Cluely arrived with a bold pitch: an invisible AI overlay that listens to your meetings, coaches you in real time, and hides from screen-share detection. The product works. But at $14.99 / month for the base plan — and with a 270 MB app that phones home to its servers — a lot of people started asking: is there a leaner alternative?
The answer is Mufal. We built the same core experience — live transcription, AI chat, screen-capture invisibility — and made it three times cheaper to start, an order of magnitude lighter, and privacy-first by design.
The problem with "AI meeting tools" today
Most tools in this space fall into one of two traps. The first is the bloated app trap: the vendor bundles their own AI infrastructure into a 200–300 MB desktop app, charges a premium subscription to cover server costs, and uses your meeting audio to improve their models. You pay for their infrastructure, not just the product.
The second trap is the fragile integration trap: the tool only works inside specific platforms (Zoom, Teams, Meet), breaks on OS updates, and requires admin permissions to install audio drivers. Real meetings don't follow a script.
Cluely, for all its cleverness, falls into both of these traps. It's a well-funded, well-designed product — but you're paying $15/month to rent AI time on someone else's servers inside a 270 MB binary.
Head-to-head: Mufal vs Cluely
| Feature | Mufal | Cluely |
|---|---|---|
| Monthly price | $5 / month (Basic) $15 / month (Pro+, keys included) |
$14.99 / month |
| App size | ~15 MB (Tauri / Rust) | ~270 MB |
| Real-time transcription | ✓ AssemblyAI Universal Streaming | ✓ |
| AI chat during meetings | ✓ Any OpenRouter model | ✓ Locked to Cluely's backend |
| Screen-share invisibility | ✓ Native content-protection API | ✓ |
| AI provider choice | Any model via OpenRouter GPT-4o, Claude, Gemini, Mistral… |
Vendor-locked |
| Your audio data | Your AssemblyAI key — your data | Processed on Cluely servers |
| Custom AI modes | ✓ Unlimited modes + templates | Limited presets |
| Session history | ✓ Full cloud sync, searchable | ✓ |
| Platform | macOS + Windows | macOS + Windows |
| Free trial | ✓ 3-day free trial | Limited free tier |
Why the price difference is so large
Cluely processes everything on its own infrastructure. Every word you speak in a meeting hits their servers, runs through their AI, and returns a response. That costs money — and you're the one paying for it, at a margin that funds a $15M valuation.
Mufal is architected differently. On the Basic plan ($5/month), you use your own API keys: your AssemblyAI account for transcription, your OpenRouter account for AI. Your audio goes to AssemblyAI directly (both have generous free tiers that cover most users), your AI requests go to OpenRouter directly. Mufal is the software layer that ties them together — and that software layer costs $5/month.
On the Pro+ plan ($15/month), we absorb the API costs for you — AssemblyAI and OpenRouter keys are included. Same price as Cluely, but with full model choice and zero data lock-in.
What 270 MB actually means
Cluely's app size isn't an accident. It bundles AI models, Electron (a full Chromium browser), Node.js, and a significant amount of telemetry infrastructure. That 270 MB sits in RAM, competes with your IDE and browser, and slows your machine down during exactly the meetings where you need it most.
Mufal is built on Tauri, which uses the OS's native WebView instead of shipping a browser. The entire app is under 15 MB. It uses less than 50 MB of RAM during a recording session. You won't notice it running — which is exactly the point.
Model freedom matters more than you think
One of Cluely's quiet limitations is that you're using whatever model they've decided to serve you, at whatever quality and speed they've configured. You can't swap to Claude 3.5 Sonnet when GPT-4o is slow. You can't use a cheaper model when you don't need top-tier reasoning. You can't experiment.
Mufal routes AI requests through OpenRouter, which gives you access to every major model in a single subscription: GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, Mistral Large, Llama 3, and dozens more. You pick the model in Settings — it takes three seconds. Different meetings call for different models, and you shouldn't be locked out of that choice.
Privacy that's structural, not just a checkbox
When Cluely processes your audio, it travels to their servers. The company has a privacy policy that says the right things. But the data leaves your machine and touches their infrastructure — and for a lot of use cases (sensitive client calls, confidential strategy discussions, job interviews), that's a hard no.
With Mufal on the Basic plan, your audio goes to AssemblyAI using your API key — you are the account holder, you control the data retention settings, you can delete your data. Your AI requests go to OpenRouter under your account. Mufal never sees your audio. It never sees your transcripts unless you enable cloud sync (opt-in).
Privacy here isn't a feature we bolted on — it's a consequence of the architecture.
The bottom line
If you're paying $14.99/month for Cluely and you're happy, that's fine. It's a good product. But if you've been looking for something leaner, cheaper, or more privacy-conscious — or if you just want to pick your own AI model — Mufal is the straightforward upgrade.
- $5/month with your own API keys (both AssemblyAI and OpenRouter have free tiers)
- $15/month with all keys included — same price as Cluely, with full model freedom
- 3-day free trial on both plans, no credit card required to start
Download Mufal and run it next to your next meeting. You'll notice the difference in the first five minutes.