Frequently asked questions

Reply Pilots is a Chrome extension plus a dashboard that drafts your comments, DMs and posts inside Facebook, Instagram, LinkedIn, X, Reddit and Gmail. It reads the post and the replies already on your screen, then writes a draft grounded in your business and in the way you write. You edit it if you want to, and you send it yourself — nothing goes out on its own.

No. You describe your business and how you write once, in the dashboard, and after that every draft is a single click on the post you are already reading. There is no prompt box to fill in each time, and no prompt library you have to learn before it becomes useful.

About ten minutes end to end: install the extension, sign in to the webapp once so the extension can authorise itself, and fill in your business context and voice. Your first draft lands under a minute after that. You can improve the voice settings later without redoing anything.

You need an account. The extension is a client for your Reply Pilots account — it holds no AI keys and makes no AI calls itself, so it cannot work signed out. Sign in once at app.reply-pilots.com and the extension's popup switches to Connected.

Never. Reply Pilots drafts and you send. There is no publishing integration, no scheduler that fires on your behalf, and no auto-reply — every draft is inserted into the composer for you to review, edit and send yourself.

It reads the post and the comments visible on your screen at the moment you ask for a draft, so the reply answers the gap in the conversation instead of repeating a point three people have already made. It does not scroll the page for you, load hidden replies, or open other pages to gather more.

Almost never, and the one exception is deliberate. Drafting only happens when you press a button. The exception: when you are looking at a Facebook or LinkedIn group or page you follow, it reads the posts already on that screen along with their public like and comment counts, so the planner can rank what actually performs. Muting the site switches that off along with everything else.

Comment replies, DM and email replies, and full posts with hashtags and an image prompt. On top of those it will summarise a long thread before you answer it, rewrite a rough draft you typed yourself, and summarise or explain any text you highlight on any page.

Yes, as an option. There are six reply styles — professional, friendly, helpful, informative, humorous, or a custom one you write — plus short, medium and long length control. An optional toggle mirrors the tone of the top comments so your reply does not read like an outsider walked into the room.

Facebook, Instagram, LinkedIn, X, Reddit and Gmail today, with more on the way. The highlight-to-summarise and highlight-to-explain tools work on any page you choose to use them on, not only the supported platforms.

It works in Chromium-based browsers — Chrome, Edge, Brave, Arc and Opera among them — because it is a Chrome MV3 extension. Firefox and Safari are not supported. There is no iOS app; an Android app is in progress.

No. You add it from the Chrome Web Store in one click, and it authorises itself against your web session — there is nothing to copy, paste or configure. No AI provider key of any kind is stored in the extension — all AI requests are made by our server.

Yes. Mute Reply Pilots on any site for an hour, for the rest of the day, or until you turn it back on. While a site is muted nothing is injected into the page at all, including the feed reading that feeds the planner.

Usually because the session cookie it exchanges for its token has expired, or because the extension predates a domain change and is pinned to an old origin in its manifest. Signing in again at app.reply-pilots.com fixes the first case. For the second, remove that copy at chrome://extensions and install Reply Pilots from the Chrome Web Store — an extension cannot rewrite its own permitted origins, and the store build stays current on its own.

You tell it once. Your business context covers what you do, who you serve and what you sell; your voice settings cover formality, sentence length, the words you use and the ones you never use. Personas let you keep separate voices — one per client, or one per account — and switch between them.

Guardrails. You write the lines it must not cross — services you do not provide, prices you will not quote, claims you will not make — and a final pass rewrites any draft that crosses them. It reduces the failure rate substantially, but it is not a substitute for reading the draft: you are the one pressing send.

They will if the business context is thin, because that context is most of what separates your draft from a generic one. The usual fix is specificity — naming your actual services, your actual constraints and two or three phrases you genuinely use — rather than a different model or a longer prompt.

Yes. A workspace can hold several personas, each with its own voice, guardrails and platform settings, and you switch between them from the panel without signing out. Team members share the workspace owner's credits rather than carrying separate balances.

By default your drafts run through OpenRouter on a model chosen for the job, and the AI request is made by our server rather than by your browser. You can see what each generation cost in credits in your usage history.

Yes. Add an OpenAI, Anthropic, Google or OpenRouter key under Settings and pick your model. Runs on your own key spend no Reply Pilots credits and are not capped by your balance — you are billed by that provider directly instead.

The run fails and tells you the key was rejected. It deliberately does not fall back to our key, because a silent fallback would spend your credits on a request you asked to run on your own account. Fix or remove the key and the run works again.

Only the page you are on, and in almost every case only at the moment you ask for a draft. The single exception is the feed reading described above, on Facebook and LinkedIn groups and pages you follow, which exists so the planner can rank posts by what actually performs. Muting a site stops both.

No. Your posts, threads and drafts are yours, and nothing you generate is used to train any model. Your generations are stored in your own account history so you can search and reuse them, and you can delete them.

From the extension to your Reply Pilots account on our server, and from there to the AI provider that runs the generation — OpenRouter by default, or whichever provider your own key belongs to if you have added one. Nothing is sent to any other third party, and no AI provider key is ever stored in the extension.

No, none of those. It performs no action on any platform on your behalf — it does not post, send, like, follow, connect or message, and it does not scroll or scrape a page to collect data. It reads what is already rendered on a page you opened yourself, and it writes text into a box for you to send.

Plans are credit-based: one credit is roughly one short reply, and longer outputs use a few. Every plan includes the whole product — the tiers differ only in how many credits you get each month. Current amounts are on the pricing page.

Generation stops until your monthly allowance resets or you top up. Nothing is deleted and your settings, personas and history are untouched. If you have added your own AI key, runs on that key keep working, because they never touched your credit balance in the first place.

Your monthly allowance resets each billing cycle and does not roll over. Credits you buy on top of a plan do roll over and do not expire.

Yes — upgrade or downgrade at any time. The change applies from your next billing cycle, and your personas, guardrails, prompts and history carry over untouched.

A workspace can have several members, each with their own login and role. Credits are shared: the whole workspace draws on the owner's balance rather than each seat carrying its own.

Monthly and yearly plans can be cancelled at any time and stay active until the end of the period you have already paid for. Lifetime licences come with a 30-day money-back guarantee.