aiagentsbay-mcp

StremAI's MCP memory server — the package keeps its original name for compatibility. StremAI is a shared memory layer for AI coding agents: connected agents can store what they learn while working, and other connected agents can recall it later across sessions, machines, and tools.

What this is

An MCP server that gives your coding agent shared memory tools. Your agent records what it learns, queries relevant history, and hands off context between sessions.

Not a marketplace. Not a chatbot wrapper. Memory, projects, and team features for real coding workflows.

Install

npm install -g aiagentsbay-mcp

Or run without installing:

npx aiagentsbay-mcp

Hosted setup first

Prefer the hosted StremAI endpoint with OAuth/browser sign-in when your MCP client supports it:

{
  "mcpServers": {
    "stremai": {
      "type": "http",
      "url": "https://stremai.com/api/mcp"
    }
  }
}

Use the local stdio package when you want offline-first memory or your client cannot complete the hosted flow:

{
  "mcpServers": {
    "stremai": {
      "command": "npx",
      "args": ["-y", "aiagentsbay-mcp"]
    }
  }
}

API keys remain available as the fallback for CI, scripts, and clients that cannot complete OAuth/browser sign-in.

No API key? The server runs fully offline against a local SQLite brain at ~/.agentbay/local.db, and your memories sync up automatically once you sign in.

Credential problems never take the server down

If your API key is invalid or expired, the server still starts — it logs one warning to stderr and falls back to local-first mode (the same local tool set, including agentbay_signup). Run agentbay login or fix AGENTBAY_API_KEY, then restart to reconnect cloud sync. If a token expires mid-session, only the affected cloud calls fail (new memories are written locally and sync after you re-authenticate); the server keeps running.

What you get

• Shared memory across sessions. Connected agents can recall lessons stored earlier.
• Projects that bundle memory, handoffs, and a shared context surface.
• Teams so multiple agents share context across the same workspace.
• Handoffs so one agent can pick up where another left off.

Docs

Full docs at stremai.com/docs/mcp-memory.

License

MIT