Frequently Asked Questions
Everything you need to know about the product and billing.
About Kosmo
Kosmo is an AI prompt tool at askkosmo.com. You say what you want in your own words; Kosmo asks what an expert would ask, then writes the prompt an expert would write, for the exact AI tool you use, such as ChatGPT, Claude, Gemini or Cursor. You paste that prompt into your tool. In short: before you ask AI, ask Kosmo.
Kosmo is built by its founder, Harshal Dorlikar, in Nagpur, Maharashtra, India, and launched in 2026. It is supported by the Google for Startups Cloud Program, and it is an independent company, not owned by Google. In 2026 it won the Education and Human Potential category prize in the Build with Gemini XPRIZE.
No. Kosmo is an independent product, and its only website is askkosmo.com. It is not KosmoAI, Cosmo by Socratic (askcosmo.ai), Kosmos AI Labs, the Kosmos AI scientist, or any other company with a similar name.
Kosmo 1.2 (October 2026) is our biggest update since launch. Kosmo now asks up to three expert questions with answers you can tap, builds the standards of your field and a check-your-work step into every prompt, and writes for the AI tool you use from verified tool profiles. Research prompts cite current sources, you can update a prompt, change its tool or start fresh without repeating yourself, and Vent Mode is now a real conversation. Plans, prices and checkout are new too. The full list is in the changelog.
No, Kosmo upgrades them. Think of Kosmo as the director, and ChatGPT as the actor. You tell Kosmo the vision, Kosmo writes the script, and your AI tool performs it.
Not at all. Kosmo works for any domain. Whether you need a legal analysis prompt, a fitness coaching system, an English tutor, or a creative writing assistant, Kosmo injects the right expertise into your prompt. If an AI tool can do it, Kosmo can write the prompt for it.
If you use AI every day, you know the frustrating cycle:
The Correction Loop: You ask for code, and it uses an outdated library. You tell it to fix the library, and it forgets the security constraints.
The Trust Issue: You ask it to format a report, but you still have to manually verify every single data point because you know it might hallucinate.
The Micro-management: You spend 20 minutes writing a massive, complex prompt just to get a 5-minute task done correctly.
Sycophancy is when an AI agrees with everything you say just to keep you happy, even when you are wrong. Most AI tools do this because they are trained to maximize user approval. Kosmo is built to do the opposite. If your idea has a flaw, Kosmo tells you before compiling. If your requirements contradict each other, Kosmo points it out instead of silently picking one. It is not rude. It is honest. Your project is better off hearing the truth early than getting a polished prompt built on a broken foundation.
Using Kosmo
Copy it and paste it into the tool you named. Each prompt ends with a short note on which model or mode to pick, and where to save it so the tool keeps following it.
Tell Kosmo what to change and press Update your prompt: it edits the prompt and shows what changed. Change tool rewrites it for another AI tool, and Start fresh begins a new prompt in the same conversation.
Name the tool you're using and Kosmo writes for it. Tools with their own verified profile include Claude, Claude Code, ChatGPT, Codex, Gemini, Antigravity, Cursor, Lovable, v0, Emergent, Perplexity and Midjourney. If you don't name one, Kosmo asks, or picks a sensible default and tells you which before you use the output.
Because it does the work before it writes. Kosmo reads current documentation for your tools, plans the prompt the way a specialist would, writes it, then reviews its own work. Chat replies come back in seconds; a compile takes about a minute.
When a question depends on current facts, Kosmo looks it up before answering. Research prompts tell your AI tool to use current sources and cite them with dates.
Conversations are the back-and-forth chat you have with Kosmo. You can bounce ideas, ask questions, and refine your thoughts without spending a compile. You only spend one when Kosmo writes your finished, expert-level prompt.
Yes, within a daily limit: 20 messages a day on the free plan (8 once your free compiles are used), 150 on Pro and 300 on Max. Only finished prompts use compiles.
Vent Mode is a mirror of your main chat that keeps the same context, so you can ask a basic question, vent about a bug, or go on a tangent without any of it touching your main thread. It's a real back-and-forth conversation, not a one-way inbox, and it never produces a compiled prompt. Nothing you say there gets saved automatically; if something's worth keeping, you save it to your Idea Library yourself.
The Idea Library is where you park a good idea when the timing isn't right to build it yet, whether it came up in your main chat or in Vent Mode. Nothing gets saved automatically; you manually choose what goes in. Each idea has a status (active, parked, resurfaced after 7 days, implemented, or archived), so a good idea never just gets forgotten.
Not available yet. Today Kosmo keeps your chats, prompts and saved ideas in your account, but it does not carry your preferences from one conversation to the next. Remembering your stack, style and past decisions across conversations is planned for later.
Yes. For agent frameworks like Hermes and OpenClaw that define an agent's personality in a dedicated SOUL.md file (separate from its system prompt), Kosmo has current documentation for both frameworks' conventions and compiles an accurate SOUL.md or system prompt for either one, grounded in their real docs, not a generic template.
Yes. Kosmo works in your phone's browser at askkosmo.com, and you can add it to your home screen straight from the browser, with no app store needed: on iPhone, tap Share, then Add to Home Screen; on Android, open the browser menu and tap Add to home screen or Install app. Kosmo then opens from its own icon.
Plans and billing
A compile is Kosmo writing your finished prompt. Chatting with Kosmo doesn't use compiles; each finished prompt uses one. Updating a prompt uses one only if the prompt actually changes, and a compile that fails on our side is never charged.
Not at all. You start with 3 free compiles, no card required. If you run out, you can top up for $5 (12 more compiles), or ₹199 in India (8 more compiles), valid 30 days, instead of subscribing. No strings attached.
How many compiles you get each month (40 or 90; 25 or 80 in India) and your daily chat limit (150 or 300 messages). Everything else is the same.
Plan compiles reset when your plan renews, so unused ones don't carry over. Your free compiles never expire. Top-ups last for the time shown when you buy them; the Starter Pack lasts 30 days.
In India, in rupees through Razorpay, including UPI and cards. Everywhere else, in your own currency through Lemon Squeezy, with cards or PayPal; Lemon Squeezy handles tax as the seller of record.
Yes, but signing in first is easier: your purchase goes straight to your account. If you pay without an account, create it right after paying on the same device, and the purchase is added to it.
Account settings, Cancel plan. Your plan stays active, with its compiles, until the end of the period you've paid for, and you won't be charged again.
If you were charged by mistake, yes, in full. Details are in the Refund Policy.
Prices can change for future purchases. If you have a plan, we tell you before a change applies to your next renewal, and you can cancel before it does.
Privacy and your account
Your messages and requests are used to generate your results and to keep Kosmo working. To do that, they are processed by the AI providers we rely on for reasoning and fact-checking. We do not sell your data, and we do not use it to train AI models. Full detail is in our Privacy Policy.
Account settings, Delete account. We keep it for 30 days in case you change your mind; sign in within that time to restore it. After that, your content is permanently deleted.