Percept
Percept is a real-world context layer for AI agents — giving agents ears (Omi pendant + Chrome extension for ambient audio capture and transcription), eyes (Apple Watch with push-to-talk and raise-to-
About
Percept is a real-world context layer for AI agents — giving agents ears (Omi pendant + Chrome extension for ambient audio capture and transcription), eyes (Apple Watch with push-to-talk and raise-to-speak), and awareness (a knowledge graph connecting people, projects, conversations, and events from Gmail, GitHub, Linear, Calendar, and Slack). An initiative engine detects patterns and triggers agent actions without being asked. MIT licensed, Python-based, and open source.
Technical founders and developers who want their AI agent to have continuous ambient awareness — capturing meeting context, syncing work tools, and proactively acting on patterns — and who are comfortable with the privacy tradeoffs of always-on audio capture and local self-hosting.
Pros & Cons
Pros
- check Uniquely bridges physical hardware (Omi pendant, Apple Watch) with digital work context — not just another app integration
- check Knowledge graph connects disparate data sources into a unified understanding of who/what/when across all inputs
- check Initiative engine enables proactive agent behavior — agent acts on patterns without waiting for explicit prompts
- check Broad connector coverage: Gmail, GitHub, Linear, Calendar, Slack — covers most developer and business workflows
- check Open source and self-hosted with MIT license — full data control, no cloud dependency for sensitive conversation data
Cons
- close Requires hardware (Omi pendant, Apple Watch) for the most compelling use cases — creates a purchase barrier
- close Very early project: 2 stars on GitHub, no public production deployments cited
- close Privacy implications of ambient audio capture are significant — all meetings and conversations being transcribed
- close Python installation and self-hosting adds complexity compared to SaaS alternatives
- close Initiative engine triggering autonomous actions based on conversations is a trust boundary that needs careful evaluation
More Memory
Other tools in the same category.
