diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2d80151..72deba2 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -13,6 +13,7 @@ The format is intentionally lightweight and human-readable. Group entries by rel
 - Added optional `image` provider metadata (`max_outputs`, `max_side_px`, `supported_sizes`) so image-capable providers can be ranked against `n` and `size`
 - Added top-level capability coverage to `GET /health` plus `GET /api/providers` for filtered provider inventory and dashboard coverage views
 - Added shared request validation for image-generation, image-editing, and image-route preview payloads so invalid `size`, `n`, and scalar fields fail fast before provider calls
+- Added optional `image.policy_tags` plus request-side image-policy hints so image routing can prefer providers tagged for `quality`, `cost`, `balanced`, `batch`, or `editing`
 
 ## v0.5.0 - 2026-03-12
 
diff --git a/README.md b/README.md
index 5e169dd..eefd4b2 100644
--- a/README.md
+++ b/README.md
@@ -220,6 +220,7 @@ OpenAI-compatible image generation endpoint.
 - `model: "auto"` selects the best loaded provider with `capabilities.image_generation: true`
 - `model: "<provider-id>"` routes directly to a loaded image-capable provider
 - validates `prompt`, `n`, and `size` before any provider call
+- optional image-policy hints can be passed via `metadata.image_policy` or `X-FoundryGate-Image-Policy`
 
 ```bash
 curl -fsS http://127.0.0.1:8090/v1/images/generations \
@@ -240,6 +241,7 @@ OpenAI-compatible image editing endpoint.
 - `model: "auto"` selects the best loaded provider with `capabilities.image_editing: true`
 - `model: "<provider-id>"` routes directly to a loaded image-edit-capable provider
 - validates scalar fields such as `prompt`, `n`, and `size` before any provider call
+- optional image-policy hints can be passed via form field `image_policy`, `metadata.image_policy`, or `X-FoundryGate-Image-Policy`
 
 ```bash
 curl -fsS http://127.0.0.1:8090/v1/images/edits \
@@ -296,6 +298,8 @@ If request hooks are enabled, `POST /api/route` also shows the applied hook name
 
 `GET /api/providers` returns the current provider inventory, including capability flags and optional image metadata such as `max_outputs`, `max_side_px`, and `supported_sizes`.
 
+For image-capable providers, `image.policy_tags` can be used as lightweight presets such as `quality`, `cost`, `balanced`, `batch`, or `editing`. When a request carries `metadata.image_policy` or `X-FoundryGate-Image-Policy`, routing prefers providers whose `image.policy_tags` match that hint.
+
 `GET /api/stats`, `GET /api/recent`, and `GET /api/traces` also accept optional `provider`, `modality`, `client_profile`, `client_tag`, `layer`, and `success` filters. The built-in dashboard uses the same filtered endpoints.
 
 `GET /api/traces` returns recent enriched routing records from the metrics store, including requested model, modality, resolved client profile, client tag, decision reason, confidence, and attempt order.
diff --git a/config.yaml b/config.yaml
index a2bbefc..c9703e9 100644
--- a/config.yaml
+++ b/config.yaml
@@ -46,6 +46,11 @@ server:
 #     read_s     : read/response timeout (default: 120)
 #   pricing      : USD per 1 000 000 tokens (input / output / cache_read)
 #                  Used for cost tracking only – not enforced.
+#   image        : optional image-routing metadata for `contract: image-provider`
+#     max_outputs    : maximum supported `n`
+#     max_side_px    : largest supported edge size
+#     supported_sizes: optional exact allowed size strings such as 1024x1024
+#     policy_tags    : optional routing tags such as quality | cost | balanced
 #
 # Sections
 # ─────────
@@ -190,6 +195,11 @@ providers:
   #  capabilities:
   #    # image_generation is enabled automatically by the contract
   #    image_editing: true
+  #  image:
+  #    max_outputs: 4
+  #    max_side_px: 2048
+  #    supported_sizes: ["1024x1024", "2048x2048"]
+  #    policy_tags: ["quality", "editing", "batch"]
 
