diff --git a/plugins/memory/hindsight/README.md b/plugins/memory/hindsight/README.md
index 4c7e0f6be30e..898c1bb4f668 100644
--- a/plugins/memory/hindsight/README.md
+++ b/plugins/memory/hindsight/README.md
@@ -75,8 +75,16 @@ Config file: `~/.hermes/hindsight/config.json`
 | `recall_prompt_preamble` | — | Custom preamble for recalled memories in context |
 | `recall_tags` | — | Tags to filter when searching memories |
 | `recall_tags_match` | `any` | Tag matching mode: `any` / `all` / `any_strict` / `all_strict` |
+| `recall_prefetch_tags` | — | Auto-prefetch recall only: explicit tags to use instead of generic `recall_tags` |
+| `recall_prefetch_tags_match` | `any` | Auto-prefetch recall only: tag matching mode for `recall_prefetch_tags` |
 | `auto_recall` | `true` | Automatically recall memories before each turn |
 
+`recall_prefetch_tags` and `recall_prefetch_tags_match` apply only to automatic prefetch recall. When unset, auto-prefetch falls back to the generic `recall_tags` and `recall_tags_match` settings. They do not affect the manual `hindsight_recall` tool and are ignored when `recall_prefetch_method=reflect`.
+
+If you want hard tag scoping, use a `_strict` match mode. Hindsight's non-strict `any` and `all` modes can still include untagged memories.
+
+This design changes only the read path. It does not add implicit session-derived tags, metadata-to-tag bridging, or any write-path changes to auto-retain or manual `hindsight_retain`. As a result, `recall_prefetch_tags` only matches memories that already carry those tags; untagged memories are not automatically backfilled or inferred from metadata.
+
 ### Retain
 
 | Key | Default | Description |
diff --git a/plugins/memory/hindsight/__init__.py b/plugins/memory/hindsight/__init__.py
index bc82bc40fb5d..498df10bcd18 100644
--- a/plugins/memory/hindsight/__init__.py
+++ b/plugins/memory/hindsight/__init__.py
@@ -46,6 +46,7 @@
 _MIN_CLIENT_VERSION = "0.4.22"
 _DEFAULT_TIMEOUT = 120  # seconds — cloud API can take 30-40s per request
 _VALID_BUDGETS = {"low", "mid", "high"}
+_VALID_TAG_MATCH_MODES = {"any", "all", "any_strict", "all_strict"}
 _PROVIDER_DEFAULT_MODELS = {
     "openai": "gpt-4o-mini",
     "anthropic": "claude-haiku-4-5",
@@ -217,7 +218,7 @@ def _load_config() -> dict:
     }
 
 
-def _normalize_retain_tags(value: Any) -> List[str]:
+def _normalize_tags(value: Any) -> List[str]:
     """Normalize tag config/tool values to a deduplicated list of strings."""
     if value is None:
         return []
@@ -254,6 +255,12 @@ def _normalize_retain_tags(value: Any) -> List[str]:
     return normalized
 
 
+def _normalize_tags_match(value: Any) -> str:
+    """Normalize tag match mode to a supported choice."""
+    candidate = str(value or "any").strip().lower() or "any"
+    return candidate if candidate in _VALID_TAG_MATCH_MODES else "any"
+
+
 def _utc_timestamp() -> str:
     """Return current UTC timestamp in ISO-8601 with milliseconds and Z suffix."""
     return datetime.now(timezone.utc).isoformat(timespec="milliseconds").replace("+00:00", "Z")
@@ -424,6 +431,8 @@ def __init__(self):
         self._tags: list[str] | None = None
         self._recall_tags: list[str] | None = None
         self._recall_tags_match = "any"
+        self._recall_prefetch_tags: list[str] | None = None
+        self._recall_prefetch_tags_match = "any"
 
         # Retain controls
         self._auto_retain = True
@@ -678,6 +687,8 @@ def get_config_schema(self):
             {"key": "recall_budget", "description": "Recall thoroughness", "default": "mid", "choices": ["low", "mid", "high"]},
             {"key": "memory_mode", "description": "Memory integration mode", "default": "hybrid", "choices": ["hybrid", "context", "tools"]},
             {"key": "recall_prefetch_method", "description": "Auto-recall method", "default": "recall", "choices": ["recall", "reflect"]},
+            {"key": "recall_prefetch_tags", "description": "Auto-prefetch recall only: explicit tags to use instead of generic recall_tags when searching memories (comma-separated)", "default": ""},
+            {"key": "recall_prefetch_tags_match", "description": "Auto-prefetch recall only: tag matching mode for recall_prefetch_tags", "default": "any", "choices": ["any", "all", "any_strict", "all_strict"]},
             {"key": "retain_tags", "description": "Default tags applied to retained memories (comma-separated)", "default": ""},
             {"key": "retain_source", "description": "Metadata source value attached to retained memories", "default": ""},
             {"key": "retain_user_prefix", "description": "Label used before user turns in retained transcripts", "default": "User"},
@@ -834,13 +845,17 @@ def initialize(self, session_id: str, **kwargs) -> None:
         self._bank_retain_mission = self._config.get("bank_retain_mission") or None
 
