Confer Labs

license

none-declared(no LICENSE file in either public repo)

source

partial(confer-proxy + confer-image public,client-app private)

self-host

no(depends on vendor attestation infra)

models

open source(undisclosed)

Marketed as open source, but neither public repository declares any license - GitHub reports no license at all - so re-use rights are undefined, and the end-user client is not public. The source that is published exists for verifiability and reproducible builds rather than for self-deployment, which makes this source-available rather than open source. No license is declared anywhere the publisher could declare one, so the license tier is unstated.

• https://confer.to/blog/2026/01/private-inference/ recorded 2026-08-13

  architecture (TEE, Noise pipes, dm-verity, reproducible nix/mkosi builds, transparency log); names the confer-proxy and confer-image repos

• https://github.com/ConferLabs/confer-proxy recorded 2026-08-13

  public proxy repo with no LICENSE file (no declared license)

Young product (repos created Dec 2025, launched late 2025/early 2026); ~70-75 stars per repo and a reported partnership "bringing foundational AI privacy to Meta", but no user/usage numbers. Notability is driven mainly by the founder's reputation.

• https://confer.to/blog/ recorded 2026-10-08

  Blog index with the product's feature timeline and the post "Confer is bringing foundational AI privacy to Meta" (Moxie Marlinspike, 17 Mar 2026); no user or usage figure.

Confer combines end-to-end encrypted chat, inference inside attested confidential VMs, encrypted memory and folders, client-side connectors and confidential workers that run code. Its chat features are otherwise conventional, and the open-weight models it uses are not named.

• https://confer.to/ recorded 2026-09-29

  Homepage describes the product as "A complete AI assistant — writing, images, web search, memory." and its structured data lists the operating systems as "Web, iOS".

• https://confer.to/blog/ recorded 2026-09-29

  Blog index listing posts including "Confidential workers for private AI", plus encrypted folders, memory, connectors, folder files and confidential workers, dated January to September 2026.

• https://confer.to/blog/2026/01/private-inference/ recorded 2026-09-29

  Post describing inference inside a confidential VM: "Your prompts are encrypted from your device directly into the TEE" using Noise Pipes, with remote attestation, dm-verity and a transparency log.

• https://confer.to/blog/2026/03/encrypted-memory/ recorded 2026-09-29

  Post announcing cross-conversation memory: "Confer now has memory, and it works differently." Facts are stored encrypted with passkey-derived keys.

• https://confer.to/blog/2026/03/encrypted-meta/ recorded 2026-09-29

  Post says "Confer’s technology has been built on top of open weight models."

• https://confer.to/blog/2026/04/connectors/ recorded 2026-09-29

  Post announcing client-side connectors: "Confer now supports connectors" for Google Drive, Gmail and Google Calendar, run on the device.

• https://confer.to/blog/2026/08/private-ai-document-search/ recorded 2026-09-29

  Post announcing folder files: "You can add PDFs, Word documents, presentations, spreadsheets, HTML, Markdown, or plain text to a folder" with search and inference inside the attested TEE.

• https://confer.to/blog/2026/09/confidential-workers-for-private-ai/ recorded 2026-09-29

  Post introducing confidential workers, TEE-backed Linux VMs where "The model can write and run programs, copy an attachment into the workspace" and return completed files such as PDFs, spreadsheets and presentations.