Char is a significant player in the agent ecosystem because it serves as a high-fidelity data ingestion layer. By converting unstructured voice conversations into structured, searchable summaries and transcripts without a cloud intermediary, it provides the "raw material" that personal AI agents need to function effectively. It acts as a memory buffer for an individual's professional life.
Technically, Char sits at the edge of the agent stack, focusing on secure data collection. Its support for the "Bring Your Own Key" model and its local-first processing means it can be integrated into private agent workflows without the data leak risks associated with centralized SaaS recorders. For developers building agents that require context from real-world meetings, Char offers a way to capture that context while maintaining strict data sovereignty.
Char, developed by the startup Hyprnote, represents a move away from the standard "recorder bot" model that has dominated the AI meeting assistant market. While tools like Otter and Fireflies rely on bots joining calls as visible participants, Char functions as a desktop application that captures system audio. This approach eliminates the social friction and privacy concerns associated with bot intrusion. The company specifically targets users in high-stakes environments—such as psychiatry or executive engineering sessions—where the presence of a third-party recording bot is often inappropriate or strictly prohibited.
The application is built using Rust and the Tauri framework, which allows it to maintain a low memory footprint while handling heavy audio processing tasks. A key differentiator for Char is its local-first philosophy. Transcription and initial processing can occur entirely on the user's device, ensuring that sensitive audio data never leaves the local environment unless cloud features are specifically enabled. For developers and privacy advocates, Char offers a "Bring Your Own Key" (BYOK) model. This allows users to connect their own API providers, such as OpenRouter or Azure OpenAI, giving them direct control over their data retention policies and model selection.
Char maintains an open-source repository, which distinguishes it from many of its closed-source competitors in the Y Combinator ecosystem. This transparency is a deliberate choice to build trust with a technical audience that is increasingly wary of how AI companies handle conversational data. The project has gained traction on platforms like Hacker News, where users have highlighted its clean UI and its utility as a case study for building high-performance audio pipelines in Rust. By keeping the core product open, the founders have created a feedback loop with their most sophisticated users, often shipping features like the "context" enhancement within 24 hours of user feedback.
Based in Seoul and active in San Francisco as part of the Y Combinator S25 cohort, Char is founded by John Jeong and Yujong Lee. The team is leaning into a specialized niche of the productivity market: the "AI Notepad." Rather than attempting to be a comprehensive project management suite, they focus on the specific transition from live conversation to structured knowledge. The product includes a floating panel for real-time controls, customizable templates for different meeting types (sales discovery, sprint planning, etc.), and a chat interface that allows users to query their transcript history using natural language. This narrow but deep focus allows them to compete effectively against larger incumbents by serving the specific segment of the market that prioritizes speed, privacy, and technical control.
An AI notepad for private meetings that works on-device without bots.
Char is hiring