Keen Bean Launches 'Invisible' AI Assistant for macOS to Draft Technical Specs
The new tool captures system audio locally to generate real-time diagrams and task lists without using visible meeting bots.
Keen Bean has released a macOS-exclusive AI meeting assistant that generates first drafts of technical specifications and diagrams in real-time. The tool aims to eliminate the manual labor of converting call notes into project deliverables for consultants, architects, and founders.
Unlike traditional AI note-takers, Keen Bean operates as a native application for Apple Silicon and Intel Macs and does not join calls as a visible bot. Instead, it captures audio directly from the system, allowing it to function in corporate environments where third-party recording bots are banned. While a meeting is in progress, the app generates "artifacts," which include draft specifications, rough mock-ups, task cards, and diagrams.
A Shift Toward Local Control
The AI meeting assistant market has become increasingly crowded with tools that prioritize summaries and transcripts. Keen Bean differentiates itself by focusing on the immediate production of structured technical drafts rather than simple text logs. To support this workflow, the app stores all data locally on the user's Mac as Markdown and JSON files, enabling seamless export to project folders or knowledge management tools like Obsidian.
Privacy and Pricing
To address security concerns common in high-compliance industries, Keen Bean utilizes a privacy architecture where audio travels directly from the Mac to a transcription service and then to the AI model. According to the company, no proxy or server managed by Keen Bean exists in that data path.
The service is available via a subscription model with two tiers. The "Seed" plan is priced at $19 per month, while the "Sprout" plan costs $39 per month; both tiers include the necessary AI usage for transcription and model processing.
Industry Implications
For professionals in high-privacy environments, such as early-stage founder discovery or specialized consulting, the "no-bot" approach removes significant social and technical friction. By shifting the output from a passive transcript to an active first draft of a deliverable, the tool targets the "blank page" problem that typically follows a technical discovery call.
What's Next
As the tool rolls out, the industry will be watching to see if the "invisible" capture method gains wider adoption over the bot-based standard. Users can also leverage the app to query their entire meeting history to recall specific decisions or objections across multiple sessions. This capability transforms a collection of isolated calls into a searchable knowledge base, further reducing the administrative overhead of technical project management.