   # ── Anthropic ───────────────────────────────────────────────────────────
   # Auth: ANTHROPIC_API_KEY (or setup-token)
diff --git a/foundrygate/config.py b/foundrygate/config.py
index f3511a8..bf7368e 100644
--- a/foundrygate/config.py
+++ b/foundrygate/config.py
@@ -347,6 +347,24 @@ def _normalize_provider_image(name: str, cfg: dict[str, Any]) -> dict[str, Any]:
     if normalized_sizes:
         image["supported_sizes"] = normalized_sizes
 
+    policy_tags = raw.get("policy_tags", [])
+    if policy_tags in (None, ""):
+        policy_tags = []
+    if isinstance(policy_tags, str):
+        policy_tags = [policy_tags]
+    if not isinstance(policy_tags, list):
+        raise ConfigError(f"Provider '{name}' field 'image.policy_tags' must be a list")
+
+    normalized_tags = []
+    for value in policy_tags:
+        if not isinstance(value, str) or not value.strip():
+            raise ConfigError(
+                f"Provider '{name}' field 'image.policy_tags' must contain non-empty strings"
+            )
+        normalized_tags.append(value.strip().lower())
+    if normalized_tags:
+        image["policy_tags"] = normalized_tags
+
     return image
 
 
diff --git a/foundrygate/main.py b/foundrygate/main.py
index 4cc963d..53a593b 100644
--- a/foundrygate/main.py
+++ b/foundrygate/main.py
@@ -276,6 +276,7 @@ def _estimate_image_request_dimensions(body: dict[str, Any], *, capability: str)
         "prompt_chars": len(str(body.get("prompt") or "")),
         "requested_size": body.get("size") or "",
         "requested_outputs": body.get("n") if isinstance(body.get("n"), int) else 1,
+        "image_policy": _collect_request_image_policy(body),
         "capability": capability,
     }
 
@@ -288,12 +289,25 @@ def _collect_request_cache_preference(body: dict[str, Any]) -> str:
     return ""
 
 
+def _collect_request_image_policy(body: dict[str, Any]) -> str:
+    """Return one optional image-policy hint from request data."""
+    if isinstance(body.get("image_policy"), str) and body["image_policy"].strip():
+        return body["image_policy"].strip().lower()
+    metadata = body.get("metadata") if isinstance(body.get("metadata"), dict) else {}
+    if isinstance(metadata.get("image_policy"), str) and metadata["image_policy"].strip():
+        return metadata["image_policy"].strip().lower()
+    return ""
+
+
 def _merge_routing_context_headers(headers: dict[str, str], body: dict[str, Any]) -> dict[str, str]:
     """Return routing headers plus request-body dimension hints."""
     merged = dict(headers)
     cache_preference = _collect_request_cache_preference(body)
-    if cache_preference:
+    if cache_preference and "x-foundrygate-cache" not in merged:
         merged["x-foundrygate-cache"] = cache_preference
+    image_policy = _collect_request_image_policy(body)
+    if image_policy and "x-foundrygate-image-policy" not in merged:
+        merged["x-foundrygate-image-policy"] = image_policy
     return merged
 
 
@@ -464,7 +478,17 @@ def _normalize_image_request_body(body: dict[str, Any], *, capability: str) -> d
     if metadata is not None:
         if not isinstance(metadata, dict):
             raise ValueError("Field 'metadata' must be an object when provided")
-        normalized["metadata"] = metadata
+        normalized["metadata"] = dict(metadata)
+
+    image_policy = body.get("image_policy")
+    if image_policy in (None, "") and isinstance(normalized.get("metadata"), dict):
+        image_policy = normalized["metadata"].get("image_policy")
+    if image_policy not in (None, ""):
+        if not isinstance(image_policy, str) or not image_policy.strip():
+            raise ValueError("Field 'image_policy' must be a non-empty string when provided")
+        cleaned_policy = image_policy.strip().lower()
+        normalized["image_policy"] = cleaned_policy
+        normalized.setdefault("metadata", {})["image_policy"] = cleaned_policy
 
     return normalized
 
@@ -503,6 +527,7 @@ async def _resolve_image_route_preview(
     """Resolve one image-generation request without calling a provider."""
     body, hook_state = await _apply_request_hooks(body, headers)
     body = _normalize_image_request_body(body, capability=capability)
+    headers = _merge_routing_context_headers(headers, body)
     prompt = body["prompt"]
 
