BerriAI/litellm · error · HTTPException

Key requests MCP toolsets not allowed by team '{team_obj.tea

Error message

Key requests MCP toolsets not allowed by team '{team_obj.team_id}': {sorted(disallowed_toolsets)}. Team allows: {sorted(team_mcp_toolsets)}.

What it means

HTTPException(403) from _validate_requested_toolsets: the key request asks for MCP toolsets beyond the team's own allowlist (a non-empty team mcp_toolsets list). The message names the offending toolsets and the team's allowed set so the caller can correct the request.

Source

Thrown at litellm/proxy/management_helpers/object_permission_utils.py:684

            return
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail={
                "error": (
                    "Key is not in a team. MCP toolsets cannot be assigned to "
                    "personal keys by non-admin callers. Disallowed toolsets: "
                    f"{sorted(requested_toolsets)}."
                )
            },
        )
    team_op: Final = team_obj.object_permission
    team_mcp_toolsets: Final = team_op.mcp_toolsets if team_op is not None else None
    if not team_mcp_toolsets:
        return
    disallowed_toolsets: Final = requested_toolsets - set(team_mcp_toolsets)
    if not disallowed_toolsets:
        return
    raise HTTPException(
        status_code=status.HTTP_403_FORBIDDEN,
        detail={
            "error": (
                f"Key requests MCP toolsets not allowed by team '{team_obj.team_id}': "
                f"{sorted(disallowed_toolsets)}. "
                f"Team allows: {sorted(team_mcp_toolsets)}."
            )
        },
    )


def _extract_requested_vector_stores(
    object_permission: ObjectPermissionDict | None,
) -> set[str]:
    """Return vector_store IDs from a key's object_permission dict."""
    if not object_permission or not isinstance(object_permission, dict):
        return set()
    raw: Final = object_permission.get("vector_stores")

View on GitHub (pinned to 77b7c6c40c)

Solutions

1. Remove the disallowed MCP toolsets from the key request, or ask an admin to allow them for the team.

When it happens

Trigger: Thrown at litellm/proxy/management_helpers/object_permission_utils.py:684 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/d755a80b31643003. Report an issue: GitHub.