modelcontextprotocol / modelcontextprotocol/python-sdk

OAuth: 403 responses without insufficient_scope incorrectly retry with same token

Open
#1,602 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

auth P3
Dominant language
Python
Stars
24.3k
Forks
4k
Avg merge
1d 1h
Merged PRs (30d)
31

Description

Summary

The OAuth client unconditionally retries all 403 responses, even when the error is not insufficient_scope. This causes an unnecessary retry attempt with the same token that will fail for the same reason.

Location

src/mcp/client/auth/oauth2.py, lines 662-681

The Bug

elif response.status_code == 403:
    error = self._extract_field_from_www_auth(response, "error")
    
    # Only performs step-up if error == "insufficient_scope"
    if error == "insufficient_scope":
        self._select_scopes(response)
        token_response = yield await self._perform_authorization()
        await self._handle_token_response(token_response)
    
    # BUG: Retries unconditionally, even when no new tokens were obtained
    self._add_auth_header(request)
    yield request

Lines 679-681 execute regardless of whether step-up authorization occurred, causing a retry with the same credentials.

Expected vs Actual Behavior

Scenario Expected Actual
403 with insufficient_scope Get new tokens → retry ✅ Correct
403 with different error (e.g., invalid_token) Raise error immediately ❌ Retries once with same token, then fails
403 with no error field Raise error immediately ❌ Retries once with same token, then fails

Impact

  • Wasted network round-trip: Client makes doomed retry request that will fail for the same reason
  • Poor error feedback: Delays error reporting by one request cycle
  • Spec non-compliance: MCP Authorization Spec implies retry only for insufficient_scope
  • Resource waste: Unnecessary load on server and client

Fix

Move the retry logic inside the if error == "insufficient_scope": block and raise an error otherwise:

elif response.status_code == 403:
    error = self._extract_field_from_www_auth(response, "error")
    
    if error == "insufficient_scope":
        try:
            self._select_scopes(response)
            token_response = yield await self._perform_authorization()
            await self._handle_token_response(token_response)
            
            # Retry with new tokens
            self._add_auth_header(request)
            yield request
        except Exception:
            logger.exception("OAuth flow error")
            raise
    else:
        # Permanent authorization failure - cannot be resolved by retry
        raise OAuthFlowError(
            f"Access forbidden: {error or 'insufficient permissions'}"
        )

References


Authored by Claude, reviewed by @maxisbey

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Read src/mcp/client/auth/oauth2.py around lines 662-681, focusing on how 403 responses are handled after the error field is extracted. Confirm the behavior for insufficient_scope, another error, and no error field; done means only the scope-challenge path retries and other 403 responses raise immediately.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
authentication
Issue type
Bug
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.