     model_requested = str(body.get("model", "auto"))
diff --git a/foundrygate/router.py b/foundrygate/router.py
index 10ee657..58a57fb 100644
--- a/foundrygate/router.py
+++ b/foundrygate/router.py
@@ -197,6 +197,7 @@ async def route(
             requested_image_outputs=1,
             requested_image_side_px=0,
             requested_image_size="",
+            requested_image_policy="",
             required_capability="",
             cache_preference=(headers or {}).get("x-foundrygate-cache", "").strip().lower(),
             model_requested=model_requested.lower().strip(),
@@ -289,6 +290,9 @@ def route_capability_request(
             requested_image_outputs=requested_outputs or 1,
             requested_image_side_px=_parse_image_size_max_side(requested_size),
             requested_image_size=requested_size.strip().lower() if requested_size else "",
+            requested_image_policy=(
+                (headers or {}).get("x-foundrygate-image-policy", "").strip().lower()
+            ),
             required_capability=capability,
             cache_preference=(headers or {}).get("x-foundrygate-cache", "").strip().lower(),
             model_requested=model_requested.lower().strip(),
@@ -662,9 +666,12 @@ def _provider_dimension_details(
             )
 
         image_score = 0
+        image_policy_score = 0
         image_outputs_fit = True
         image_size_fit = True
         image_supported_size = True
+        image_policy_match = not bool(ctx.requested_image_policy)
+        image_policy_tags = image_cfg.get("policy_tags", [])
         if ctx.required_capability in {"image_generation", "image_editing"}:
             max_outputs = int(image_cfg.get("max_outputs") or 0)
             max_side_px = int(image_cfg.get("max_side_px") or 0)
@@ -694,6 +701,12 @@ def _provider_dimension_details(
             elif ctx.requested_image_size:
                 image_score += 1
 
+            if ctx.requested_image_policy:
+                image_policy_match = ctx.requested_image_policy in image_policy_tags
+                image_policy_score = 12 if image_policy_match else 0
+            elif image_policy_tags:
+                image_policy_score = 1
+
         fit = self._provider_fits_request_dimensions(name, provider, ctx)
         score_total = (
             health_score
@@ -705,6 +718,7 @@ def _provider_dimension_details(
             + input_score
             + output_score
             + image_score
+            + image_policy_score
         )
         return {
             "fit": fit,
@@ -718,6 +732,7 @@ def _provider_dimension_details(
             "input_score": input_score,
             "output_score": output_score,
             "image_score": image_score,
+            "image_policy_score": image_policy_score,
             "headroom": headroom,
             "context_ratio": round(context_ratio, 3),
             "input_ratio": round(input_ratio, 3),
@@ -725,6 +740,9 @@ def _provider_dimension_details(
             "image_outputs_fit": image_outputs_fit,
             "image_size_fit": image_size_fit,
             "image_supported_size": image_supported_size,
+            "image_policy_match": image_policy_match,
+            "requested_image_policy": ctx.requested_image_policy,
+            "image_policy_tags": image_policy_tags,
             "requested_image_outputs": ctx.requested_image_outputs,
             "requested_image_size": ctx.requested_image_size,
             "max_image_outputs": image_cfg.get("max_outputs"),
@@ -1015,6 +1033,7 @@ class _RoutingContext:
         "requested_image_outputs",
         "requested_image_side_px",
         "requested_image_size",
+        "requested_image_policy",
         "required_capability",
         "cache_preference",
         "model_requested",
diff --git a/tests/test_capabilities.py b/tests/test_capabilities.py
index 001c7d0..d0a04ed 100644
--- a/tests/test_capabilities.py
+++ b/tests/test_capabilities.py
@@ -241,3 +241,24 @@ def test_image_provider_contract_rejects_non_openai_backend(tmp_path):
 
