01 / The operating case
What changes in a real workflow?
An agent switches from the test MCP endpoint to the production endpoint, whose authorization server and supported scopes differ.
- 01Client / task
- 02MCP server
- 03Tool call + arguments
- 04Downstream effect
02 / Decision contract
What the executor must check
Resolve and validate the production resource metadata, use its authorization-server discovery, and bind token acquisition to the canonical resource URI.
03 / Failure and evidence
What goes wrong, and what can be proven?
Hard-coded authorization endpoints or copied test scopes lead to token confusion when the resource changes.
Keep metadata URL and version, discovered issuer, canonical resource and validation result; never log access tokens.
04 / Canonical scope
Why this reference stands alone
Resource indicators own token targeting; this page owns discovery of who can issue for the resource.
Protocol or attack trace
Discovery boundary
Client moves from test resource R1 to production R2. R2 metadata identifies a different authorization server and canonical resource URI.
Replace R2 metadata with an issuer not approved for R2; token acquisition must stop rather than fall back to R1.
Primary references