Trust tools for AI agents and assistants.
Prompts are not permission. NEUS gives an agent verifiable identity, scoped permissions, and reusable trust receipts — checked before sensitive actions.
Identity
Who is this agent, and who approved it?
NEUS records the agent's operating account and public identity as a receipt. A second receipt records who approved the agent and which permissions were granted. Both are reusable — connected runtimes check them without repeating the setup.
- Agent account: the operating account and public identity.
- Approved permissions: who approved the agent and the granted scope.
- Portability: the same receipts work across every connected host.
Connected tools confirm limits before a protected action runs.
Permissions
Allowed actions, spending caps, and revocation.
A NEUS permission receipt includes allowed and denied actions, a USDC spending cap, an expiration date, and revocation status. Connected runtimes check those limits before a supported tool call runs.
Identity and limits. Checked before tools run.
One proof. Every surface that accepts it.
- First useVerify once
- Next appReuse
- Agent actionReuse
- Network effectCompounds
Profile owners control visibility. Connected systems check status.
Before the action
The runtime checks the receipt, then runs.
Identity alone is not enough. NEUS records who the agent is, what it may do and spend, then connected runtimes check that receipt before a supported action runs. If the action is outside the limits, the runtime stops it before it happens.
Frameworks run the tools. NEUS checks identity, context, and authority before they run.
MCP FAQ
Trust tools for AI agents and assistants.
Ready to add control?
One setup. Same identity, context, and limits in Cursor, Claude Code, Codex, and production.