Whispr primary sources
Reviewed: 2026-05-14 UTC
Organization README / repo map
- https://github.com/whisprchat/.github/blob/main/README.md
- https://raw.githubusercontent.com/whisprchat/.github/main/README.md
- Best concise source for the current public architecture map.
- Used to confirm that Whispr is split across separate protocol, app, relay, identity, and cryptography repositories rather than presented as one monolith.
- Also useful because it documents the older protocol framing in one place: PRGE pairwise-ratcheted group encryption with Ed25519, X25519, ChaCha20-Poly1305, dumb fan-out relays, and out-of-band fingerprint verification.
Whitepaper repository
- https://github.com/whisprchat/whitepaper
- https://raw.githubusercontent.com/whisprchat/whitepaper/main/whitepaper.tex
- Best source for the newer protocol design.
- Used to confirm the APQ-MLS construction: a post-quantum MLS session (ML-KEM-768, ML-DSA-65, AES-256-GCM) linked by periodic PSK export into a traditional MLS session (X25519, Ed25519, ChaCha20-Poly1305) that carries application messages.
- Also used to confirm the relay / identity split, BBS+-based sealed sender, hybrid PQ transport, SHA3 fingerprints over composite key bundles, multi-relay federation, and the richer application feature surface.
Public docs surface
- https://docs.whspr.one/
- The live public docs endpoint was very thin in this pass and mainly confirmed the existence of a
Whispr White Papersurface. - Still worth keeping because it suggests the project intends a documentation site beyond the GitHub repos, even though the currently queryable content is sparse.
- The live public docs endpoint was very thin in this pass and mainly confirmed the existence of a
Main takeaways from this pass
- Whispr is worth cataloging because it exposes a messaging-stack split that many
private chatprojects hide: relay anonymity goals, identity-server control, transport security, and group-state evolution are all separated. - The most useful analytical finding is source drift. The organization README still describes an earlier PRGE architecture, while the March 2026 whitepaper specifies a more ambitious APQ-MLS plus sealed-sender design. That mismatch should be preserved as an implementation-maturity and roadmap signal, not papered over.
- Treat this entry as a mechanism page first. The public source set is promising but still young, and the docs/web surface looked incomplete or unstable during this pass.