming the CRM surface: crm: pipedrive.com. If your CRM is HubSpot, change that line to hubspot.com. If your CRM is a Google Sheet, change it to the sheet URL. Clone's Computer Agent opens whatever surface you name. It does not require a specific CRM, does not require OAuth scopes into any of them, and does not import your deals into a new system. Your CRM stays exactly where it is.
What about redacting client data from the weekly LinkedIn post?+
weekly-linkedin.md references a second file at memory/rituals/own-growth/redaction.md. Standard redaction rules: no client name, no client domain, no list name, no open/click rate, no revenue figure, no product SKU. The post extracts a reusable principle (e.g., 'a question in the subject line outperformed a declarative one') without naming who you applied it for. Many freelance marketers have an explicit NDA on this, and the redaction file is where the NDA rules live. When an NDA gets tighter, you edit one file, not 12 Jasper prompts.
Does this work with LinkedIn given LinkedIn's automation rules?+
weekly-linkedin.md does not scrape LinkedIn search, does not auto-connect, does not auto-message, and does not send InMail. It opens your LinkedIn author page in your own Chrome profile and uses LinkedIn's native scheduled-posts feature to queue a post for you to review. That is the same flow a human would do if they opened LinkedIn, clicked Start a post, typed, and clicked Schedule. It is not a LinkedIn automation tool in the prohibited sense.
How does Clone compare to hiring a virtual assistant to do these three things?+
A VA at $15 to $30/hr doing these three rituals costs 4 to 6 hours a week, roughly $240 to $720/mo, on top of the SaaS tools they use. They also need access to your Gmail, your LinkedIn, your Pipedrive, and your Notion. The trust surface and the training surface are both non-trivial. Clone is $49/mo flat, runs under your own user account on your own Mac, and the configuration surface is three text files you can read in under five minutes. VAs remain useful for judgment work: approving the drafts, negotiating pricing replies, handling edge cases. They are not required to run the base rituals.
What if the prospect's website is behind a login or their LinkedIn is private?+
The Computer Agent falls back gracefully. If the site returns a login wall, the ritual skips to the email signature and any URL embedded in the email body. If the LinkedIn is private (no public posts), the ritual skips the LinkedIn source. The draft still runs, using voice.md + pricing.md + the inbound email text itself. The first sentence of the draft is still grounded; it is just grounded in what the prospect wrote, not in what their site shows.
Does Clone work on Windows?+
The product is Mac-first today. Clone operates your desktop apps from your desktop, which means the Computer Agent runs as a native Mac app that controls Chrome, Gmail, LinkedIn, and other surfaces through OS-level input. A Windows build is on the roadmap. If you are on Windows, the right answer right now is to wait or run Clone on a secondary Mac alongside your main Windows machine.
What do I actually have to write before I run this?+
Three things. First, memory/voice.md, which is plain-text paste of your last 10 outbound emails (maybe 300 to 600 words total). Second, memory/pricing.md, which is your tier list, three to five tiers, one line each, with a range not a point price. Third, a Gmail filter that applies the label marketing-inbound to new inbound from non-clients. Total setup time is roughly 20 minutes. After that, the three ritual files are the starters on this page, copy-paste with your Chrome profile name swapped in.