Zero-Knowledge LLM Routing Mesh Launches on Pure Web Stack

A new open-source project has emerged that radically simplifies how developers route requests across multiple large language models while preserving user privacy. BringYourOwnBrain.org, unveiled today on Hacker News, offers a zero-knowledge routing mesh built entirely on standard web primitives, eliminating the need for custom server infrastructure, API proxies, or centralized gateways that traditionally log every prompt and response.

What Makes This Routing Mesh Different

Unlike conventional LLM orchestration platforms that funnel requests through proprietary backends, BringYourOwnBrain operates exclusively in the browser using native Web APIs. The architecture leverages Service Workers, WebAssembly, and the SubtleCrypto API to perform cryptographic operations and routing decisions entirely client-side. No intermediate server ever sees plaintext prompts, model outputs, or user credentials.

This approach addresses a critical blind spot in existing multi-model workflows. When developers use traditional routing layers or API gateways to switch between providers like OpenAI, Anthropic, or Google, those intermediaries necessarily observe all traffic. BringYourOwnBrain.org eliminates that surveillance surface by design, ensuring that only the user's browser and the target model provider handle sensitive data.

Zero-Overhead Architecture Using Web Standards

The platform achieves its zero-overhead promise by compiling routing logic into WebAssembly modules that execute with near-native performance in the browser. Key technical components include:

• Service Worker-based request interception that reroutes API calls without modifying application code
• Client-side credential vaults using IndexedDB and SubtleCrypto for secure key storage
• Dynamic provider selection based on model availability, latency, and user-defined policies
• Fallback chains that automatically retry failed requests against alternate providers

Because the entire mesh runs in standard browser contexts, deployment requires no Docker containers, Kubernetes clusters, or cloud functions. Developers can integrate the routing layer by including a single JavaScript module in their web applications.

Release Date and Availability

BringYourOwnBrain.org was officially launched on November 2, 2026, via a Show HN post that quickly gained traction in the developer community. The project is released under an MIT license, with full source code available on the project website. Early adopters can begin testing the routing mesh immediately by following the quickstart guide published alongside the launch announcement.

Privacy Guarantees and Cryptographic Design

The zero-knowledge claim rests on a formal threat model: the routing mesh itself never possesses decrypted API keys, prompts, or model responses. User credentials remain encrypted at rest in IndexedDB, decrypted only in memory during active requests, and are never transmitted to any server controlled by the BringYourOwnBrain team.

Request metadata like model choice, latency, and retry counts are aggregated locally using differential privacy techniques before optional anonymous telemetry submission. Even if users opt into usage analytics, the system cannot reconstruct individual queries or link them to specific accounts.

What This Means

BringYourOwnBrain.org represents a shift toward privacy-first AI infrastructure that respects user autonomy without sacrificing developer convenience. By proving that sophisticated LLM orchestration can run entirely in the browser using only web standards, the project challenges the assumption that multi-model workflows require centralized platforms. As enterprises increasingly scrutinize data residency and prompt logging practices, architectures that keep sensitive information client-side may become the new baseline for production AI applications.