Omnesis
omnesis.dev- Category
- AI Tools
- Rank
- No. 1392Tools index
Previous survey · No. 1386 ·
- Type
- TOOL
- Use case
- Data, Retrieval & Knowledge
- GitHub
- 50 stars
- Latest release
- v0.7.8
- Date
About
Omnesis is a self-hosted context layer that indexes a user's personal data, including WhatsApp/iMessage, bank transactions, health data, browsing history, photo metadata, emails, call logs, and voicemails, into a connected graph so AI agents like ChatGPT, Claude, or Codex can query it. Access is controlled per-agent through configurable data-source permissions and privacy policies written in prose, with voice notes transcribed and OCR run on images in emails and PDFs.
At a glance
Omnesis collects information from several personal services into a searchable index you operate. It is worth investigating if you want an agent to find related notes and records without searching each source separately.
Good fit
People who can run an alpha service and start with a small, low-sensitivity collection.
Consider something else when
Keep records in their existing services if you cannot safely hold a combined copy or limit who can query it.
Start here
Index a folder of synthetic notes and verify a known search result. Compare the information returned through the privacy gate with what a direct query can see before connecting a real account.
Cost and requirements
AGPL-licensed alpha project with documented macOS, Linux and container setup. Model and connector requirements depend on the selected workflow.
Limits to know
A local index can still hold sensitive material from many sources in one place. Direct access and gated access have different exposure, and selected models may receive context.
Sources checked 2026-11-05. Source-based assessment.
Why it made the leaderboard
It's a differentiated approach to giving agents durable personal context under explicit, self-hosted access controls rather than relying on each app's own siloed memory.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.