         # Tags
-        self._retain_tags = _normalize_retain_tags(
+        self._retain_tags = _normalize_tags(
             self._config.get("retain_tags")
             or os.environ.get("HINDSIGHT_RETAIN_TAGS", "")
         )
         self._tags = self._retain_tags or None
-        self._recall_tags = self._config.get("recall_tags") or None
-        self._recall_tags_match = self._config.get("recall_tags_match", "any")
+        self._recall_tags = _normalize_tags(self._config.get("recall_tags")) or None
+        self._recall_tags_match = _normalize_tags_match(self._config.get("recall_tags_match", "any"))
+        self._recall_prefetch_tags = _normalize_tags(self._config.get("recall_prefetch_tags")) or None
+        self._recall_prefetch_tags_match = _normalize_tags_match(
+            self._config.get("recall_prefetch_tags_match", "any")
+        )
         self._retain_source = str(
             self._config.get("retain_source") or os.environ.get("HINDSIGHT_RETAIN_SOURCE", "")
         ).strip()
@@ -877,10 +892,10 @@ def initialize(self, session_id: str, **kwargs) -> None:
                          self._bank_id_template, self._agent_identity, self._agent_workspace,
                          self._platform, self._user_id, self._bank_id)
         logger.debug("Hindsight config: auto_retain=%s, auto_recall=%s, retain_every_n=%d, "
-                     "retain_async=%s, retain_context=%s, recall_max_tokens=%d, recall_max_input_chars=%d, tags=%s, recall_tags=%s",
+                     "retain_async=%s, retain_context=%s, recall_max_tokens=%d, recall_max_input_chars=%d, tags=%s, recall_tags=%s, prefetch_tags=%s, prefetch_tags_match=%s",
                      self._auto_retain, self._auto_recall, self._retain_every_n_turns,
                      self._retain_async, self._retain_context, self._recall_max_tokens, self._recall_max_input_chars,
-                     self._tags, self._recall_tags)
+                     self._tags, self._recall_tags, self._recall_prefetch_tags, self._recall_prefetch_tags_match)
 
         # For local mode, start the embedded daemon in the background so it
         # doesn't block the chat. Redirect stdout/stderr to a log file to
@@ -991,9 +1006,15 @@ def _run():
                         "bank_id": self._bank_id, "query": query,
                         "budget": self._budget, "max_tokens": self._recall_max_tokens,
                     }
-                    if self._recall_tags:
-                        recall_kwargs["tags"] = self._recall_tags
-                        recall_kwargs["tags_match"] = self._recall_tags_match
+                    prefetch_tags = self._recall_prefetch_tags or self._recall_tags
+                    prefetch_tags_match = (
+                        self._recall_prefetch_tags_match
+                        if self._recall_prefetch_tags
+                        else self._recall_tags_match
+                    )
+                    if prefetch_tags:
+                        recall_kwargs["tags"] = prefetch_tags
+                        recall_kwargs["tags_match"] = prefetch_tags_match
                     if self._recall_types:
                         recall_kwargs["types"] = self._recall_types
                     logger.debug("Prefetch: calling recall (bank=%s, query_len=%d, budget=%s)",
@@ -1075,8 +1096,8 @@ def _build_retain_kwargs(
             kwargs["document_id"] = document_id
         if retain_async is not None:
             kwargs["retain_async"] = retain_async
-        merged_tags = _normalize_retain_tags(self._retain_tags)
-        for tag in _normalize_retain_tags(tags):
+        merged_tags = _normalize_tags(self._retain_tags)
+        for tag in _normalize_tags(tags):
             if tag not in merged_tags:
                 merged_tags.append(tag)
         if merged_tags:
diff --git a/tests/plugins/memory/test_hindsight_provider.py b/tests/plugins/memory/test_hindsight_provider.py
index 5f1290b2f16a..6198246b2ca6 100644
--- a/tests/plugins/memory/test_hindsight_provider.py
+++ b/tests/plugins/memory/test_hindsight_provider.py
@@ -18,7 +18,7 @@
     REFLECT_SCHEMA,
     RETAIN_SCHEMA,
     _load_config,
-    _normalize_retain_tags,
+    _normalize_tags,
     _resolve_bank_id_template,
     _sanitize_bank_segment,
 )
@@ -75,6 +75,12 @@ async def _aretain(
     return client
 
