Skip to main content
Sign in once. Your assistant loads your identity, trusted context, and permissions across Cursor, Claude Code, Codex, and VS Code. It reuses saved proofs before sensitive actions and checks agent limits before tools run. First run: Run setup. Access keys: MCP setup. OAuth details: MCP OAuth.

​
How it works

Assistants load your signed-in profile once, then reuse saved proofs when they can.
  • Look up a saved proof before a sensitive action
  • Read proofs and saved context with neus_proofs_get
  • Guide any missing step
  • Confirm agent identity and authority before tools run
  • Store or revoke secrets when signed in
You see Passed, Action needed, or Blocked in chat. Not raw tool output. Journeys: MCP journeys · Tools: MCP tools, neus_proofs_get

​
Agent setup

  1. Call neus_agent_link.
  2. If linked, reuse the proofs.
  3. If missing, call neus_agent_create and finish the hosted steps.
  4. Call neus_agent_link again until linked: true.
Details: Agent verification flow · Connect agent context

​
Connect an agent to a project

Loads the agent’s verified identity, scoped authority, and host rules into the project. Run again after permission changes.

​
Reference

Run setup

Cursor, Claude Code, Codex, VS Code.

OAuth

Sign-in flow and tokens.

Setup

Install and sign in.
Last modified on August 10, 2026