Inspiration
I often switch between Claude Code, Codex, and Hermes, and I was frustrated by having to repeat the same preferences, instructions, and project context every time. I built Context Relay because I wanted my AI tools to feel like one continuous workspace without giving up privacy or control of my data.
AI assistants are most useful when they understand how you work, but that context is usually trapped inside one tool or machine. We built Context Relay so users can carry their AI memory and harness configuration with them without surrendering control of their data.
What it does
Context Relay keeps one encrypted workspace synchronized across Claude Code, Codex, Hermes, and other harnesses on Windows and macOS. It gives supported assistants consistent access to the user’s approved memories, instructions, tasks, and settings while preserving project boundaries and user ownership.
How we built it
We built a cross-platform Tauri and React desktop app backed by Rust. A local daemon, typed protocol, encrypted vault, MCP integration, and conflict-aware synchronization form the core. We designed cloud storage as untrusted ciphertext storage and added strict validation, scoped access, transactional configuration updates, and recovery safeguards.
Challenges we ran into
The hardest challenge was safely translating shared context between tools with different formats and permission models. We also had to handle concurrent edits, deletion conflicts, cross-platform filesystem behavior, crash recovery, and secure native configuration changes without overwriting content Context Relay does not own.
Accomplishments that we're proud of
We created a privacy-first architecture that treats external inputs and cloud infrastructure as untrusted. We are especially proud of the typed versioned protocol, encrypted local workspace, deterministic conflict handling, secure MCP boundary, and guarded native update process designed to preserve existing user configuration.
What we learned
We learned that synchronizing AI context is not simply a file-copying issue. Trust, ownership, provenance, permissions, and conflict handling must be part of the data model from the beginning; and reliable recovery is just as important as the successful path.
What's next for Context Relay
Next, we plan to complete the v1 synchronization experience, expand and polish the Claude Code, Codex, and Hermes adapters, improve onboarding and conflict resolution, strengthen security testing, and prepare stable Windows and macOS releases.
Log in or sign up for Devpost to join the conversation.