 
+def _run_prefetch(provider, query="test query"):
+    provider.queue_prefetch(query)
+    if provider._prefetch_thread:
+        provider._prefetch_thread.join(timeout=5.0)
+
+
 class _FakeSessionDB:
     def __init__(self, messages=None):
         self._messages = list(messages or [])
@@ -136,16 +142,16 @@ def _make(**overrides):
     return _make
 
 
-def test_normalize_retain_tags_accepts_csv_and_dedupes():
-    assert _normalize_retain_tags("agent:fakeassistantname, source_system:hermes-agent, agent:fakeassistantname") == [
+def test_normalize_tags_accepts_csv_and_dedupes():
+    assert _normalize_tags("agent:fakeassistantname, source_system:hermes-agent, agent:fakeassistantname") == [
         "agent:fakeassistantname",
         "source_system:hermes-agent",
     ]
 
 
-def test_normalize_retain_tags_accepts_json_array_string():
+def test_normalize_tags_accepts_json_array_string():
     value = json.dumps(["agent:fakeassistantname", "source_system:hermes-agent"])
-    assert _normalize_retain_tags(value) == ["agent:fakeassistantname", "source_system:hermes-agent"]
+    assert _normalize_tags(value) == ["agent:fakeassistantname", "source_system:hermes-agent"]
 
 
 # ---------------------------------------------------------------------------
@@ -204,6 +210,8 @@ def test_custom_config_values(self, provider_with_config):
             retain_source="hermes",
             retain_user_prefix="User (fakeusername)",
             retain_assistant_prefix="Assistant (fakeassistantname)",
+            recall_prefetch_tags=["prefetch-tag"],
+            recall_prefetch_tags_match="all_strict",
             recall_tags=["recall-tag"],
             recall_tags_match="all",
             auto_retain=False,
@@ -222,6 +230,8 @@ def test_custom_config_values(self, provider_with_config):
         assert p._retain_source == "hermes"
         assert p._retain_user_prefix == "User (fakeusername)"
         assert p._retain_assistant_prefix == "Assistant (fakeassistantname)"
+        assert p._recall_prefetch_tags == ["prefetch-tag"]
+        assert p._recall_prefetch_tags_match == "all_strict"
         assert p._recall_tags == ["recall-tag"]
         assert p._recall_tags_match == "all"
         assert p._auto_retain is False
@@ -251,6 +261,10 @@ def test_config_from_env_fallback(self, tmp_path, monkeypatch):
         assert cfg["banks"]["hermes"]["bankId"] == "env-bank"
         assert cfg["banks"]["hermes"]["budget"] == "high"
 
+    def test_default_prefetch_tag_config(self, provider):
+        assert provider._recall_prefetch_tags is None
+        assert provider._recall_prefetch_tags_match == "any"
+
 
 class TestPostSetup:
     def test_local_embedded_setup_materializes_profile_env(self, tmp_path, monkeypatch):
@@ -446,6 +460,29 @@ def test_recall_error_handling(self, provider):
         ))
         assert "error" in result
 
+    def test_recall_tool_ignores_prefetch_tag_config(self, provider_with_config):
+        p = provider_with_config(
+            recall_tags=["manual-tag"],
+            recall_tags_match="all",
+            recall_prefetch_tags=["prefetch-tag"],
+            recall_prefetch_tags_match="all_strict",
+        )
+        p._client.arecall = AsyncMock(
+            return_value=SimpleNamespace(
+                results=[
+                    SimpleNamespace(
+                        text="Foreign memory",
+                    )
+                ]
+            )
+        )
+
+        result = json.loads(p.handle_tool_call("hindsight_recall", {"query": "test"}))
+        assert result["result"] == "1. Foreign memory"
+        call_kwargs = p._client.arecall.call_args.kwargs
+        assert call_kwargs["tags"] == ["manual-tag"]
+        assert call_kwargs["tags_match"] == "all"
+
 
 # ---------------------------------------------------------------------------
 # Prefetch tests
@@ -508,9 +545,7 @@ def test_queue_prefetch_passes_recall_params(self, provider_with_config):
             recall_max_tokens=1024,
             recall_types=["world"],
         )
-        p.queue_prefetch("test query")
-        if p._prefetch_thread:
-            p._prefetch_thread.join(timeout=5.0)
+        _run_prefetch(p)
 
         call_kwargs = p._client.arecall.call_args.kwargs
         assert call_kwargs["max_tokens"] == 1024
@@ -518,6 +553,83 @@ def test_queue_prefetch_passes_recall_params(self, provider_with_config):
         assert call_kwargs["tags_match"] == "all"
         assert call_kwargs["types"] == ["world"]
 
