Switchboard

Switchboard

**Switchboard gives OpenCode, Claude Code, Codex, and any Cloud Agent one shared, owner-governed memory.**

*Built by [Egoist Machines, Inc.](https://egoistmachines.com/) - efficient full-stack infrastructure for reliable AI systems.*

Switchboard is local-first, model-neutral memory for coding agents. It keeps approved preferences, facts, project context, and instructions available when you change models, editors, sessions, or checkouts. Optional hosted sync carries the same governed memory across machines.

# OpenCode, session one
You: Remember that new test files use the .spec.mjs suffix.
OpenCode: I'll save that to your Passport.

# OpenCode, a fresh session with another model
You: Add a test for the parser.
OpenCode: I'll add parser.spec.mjs to match the project's test naming convention.

# Claude Code, a fresh session
You: What suffix should new test files use?
Claude Code: Use .spec.mjs for new test files.

Why Switchboard

Coding agents usually lose context when a session ends. Claude Code and Codex have their own memory systems, while OpenCode [closed native memory as not planned](https://github.com/anomalyco/opencode/issues/8043). Even when an editor remembers something, that context stays trapped in one tool.

Switchboard gives every supported editor the same owner-governed store. Memory can follow the repository and the person doing the work without turning approval into permanent access or tying context to one checkout path.

Features

• One memory across editors. OpenCode, Claude Code, and Codex read from the same local store, subject to each client's grant.
• Owner-governed access. Pair exact clients, grant only named categories, and revoke a client or grant at any time.
• Optional cross-machine sync. Content-free lifecycle events sync separately from deletable content records, and hosted sync stays off until you link it.
• Stable project scope. Project memories are keyed to Git repository identity,