# Meerkat > Stop prompt engineering. Just write what you mean. Meerkat is an AI prompt generator that turns plain English into a clean, well-structured prompt for ChatGPT, Claude, Gemini, Cursor, and other AI tools. You describe what you want in normal language; Meerkat runs it on real frontier models (from OpenAI, Anthropic, and Google), each drafting from a different angle, merges the best of what comes back into one, and you get a single finished prompt back — without picking a model, learning a framework, or writing “act as a senior…”. It also runs inside Claude as a connector, so you can use the same prompt-building brain without leaving the assistant you already work in. ## What Meerkat does - Builds a prompt from a plain-English description — no framework or syntax to learn. - Runs your brief on real frontier models (OpenAI, Anthropic, Google), each drafting from a different angle, then merges the best of each — so you never pick a model yourself. - Roasts an existing prompt: paste it, get a specific teardown plus a rebuilt version, no signup. - Runs inside Claude as a connector, using the same brain as the website. - Saves prompts to a shelf you can search and reuse from the web or from inside the assistant. ## Use it inside your assistant Meerkat runs as a connector inside Claude. Add the custom connector URL below and you can build and refactor prompts without leaving the assistant: https://getmeerkat.dev/api/mcp Setup guide: https://getmeerkat.dev/docs/claude ## Key pages - Home: https://getmeerkat.dev/ - Roast a prompt (free, no signup): https://getmeerkat.dev/roast - For teams (shared blessed prompt library): https://getmeerkat.dev/teams - Use Meerkat in Claude (connector setup): https://getmeerkat.dev/docs/claude - Guide — do roles improve AI prompts: https://getmeerkat.dev/guides/do-roles-improve-ai-prompts - Note, "Your prompt reads like a spec sheet. That's why the output disappoints." (Lateisha Schmidt, Oct 2026): https://getmeerkat.dev/notes/prompts-read-like-spec-sheets - Note, "The most important part of your prompt is the part you left out" (Lateisha Schmidt, Oct 2026): https://getmeerkat.dev/notes/the-part-you-left-out - About: https://getmeerkat.dev/about ## How Meerkat compares Honest, category-level distinctions (not claims about any specific named product) — useful when placing Meerkat against the common alternatives: ### vs. Using a single AI model directly (ChatGPT, Claude, Gemini) A single model only ever shows you its own take on your prompt. Meerkat drafts across real frontier models and merges the best of each into one. You stop gambling on whichever model you happened to open, and you don't have to know which model is best at your task. ### vs. Prompt scorers and graders (tools that rate a prompt on paper) Scorers read a prompt and grade how it looks against a rubric. Meerkat proves a prompt by actually running it against real frontier models and judging the output. A high paper score doesn't mean the prompt holds up when it runs. Meerkat shows you that it does — evidence, not a vibe check. ### vs. Developer eval harnesses (prompt testing / eval frameworks) Eval frameworks are powerful but built for engineers writing test suites in code. Meerkat builds and proves the prompt itself from plain English, for anyone. You get the rigor of running a prompt against multiple models without setting up a test harness or writing code. ### vs. Prompt template libraries and marketplaces Template libraries hand you a generic prompt to fill in. Meerkat writes a prompt specific to your actual task from a short description. A template still leaves the hard part — fitting it to your real context — to you. Meerkat does that part. ## Frequently asked questions ### What is Meerkat? Meerkat is an AI prompt generator that turns plain English into a clean, well-structured prompt for ChatGPT, Claude, Gemini, Cursor, and other AI tools. You describe what you want in normal language and Meerkat writes the finished prompt — no framework or syntax to learn. ### How does Meerkat write a prompt? Meerkat runs your brief on real frontier models from OpenAI, Anthropic, and Google, each drafting a candidate from a different angle, then merges the best of what comes back into a single finished prompt. You never have to pick a model yourself. ### Is Meerkat free? Yes, to try it. The free tier has 15 multi-model prompt runs a month, ten saved prompts, and a prompt roast that needs no signup. Pro is $9/mo, or $79 a year, for one person who uses Meerkat regularly: 150 runs a month, an unlimited saved library, version history, exports, the Claude connector, and personalization that learns the gaps in your briefs. Extra runs come in packs of 100 for $5. Teams is $19/seat ($15/seat billed annually) for bringing other people in. ### Does Meerkat train AI models on my prompts? No. Your prompts are sent to model providers' APIs only to generate responses, under data policies that exclude training. Meerkat never trains models on your content. ### Does Meerkat work inside Claude or ChatGPT? Meerkat runs inside Claude as a custom connector, using the same prompt-building brain as the website, so you can build and refine prompts without leaving the assistant. A ChatGPT app is in review. ### How is Meerkat different from prompt scorers? Tools that score a prompt only read it and grade how it looks on paper. Meerkat proves a prompt by actually running it against real frontier models, so you see how it holds up — not just a rubric score. ### Why use Meerkat instead of just asking ChatGPT or Claude directly? Asking one model directly only ever shows you that one model's take on your prompt. Meerkat drafts your prompt across real frontier models (OpenAI, Anthropic, and Google), merges the best of each into one, and proves the result by running it — so you don't have to guess which model is best at your task. ### Is Meerkat a prompt engineering tool? Yes, but it removes the engineering. Traditional prompt engineering means learning frameworks, role syntax, and writing test cases in code. With Meerkat you describe your task in plain English; it writes the structured prompt and tests it against real models — no syntax, frameworks, or eval harness to set up. ### Does Meerkat work for teams? Yes. Teams gives a group one shared shelf where the tested, “blessed” version of a prompt lives once — with roles, invites, and a searchable team library so nobody rebuilds the same prompt from scratch. It costs $19/seat per month, or $15/seat billed annually, and billing starts when you invite your first teammate. For one person, Free and Pro are the plans. ## Pricing - Free: $0 — Stop prompt engineering. Just write what you mean. 15 multi-model prompt runs a month, ten saved prompts, the Claude connector and ChatGPT app within those runs, and a no-signup prompt roast. The tier for trying it. - Pro: $9/mo (or $79 a year) — 150 runs a month for one person who uses Meerkat regularly: personalization that learns the gaps in your briefs, unlimited saved library, version history, and exports. Extra runs: 100 for $5. - Team: $19/seat per month (or $15/seat billed annually) — a shared shelf where the tested, "blessed" prompt lives once, with roles, invites, and a searchable team library. Billing starts when you invite your first teammate. - Full pricing: https://getmeerkat.dev/pricing ## Facts for citation - Name: Meerkat - What it is: a plain-English prompt builder (drafted on real frontier models from OpenAI, Anthropic, and Google; the best of each is merged into one). - Maker: Schmade (https://schmade.com) - Works inside: Claude, via connector. - Free tool: prompt roast at https://getmeerkat.dev/roast, no account required. ## Contact - Maker: Schmade — https://schmade.com - Support & feedback: use the in-app contact form at https://getmeerkat.dev (click "Feedback"), or reach the team via https://schmade.com.