Check a proposed action
A fictional purchase is within the spending limit. The principal has supplied no signed grant. Run the free example to inspect the decision and reason.
Expected: BLOCK (GRANT_REQUIRED). The example only evaluates the proposal.
Guard a real deployment · Connect MCP · Edit a policy example
Guard one Cloudflare Worker release
Start with an existing Worker and a release agent that proposes an exact version. The operator controls the deployment credential and the execution gateway.
- The Worker owner signs a short-lived grant scoped to the agent, deployment action and target Worker.
- The gateway evaluates the proposed version with authenticated Vizier and verifies that the signed receipt binds to that exact request.
- The gateway holds
REVIEWfor a human, stops onBLOCK, and appliesALLOWonly with the required operator approval for that version.
Set up the operator-side adapter · Configure signed authority and verification
Connect Vizier over MCP
Add this HTTPS address to your MCP client using Streamable HTTP.
Main tool vizier_verify_action. Connection and request example · A2A agent card · OpenAPI
First MCP check
Run one synthetic purchase check after connecting. Expected: BLOCK with GRANT_REQUIRED. The fictional purchase is within its declared limit, but the principal has not signed a grant.
Copy the complete stateless MCP command
curl -sS 'https://vizier.agendamcp.dpdns.org/mcp' \
-H 'Content-Type: application/json' \
-H 'Accept: application/json, text/event-stream' \
-H 'MCP-Protocol-Version: 2026-07-28' \
-H 'Mcp-Method: tools/call' \
-H 'Mcp-Name: vizier_verify_action' \
--data-binary '{"jsonrpc":"2.0","id":"first-vizier-check","method":"tools/call","params":{"name":"vizier_verify_action","arguments":{"agent":{"id":"example-agent","owner":"example-owner"},"principal":{"id":"example-owner"},"action":{"type":"purchase","target":"synthetic.example.invalid","parameters":{"amount":150,"currency":"USD"}},"authority":{"allowed_actions":["purchase"],"constraints":{"max_amount":1000,"currency":"USD"}},"context":{"request_id":"synthetic-onboarding-001","timestamp":"2026-10-09T00:00:00Z","source":"mcp"}},"_meta":{"io.modelcontextprotocol/protocolVersion":"2026-07-28","io.modelcontextprotocol/clientCapabilities":{}}}}'
For an initialized session client, copy only the arguments below into vizier_verify_action. The client supplies its session headers.
{
"agent": {
"id": "example-agent",
"owner": "example-owner"
},
"principal": {
"id": "example-owner"
},
"action": {
"type": "purchase",
"target": "synthetic.example.invalid",
"parameters": {
"amount": 150,
"currency": "USD"
}
},
"authority": {
"allowed_actions": [
"purchase"
],
"constraints": {
"max_amount": 1000,
"currency": "USD"
}
},
"context": {
"request_id": "synthetic-onboarding-001",
"timestamp": "2026-10-09T00:00:00Z",
"source": "mcp"
}
}
Evaluation playground: caller-supplied authority is unverified. This deployment requires a valid principal-signed delegation grant. Requests without it return BLOCK with GRANT_REQUIRED, including anonymous evaluations and requests within the declared limit. An API key alone does not replace the grant. No external action is executed. Trusted enforcement requires an authenticated integration and server-side policy.
The examples compare proposed purchases with declared spending limits and allowed targets.
// Response will render here