New Open-Source Mac App Parrot Offers Free, open-source AI meeting recorder for Mac

A solo developer has released Parrot, a free and open-source meeting recorder for Mac that pairs call transcription with a live AI "co-pilot" capable of pulling answers from a user's own documents while a conversation is still happening.

The app, posted to Hacker News under the title "Show HN: Parrot – Open-Source Smart Meeting Recorder with Co-Pilot on Mac," detects when a call starts, suggests responses drawn from files the user has uploaded, and stores a record of past meetings so users can query them later. According to its release notes, the app is 23 MB, requires macOS 14 or later on Apple Silicon, and can also be built from source. It was released on September 30, 2026.

Users can brief Parrot in advance by uploading documents such as pricing sheets, FAQs or playbooks in PDF, text or markdown format, in most languages. These are processed locally using Apple's on-device language framework, and the developer says an English question can retrieve an answer from a Turkish or Spanish document without any file leaving the Mac. Every answer card indicates whether it came from a user's documents or from the AI's general knowledge, and users can restrict the copilot to document-only answers.

The app supports "profiles" for different call types — such as sales, interviews or customer support — each of which changes what the copilot watches for during a call and how it writes a post-call report. Seven built-in profiles are included, and users can create custom ones with their own rules, documents and tracked signals. Calendar integration can automatically select a profile based on an event's title, and individual profiles can be restricted to on-device-only processing.

On data handling, Parrot offers two operating modes that can be mixed and switched at will. A fully local mode uses Whisper for transcription and Ollama for running the copilot and generating reports, requiring no internet connection after an initial one-time model download. An alternative mode lets users connect their own API keys for providers including Claude, OpenAI, Gemini, Groq, Deepgram and others, with requests sent directly from the user's Mac to the chosen provider rather than through any company server. The developer states Parrot operates no servers of its own and collects no account data, telemetry or analytics; the associated website uses Google Analytics for visit counts only with user consent.

According to the developer's published breakdown, audio is never uploaded unless a user selects Groq or Deepgram for transcription; transcript text is sent only to the chosen copilot while it is active and can be paused; and full documents are never uploaded, with only matching passages sent alongside a question. Calendar and third-party app connections are off by default and limited to explicitly shared data.

The app was built by a developer named Uygar, who described Parrot as a personal tool that grew out of frustration with existing call-recording and meeting-assistant products, which he said were often missing needed features, paywalled, or routed client calls through third-party clouds. He said the app has been used on all of his client calls since its creation and was built with Claude as a coding assistant, mostly during late-night sessions. He described it as his first open-source project and invited users to report issues or contribute fixes. Parrot is free to use, with an optional donation option for users who find it valuable.