Contexteer uses a distributed hash table to find and connect peers without any central registry or relay server. Once connected, all communication is encrypted and direct — including queries, responses, file writes, and access control decisions.
This isn't a VPN workaround. Peers discover each other by public key across NAT boundaries, over any network, without port forwarding or configuration.
Every competing product — from enterprise knowledge bases to cloud AI assistants — sends your documents to a remote server to generate answers. Contexteer is different: indexing, embedding, retrieval, and inference all happen locally on your device.
The result: no data leaves your machine, there's no per-query API cost, and you're not dependent on any third-party uptime. If your device is on, your context is available.
Sharing a file with a peer doesn't mean sharing everything in it. Contexteer enforces policies at the content level: redact financial figures, block specific keywords, apply regex patterns, or write a policy in plain English — the engine enforces it on every response.
Policies are evaluated by an on-device AI model before any content is sent. An optional safety gate runs a second pass for high-sensitivity data.
Contexteer indexes across the full spectrum of document formats — not just plain text. PDFs, Word documents, spreadsheets, slide decks, code files, archives, and Markdown are all supported through a streaming parser pipeline that works on files of any size.
The pipeline is memory-efficient by design: files are processed as streams, not loaded into memory wholesale. Handles multi-gigabyte document sets without issue.
Contexteer is built on the Hypercore protocol family — a suite of open, cryptographically-verified distributed primitives developed for the peer-to-peer web. Each layer adds a specific capability without introducing centralization.
Hover any layer to learn more. The stack is open-source, auditable, and has no dependency on Contexteer servers.
Contexteer's P2P protocol is model-agnostic. An AI agent running on one machine can query a context server on another machine directly — with the same fine-grained access control that applies to humans.
This enables entirely new workflows: autonomous agents that share, annotate, and act on distributed knowledge — without a central orchestration layer.
Watch an animated walkthrough of the full context-sharing flow.