Integrations
OneMax Agent as MCP memory
Dictate on your Mac, keep notes in the OneMax cloud notebook, then let Claude Code or Cursor read and write that notebook over Model Context Protocol (MCP). The agent gets searchable memory of what you said — not a second chat transcript.
What you get
| Tool | Use it to |
|---|---|
save_prompt_memory | Store a note or prompt the agent should remember |
search_prompt_memory | Search past notes by meaning or keywords |
get_daily_memory | Pull what you captured on a given day |
rebuild_memory_organization | Refresh how notes are grouped into topics |
1. Create a personal access token
- Sign in to the OneMax cloud notebook.
- Open Settings → MCP access.
- Create a token. Copy it once — OneMax only stores a hash.
Endpoint: https://onemaxagent.com/mcp
Header: Authorization: Bearer omx_pat_…
2. Claude Code
claude mcp add --transport http onemax https://onemaxagent.com/mcp \
--header "Authorization: Bearer omx_pat_YOUR_TOKEN"
3. Cursor
Add to mcp.json:
{
"mcpServers": {
"onemax": {
"url": "https://onemaxagent.com/mcp",
"headers": {
"Authorization": "Bearer omx_pat_YOUR_TOKEN"
}
}
}
}
Privacy
- Transcription and notes run in the cloud (internet required).
- Notes are not used for ads.
- A personal access token can only call
/mcp, not billing or account APIs. - Revoke a token anytime in Settings; calls with that token then return 401.
FAQ
Is this a local MCP server?
No. It is a remote HTTPS MCP endpoint. Your Mac client still does dictation; the notebook and MCP live on onemaxagent.com.
Do Free plan users get MCP?
Yes. Free and Pro can create tokens. Free still has the 5,000 words/week transcription cap.
Can the agent execute my notes as commands?
MCP returns note text as data. How your agent uses that text is up to the agent — treat notes as context, not shell instructions, unless you explicitly ask it to.