hussh MCP
Connect OpenClaw, Hermes or any MCP agent to One. Discover scopes, ask for consent and retrieve only the approved encrypted export.
1. Get a developer token
2. Connect your agent
- Remote MCP
Authorization: Bearer <developer-token>Use a bare HTTPS MCP URL and keep the developer token machine-local in the host secret store.
- REST API
/api/v1Use versioned endpoints for scope discovery, consent requests, status checks, and encrypted scoped exports.
- npm bridge
npx -y @hushh/mcpUse the public launcher when an external agent still expects a local stdio process instead of remote MCP.
Using OpenClaw?
openclaw plugins install clawhub:@hushh/one-mcpThe plugin still requires a developer token and user-approved scopes. Install ClawHub plugin
3. Request access with consent
- 01
Search available scopes
Call search-user-scopes with the person's identifier. Scopes are dynamic and come from their indexed PKM.
- 02
Request a specific scope
Ask for one discovered scope with a transparent purpose, expiry, timeout, and connector key-wrapping metadata.
- 03
Wait for approval
The user approves or denies inside the hussh surface. Poll only with the returned request_ref and interval.
- 04
Retrieve the approved export
Use grant_ref and expected_scope to retrieve the encrypted scoped export in a trusted connector process. There is no plaintext fallback.
Scope rules
Discover scopes per user. A broader active grant can satisfy a narrower request; a narrow grant never expands silently. Consent MCP grants no write authority. Trusted-device writes require separate owner authorization.
attr.{domain_slug}.{scope_slug}.*- A dynamic semantic branch returned by per-user discovery. Do not hardcode a domain key.
cap.one.invoke- A reserved One task capability. It grants no user-data read or mutation authority by itself.
Tools & connection examples
Live endpoints and tools from the consent API. App authentication never replaces user consent.