Docs
Use the hosted MCP endpoint to let supported AI clients search and store shared team memory.
Point MCP-compatible clients at the hosted ContextVault server. The app handles authorization, workspace scoping, and memory access after the user signs in.
Server URL
https://mcp.contextvault.devAdd ContextVault to the project MCP configuration so Claude Code can discover the hosted server and start the OAuth sign-in flow.
{
"mcpServers": {
"contextvault": {
"url": "https://mcp.contextvault.dev"
}
}
}Save this in the project `.mcp.json` file. Claude Code should prompt for authorization the first time it connects.
Codex uses TOML configuration for MCP servers and a streamable HTTP client. Add ContextVault to your Codex config, then run the MCP login flow if your Codex version does not start it automatically.
[features]
experimental_use_rmcp_client = true
[mcp_servers.contextvault]
url = "https://mcp.contextvault.dev"This entry belongs in `~/.codex/config.toml` or the Codex config location your client uses. Recent Codex builds also support adding HTTP MCP servers with `codex mcp add contextvault --url https://mcp.contextvault.dev`.
Groups control which teammates can find and reuse each set of memories. They are the main way Team and Enterprise workspaces keep shared context scoped without creating separate accounts.
In the dashboard, open Workspace Management, then Groups. From there you can add a group, edit a group name and choose Rename from Actions, archive a group, or use Move memories to transfer records between groups.