Agent requests approval
The MCP gateway sends only an action summary and SHA-256 hash. Prompt context stays local.
Private beta · MCP-native
Human approval for agent actions.
Put a hardware-backed biometric checkpoint between autonomous agents and irreversible operations.
120-second, single-use challenges Passkeys locked to goodroom.in No prompts or agent context retained
APPROVAL REQUEST
gr_8f2a91Critical action
Deploy migration to production
agent claude-code
tool db_migrate
hash sha256:7d19c3a5...2e04
Physical presence required
Verify with your enrolled passkey
Challenge expires in 01:42
Zero-trust by design
No silent execution.
High-risk calls pause until a human proves physical presence. Approval is explicit, scoped, and auditable.
You verify physically
Touch ID, Face ID, Windows Hello, or a hardware key signs a 120-second challenge.
Execution unlocks
The agent receives a short-lived, single-use proof bound to the approved action.
Approval where risk lives
Keep agents fast. Gate the irreversible.
Use human-in-the-loop approval at the boundary where an AI agent can create real-world impact.
Production changes
Pause deployments, database migrations, infrastructure changes, and destructive shell commands until an operator approves the exact action.
Financial operations
Add explicit human authorization before an agent sends a payment, changes billing, purchases resources, or commits company funds.
Sensitive data access
Require physical presence before exporting customer records, rotating credentials, changing permissions, or opening protected systems.
Built for agent runtimes
One tool. Every agent.
Add the MCP server to Claude Code, Cursor, or any compatible runtime. Your agent can request approval without gaining access to biometric credentials.
- 120-second, single-use challenges
- Passkeys locked to goodroom.in
- No prompts or agent context retained
- Cryptographic proof for every approval
Terminal
$ git clone https://github.com/kairosgoodroom/goodroomverify.git
$ cd goodroomverify && npm install
$ npm run build -w @goodroom/mcp
✓ MCP server built for private betaCommon questions
Human approval, without ambiguity.
What is human approval for AI agents?
Human approval is a control that pauses a sensitive agent action until a person reviews and authorizes it. GoodRoom.verify binds that approval to the action with a short-lived cryptographic proof.
How does GoodRoom.verify secure MCP tool calls?
An MCP gateway submits an action summary and SHA-256 hash. The operator approves a 120-second challenge with a passkey, and the agent receives a single-use proof for only that action.
Does GoodRoom.verify store agent prompts?
No. Prompt context stays with the agent runtime. GoodRoom.verify handles the action summary, action hash, challenge, and approval proof needed for verification.
Which passkeys and agent runtimes are supported?
The approval flow is designed for Touch ID, Face ID, Windows Hello, and hardware security keys. The MCP integration works with compatible agent runtimes such as Claude Code and Cursor.