# ScopeProof

ScopeProof is a global machine-native authorization boundary for AI agents.

Canonical: https://scopeproof.davisvillelabs.com
MCP: https://scopeproof.davisvillelabs.com/mcp
OpenAPI: https://scopeproof.davisvillelabs.com/openapi.json
Capabilities: https://scopeproof.davisvillelabs.com/v1/capabilities
Task guide: https://scopeproof.davisvillelabs.com/tasks/check-action
JWKS: https://scopeproof.davisvillelabs.com/.well-known/jwks.json

## Use
1. Call preflight for free structural validation. It returns no authorization verdict.
2. Call check_action immediately before a consequential action. Price: $0.01 USD, stablecoin machine-payment rails only.
3. Enforce only one of: within_scope, outside_scope, review_required.
4. Verify the returned scopeproof-receipt.v1 using verify_receipt or the public JWKS.
5. Paid results also include an additive davisville-receipt-envelope.v1 with a canonical content digest. The native ScopeProof signed receipt remains authoritative.

Receipt Envelope contract: https://davisvillelabs.com/agents/receipt-envelope-v1.json
Receipt Envelope schema: https://davisvillelabs.com/agents/davisville-receipt-envelope.v1.schema.json
Delivery contract: https://davisvillelabs.com/agents/delivery-contract-v1.json

## Trust boundary
ScopeProof checks a proposed action against the supplied structured authority envelope. It does not authenticate the authority issuer, execute/fetch the target, accept raw prompts, or accept credentials/secrets. A within_scope result is not a safety, legality, identity, delivery, or human-approval guarantee.