     with pytest.raises(ConfigError, match="image-provider"):
         load_config(path)
+
+
+def test_image_provider_policy_tags_are_normalized(tmp_path):
+    path = _write_config(
+        tmp_path,
+        (
+            "  image-cloud:\n"
+            "    contract: image-provider\n"
+            "    backend: openai-compat\n"
+            '    base_url: "https://api.example.com/v1"\n'
+            '    api_key: "secret"\n'
+            '    model: "gpt-image-1"\n'
+            "    image:\n"
+            '      policy_tags: ["Quality", " editing "]\n'
+        ),
+    )
+
+    cfg = load_config(path)
+    provider = cfg.provider("image-cloud")
+
+    assert provider["image"]["policy_tags"] == ["quality", "editing"]
diff --git a/tests/test_route_introspection.py b/tests/test_route_introspection.py
index 9789358..d52d57e 100644
--- a/tests/test_route_introspection.py
+++ b/tests/test_route_introspection.py
@@ -168,6 +168,7 @@ def preview_config(tmp_path, monkeypatch):
       max_outputs: 1
       max_side_px: 1024
       supported_sizes: ["1024x1024"]
+      policy_tags: ["balanced", "cost", "editing"]
   image-large:
     contract: image-provider
     backend: openai-compat
@@ -181,6 +182,7 @@ def preview_config(tmp_path, monkeypatch):
       max_outputs: 4
       max_side_px: 2048
       supported_sizes: ["1024x1024", "2048x2048"]
+      policy_tags: ["quality", "batch"]
 client_profiles:
   enabled: true
   default: generic
@@ -236,6 +238,7 @@ def preview_config(tmp_path, monkeypatch):
                     "max_outputs": 1,
                     "max_side_px": 1024,
                     "supported_sizes": ["1024x1024"],
+                    "policy_tags": ["balanced", "cost", "editing"],
                 },
             ),
             "image-large": _ProviderStub(
@@ -254,6 +257,7 @@ def preview_config(tmp_path, monkeypatch):
                     "max_outputs": 4,
                     "max_side_px": 2048,
                     "supported_sizes": ["1024x1024", "2048x2048"],
+                    "policy_tags": ["quality", "batch"],
                 },
             ),
         },
@@ -414,6 +418,47 @@ async def test_image_route_preview_prefers_provider_that_fits_size_and_count(
         assert ranking[0]["image_size_fit"] is True
         assert ranking[0]["image_outputs_fit"] is True
 
+    @pytest.mark.asyncio
+    async def test_image_route_preview_prefers_matching_policy_tag(self, preview_config):
+        response = await preview_image_route(
+            _json_request(
+                "/api/route/image",
+                {
+                    "model": "auto",
+                    "capability": "image_generation",
+                    "prompt": "Create a polished product render.",
+                    "size": "1024x1024",
+                    "metadata": {"image_policy": "quality"},
+                },
+            )
+        )
+
+        assert response["effective_request"]["image_policy"] == "quality"
+        assert response["decision"]["provider"] == "image-large"
+        ranking = response["decision"]["details"]["candidate_ranking"]
+        assert ranking[0]["provider"] == "image-large"
+        assert ranking[0]["image_policy_match"] is True
+        assert ranking[0]["requested_image_policy"] == "quality"
+
+    @pytest.mark.asyncio
+    async def test_image_route_preview_header_policy_overrides_metadata(self, preview_config):
+        response = await preview_image_route(
+            _json_request(
+                "/api/route/image",
+                {
+                    "model": "auto",
+                    "capability": "image_generation",
+                    "prompt": "Create a cheap concept sketch.",
+                    "size": "1024x1024",
+                    "metadata": {"image_policy": "quality"},
+                },
+                headers={"x-foundrygate-image-policy": "cost"},
+            )
+        )
+
+        assert response["routing_headers"]["x-foundrygate-image-policy"] == "cost"
+        assert response["decision"]["provider"] == "image-cloud"
+
     def test_extract_image_edit_request_fields_requires_prompt(self):
         with pytest.raises(ValueError, match="non-empty 'prompt'"):
             _extract_image_edit_request_fields({"model": "auto"})
@@ -427,6 +472,7 @@ def test_extract_image_edit_request_fields_parses_scalars(self):
                 "size": "1024x1024",
                 "response_format": "b64_json",
                 "user": "tester",
+                "image_policy": "editing",
             }
         )
 
@@ -436,6 +482,8 @@ def test_extract_image_edit_request_fields_parses_scalars(self):
         assert payload["size"] == "1024x1024"
         assert payload["response_format"] == "b64_json"
         assert payload["user"] == "tester"
+        assert payload["image_policy"] == "editing"
+        assert payload["metadata"]["image_policy"] == "editing"
 
     def test_normalize_image_request_body_validates_size_and_n(self):
         payload = _normalize_image_request_body(