Skip to main content
The Model Context Protocol (MCP) is an open standard that lets AI assistants read from external data sources. Hyprcore ships a built-in MCP server, so Claude Desktop, Claude Code, Cursor, Codex, and any other MCP-aware tool can search and read your meetings directly.
MCP requires a Pro or Team plan. The server runs locally on your Mac and follows whichever account is signed into Hyprcore.

​
What it does

Once connected, your assistant can:
  • Search your meetings by keyword or by meaning.
  • Read meeting pages — summaries, action items, and notes.
  • Read full transcripts.
  • List recent meetings to ground a follow-up question.
  • Look up people — who attended what, and a person’s history across meetings.
Practical example: in Claude Desktop, ask “What did we decide about onboarding in the design review last Tuesday?” and Claude pulls the relevant meeting from Hyprcore, reads it, and answers.

​
Set it up

Everything happens in Settings → Integrations → MCP. Hyprcore detects the AI tools installed on your Mac and lists them with their current status. For supported clients, connecting is one click — no tokens, no config files to edit by hand.
1

Open the MCP panel

Settings → Integrations → MCP. You’ll see a row for each supported client: Claude Code, Claude Desktop, Cursor, and Codex CLI.
2

Install

Click Install on the client you want. Hyprcore writes the hyprcore server entry into that client’s config file for you, pointing at the MCP server bundled inside the app.
3

Restart the client

Quit and reopen the AI tool so it picks up the new server. (Codex and Claude Code pick it up on their next run.)
4

Test the connection

Back in Hyprcore, click Test on the row. A healthy connection reports the number of tools available and the round-trip time.
Each row shows where it stands:
StatusMeaning
Not detectedThat tool isn’t installed on this Mac.
Not configuredDetected, but Hyprcore isn’t connected yet. Click Install.
ConfiguredConnected and ready.
StaleConnected, but the server path is out of date (for example, after moving the app). Click Repair.
To disconnect, click Remove — Hyprcore strips only its own entry and leaves the rest of the client’s config untouched.

​
Connect another tool

For any MCP client Hyprcore doesn’t list a one-click row for — Zed, Windsurf, Continue, JetBrains, and others — use the Connect another tool section at the bottom of the panel. Click Copy, then paste the snippet into that client’s MCP config. It looks like this: 
{
  "mcpServers": {
    "hyprcore": {
      "command": "/Applications/Hyprcore.app/Contents/MacOS/hyprcore-mcp",
      "args": [],
      "env": {
        "HYPRCORE_VAULT": "…",
        "HYPRCORE_CACHE": "…"
      }
    }
  }
}
Codex-style clients use TOML instead — use the Copy config button on the Codex row to get the right shape: 
[mcp_servers.hyprcore]
command = "/Applications/Hyprcore.app/Contents/MacOS/hyprcore-mcp"
args = []
env = { HYPRCORE_VAULT = "…", HYPRCORE_CACHE = "…" }
The actual command and env paths are filled in for you when you copy from Hyprcore — don’t hand-edit them.

​
What the assistant can see

  • All meetings in your vault — titles, summaries, action items, transcripts.
  • Notes and folder structure.
  • People and attendees across your meetings.
It cannot:
  • Modify your vault.
  • See your dictation history.
  • Access settings, account info, or raw audio and video files.

​
Disconnect

Open Settings → Integrations → MCP and click Remove on any connected client. The next time that assistant tries to reach Hyprcore, it gets nothing back.

​
Privacy

The MCP server runs as a local process on your Mac. There’s no token to generate, share, or leak — access is scoped to whichever account is signed into Hyprcore, and it follows you automatically when you sign in or out. Your meetings don’t leave the machine to reach the assistant. The assistant will send the excerpts it pulls to its own LLM (Anthropic, OpenAI, etc.) as part of generating an answer. If you’d rather keep everything inside Hyprcore, use session chat instead, which routes through your configured Hyprcore LLM provider.