chore: Update SDK documentation (#3551)

Co-authored-by: marvin-context-protocol[bot] <225465937+marvin-context-protocol[bot]@users.noreply.github.com>

docs/python-sdk/fastmcp-utilities-token_cache.mdx

@@ -0,0 +1,87 @@
+---
+title: token_cache
+sidebarTitle: token_cache
+---
+
+# `fastmcp.utilities.token_cache`
+
+
+In-memory cache for token verification results.
+
+Provides a generic TTL-based cache for ``AccessToken`` objects, designed to
+reduce repeated network calls during opaque-token verification.  Only
+*successful* verifications should be cached; errors and failures must be
+retried on every request.
+
+Example:
+    ```python
+    from fastmcp.utilities.token_cache import TokenCache
+
+    cache = TokenCache(ttl_seconds=300, max_size=10000)
+
+    # On cache miss, call the upstream verifier and store the result.
+    hit, token = cache.get(raw_token)
+    if not hit:
+        token = await _call_upstream(raw_token)
+        if token is not None:
+            cache.set(raw_token, token)
+    ```
+
+
+## Classes
+
+### `TokenCache` <sup><a href="https://github.com/PrefectHQ/fastmcp/blob/main/src/fastmcp/utilities/token_cache.py#L46" target="_blank"><Icon icon="github" style="width: 14px; height: 14px;" /></a></sup>
+
+
+TTL-based in-memory cache for ``AccessToken`` objects.
+
+Features:
+- SHA-256 hashed cache keys (fixed size, regardless of token length).
+- Per-entry TTL that respects both the configured ``ttl_seconds`` and the
+  token's own ``expires_at`` claim (whichever is sooner).
+- Bounded size with FIFO eviction when the cache is full.
+- Periodic cleanup of expired entries to prevent unbounded growth.
+- Defensive deep copies on both store and retrieve to prevent
+  callers from mutating cached values.
+
+Caching is disabled when ``ttl_seconds`` is ``None`` or ``0``, or
+when ``max_size`` is ``0``.  Negative values raise ``ValueError``.
+
+
+**Methods:**
+
+#### `enabled` <sup><a href="https://github.com/PrefectHQ/fastmcp/blob/main/src/fastmcp/utilities/token_cache.py#L89" target="_blank"><Icon icon="github" style="width: 14px; height: 14px;" /></a></sup>
+
+```python
+enabled(self) -> bool
+```
+
+Return whether caching is active.
+
+
+#### `get` <sup><a href="https://github.com/PrefectHQ/fastmcp/blob/main/src/fastmcp/utilities/token_cache.py#L95" target="_blank"><Icon icon="github" style="width: 14px; height: 14px;" /></a></sup>
+
+```python
+get(self, token: str) -> tuple[bool, AccessToken | None]
+```
+
+Look up a cached verification result.
+
+**Returns:**
+- ``(True, AccessToken)`` on a cache hit, ``(False, None)`` on a miss
+- or when caching is disabled.  The returned ``AccessToken`` is a deep
+- copy that is safe to mutate.
+
+
+#### `set` <sup><a href="https://github.com/PrefectHQ/fastmcp/blob/main/src/fastmcp/utilities/token_cache.py#L118" target="_blank"><Icon icon="github" style="width: 14px; height: 14px;" /></a></sup>
+
+```python
+set(self, token: str, result: AccessToken) -> None
+```
+
+Store a *successful* verification result.
+
+Only successful verifications should be cached.  Failures (inactive
+tokens, missing scopes, HTTP errors, timeouts) must **not** be cached
+so that transient problems do not produce sticky false negatives.
+