DocsLog in

The AI trust layer between agents: identity, agreements, safeguards, and outcome intelligence for A2A.

Commit 82a1651a

82a1651a33d0Merged

aryaamin <[email protected]> · 2 months ago · 2 files · ~2 modified

Browse files at this commit

Description

fix: align MCP scope with Descope issuer

Details

Commit
82a1651a33d0
Parent
c95a8380814f

README.mdmodifiedView file

@@ -35,6 +35,10 @@
3535 consent page, and retry with an audience-bound bearer token. The local stdio server remains for
3636 development and obtains credentials through its host environment.
3737
38+The provisioned Descope project initially issues `full-access`. Before a public beta, configure and
39+enforce separate `profile:read`, `interaction:write`, `feedback:write`, `agent:manage`, and
40+`network:contribute` scopes in Descope.
41+
3842 ## What the demo proves
3943
4044 The deterministic demo creates requester/provider/subagent identities, verifies scoped delegation, signs a contract, sends an A2A-shaped message with the OpenClasp extension, keeps its raw body local, challenges a contradicted claim, blocks a deterministic violation, mediates with mutual consent, verifies bilateral feedback and receipts, updates task-specific history, and reduces confidence for a new agent version. It also rejects signature tampering, expired delegation, replay, and a tampered receipt.