+    def test_queue_prefetch_default_config_leaves_results_unfiltered(self, provider):
+        provider._client.arecall = AsyncMock(
+            return_value=SimpleNamespace(
+                results=[
+                    SimpleNamespace(
+                        text="Foreign memory",
+                        metadata={"chat_id": "other-chat", "thread_id": "other-thread"},
+                    ),
+                    SimpleNamespace(text="Unscoped memory"),
+                ]
+            )
+        )
+
+        _run_prefetch(provider)
+
+        call_kwargs = provider._client.arecall.call_args.kwargs
+        assert "tags" not in call_kwargs
+        assert "tags_match" not in call_kwargs
+        assert provider._prefetch_result.splitlines() == ["- Foreign memory", "- Unscoped memory"]
+
+    def test_queue_prefetch_uses_prefetch_tags_when_configured(self, provider_with_config):
+        p = provider_with_config(
+            recall_tags=["fallback-tag"],
+            recall_tags_match="all",
+            recall_prefetch_tags=["prefetch-tag", "team:cli"],
+            recall_prefetch_tags_match="all_strict",
+            recall_max_tokens=1024,
+            recall_types=["world"],
+        )
+        p._client.arecall = AsyncMock(
+            return_value=SimpleNamespace(
+                results=[
+                    SimpleNamespace(text="Tagged memory"),
+                ]
+            )
+        )
+
+        _run_prefetch(p)
+
+        call_kwargs = p._client.arecall.call_args.kwargs
+        assert call_kwargs["tags"] == ["prefetch-tag", "team:cli"]
+        assert call_kwargs["tags_match"] == "all_strict"
+        assert call_kwargs["max_tokens"] == 1024
+        assert call_kwargs["types"] == ["world"]
+        assert p._prefetch_result == "- Tagged memory"
+
+    def test_queue_prefetch_with_prefetch_tags_does_not_post_filter_results(self, provider_with_config):
+        p = provider_with_config(
+            recall_prefetch_tags=["prefetch-tag"],
+            recall_prefetch_tags_match="all",
+        )
+        p._client.arecall = AsyncMock(
+            return_value=SimpleNamespace(
+                results=[
+                    SimpleNamespace(text="Unscoped memory"),
+                    SimpleNamespace(text="Foreign memory", metadata={"chat_id": "other-chat"}),
+                ]
+            )
+        )
+
+        _run_prefetch(p)
+
+        assert p._prefetch_result.splitlines() == ["- Unscoped memory", "- Foreign memory"]
+
+    def test_queue_prefetch_reflect_ignores_prefetch_tags(self, provider_with_config):
+        p = provider_with_config(
+            recall_prefetch_method="reflect",
+            recall_prefetch_tags=["prefetch-tag"],
+            recall_prefetch_tags_match="all_strict",
+        )
+        p._client.areflect = AsyncMock(return_value=SimpleNamespace(text="Synthesized answer"))
+
+        _run_prefetch(p)
+
+        assert p._prefetch_result == "Synthesized answer"
+        p._client.arecall.assert_not_called()
+
 
 # ---------------------------------------------------------------------------
 # sync_turn tests
@@ -786,6 +898,7 @@ def test_schema_has_all_new_fields(self, provider):
             "mode", "api_url", "api_key", "llm_provider", "llm_api_key",
             "llm_model", "bank_id", "bank_id_template", "bank_mission", "bank_retain_mission",
             "recall_budget", "memory_mode", "recall_prefetch_method",
+            "recall_prefetch_tags", "recall_prefetch_tags_match",
             "retain_tags", "retain_source",
             "retain_user_prefix", "retain_assistant_prefix",
             "recall_tags", "recall_tags_match",
diff --git a/website/docs/user-guide/features/memory-providers.md b/website/docs/user-guide/features/memory-providers.md
index afbdac5fca41..e4abb5f0afa5 100644
--- a/website/docs/user-guide/features/memory-providers.md
+++ b/website/docs/user-guide/features/memory-providers.md
@@ -365,6 +365,10 @@ The setup wizard installs dependencies automatically and only installs what's ne
 | `retain_user_prefix` | `User` | Label used before user turns in auto-retained transcripts |
 | `retain_assistant_prefix` | `Assistant` | Label used before assistant turns in auto-retained transcripts |
 | `recall_tags` | — | Tags to filter on recall |
+| `recall_prefetch_tags` | — | Auto-prefetch only: explicit tags to use instead of generic `recall_tags` |
+| `recall_prefetch_tags_match` | `any` | Auto-prefetch only: tag matching mode for `recall_prefetch_tags` |
+
+`recall_prefetch_tags` applies only to automatic prefetch recall. When unset, auto-prefetch falls back to `recall_tags`. It does not affect the manual `hindsight_recall` tool, and it only matches memories that already carry those tags.
 
 See [plugin README](https://github.com/NousResearch/hermes-agent/blob/main/plugins/memory/hindsight/README.md) for the full configuration reference.