Shared memory for your team's AI coding agents
An AI agent's memory ends when its context window does. The next agent — a new session, another tool, a teammate's machine — starts blank and repeats the work. Vibsync gives your agents a durable, shared team memory over MCP: what one agent learns, any agent can recall.
What shared memory means for agents
A single agent's context is ephemeral and trapped inside one session. A team needs the opposite:
- Durable — it survives after the session ends.
- Shared — every teammate's agents (and people) read and write the same memory.
- Cross-machine — the same memory across your laptop, CI, and a teammate's box.
- Attributable — you can tell which person and which agent wrote what.
How it works
Vibsync is a remote MCP server that adds a small, shared set of tools:
remember/recall— write and search durable team memories (a decision, a gotcha, a fix), optionally scoped to a code area so they surface automatically when an agent works there.onboard— a new session inherits open tasks, unanswered questions, teammates' active work, and a "start here" note in one call.
The "why we decided X" one agent records is inherited the next day by a different agent on a different machine.
Memory plus coordination
Shared knowledge is half the problem; the other half is not touching the same place at once. Vibsync also carries who is doing what right now: file claim/release and a shared task board keep two agents from unknowingly editing the same file. See agent coordination →
Why Vibsync
- Vendor-neutral, bring your own model — Claude Code, Cursor, and Codex share one brain; you keep your own models and keys.
- One hosted endpoint — no database to run; GitHub/Google login; free during beta.
- No code ingest — Git stays the source of truth; Vibsync holds only coordination data.
- Tamper-evident audit log — every mutation is recorded to an append-only log keyed to a token-attested actor.
Connect in a minute
{
"mcpServers": {
"vibsync": { "type": "http", "url": "https://mcp.vibsync.com/mcp" }
}
}Step-by-step for each client is in the getting started guide.
claims are advisory coordination signals, not a hard OS-level lock. Free during beta.Related reading: How to Share Claude Code Learnings Across Your Team.