diff --git a/nix/host-modules/swarm-bao.nix b/nix/host-modules/swarm-bao.nix
index f5e632c9..b43ab84e 100644
--- a/nix/host-modules/swarm-bao.nix
+++ b/nix/host-modules/swarm-bao.nix
@@ -195,9 +195,9 @@ let
 
   # Metrics get their own listener rather than a flag on the one above, and
   # that follows from what a scraper can express: `swarm.otel.scrapeTargets`
-  # is `host:port` with no scheme and no credential, while the API listener is
-  # TLS and — once a client CA is set — demands a client certificate. The
-  # collector cannot reach it at all.
+  # carries no scheme and no credential, while the API listener is TLS and —
+  # once a client CA is set — demands a client certificate. The collector
+  # cannot reach it at all.
   #
   # `metrics_only` narrows this one to the metrics path (every other path 404s)
   # and the unauthenticated access is confined to loopback. **Deliberately
@@ -576,7 +576,11 @@ in
       ];
 
       services.hyperhive.swarm.otel.scrapeTargets = lib.mkIf scrapeHere {
-        bao = "127.0.0.1:${toString baoDeploy.metricsPort}";
+        # Path and query, not just `host:port`: openbao serves no `/metrics`
+        # at all, and `/v1/sys/metrics` answers JSON unless the format is
+        # asked for. A scrape of the default path 404s, which reads as a
+        # dead exporter rather than a wrong address.
+        bao = "127.0.0.1:${toString baoDeploy.metricsPort}/v1/sys/metrics?format=prometheus";
       };
     })
 
diff --git a/nix/host-modules/swarm-grafana.nix b/nix/host-modules/swarm-grafana.nix
index 4c4b159b..26c0d8ae 100644
--- a/nix/host-modules/swarm-grafana.nix
+++ b/nix/host-modules/swarm-grafana.nix
@@ -61,6 +61,7 @@ let
       [
         "agents.json"
         "authelia.json"
+        "bao.json"
         "claude-usage.json"
         "forge.json"
         "logstore.json"
diff --git a/nix/host-modules/swarm-grafana/dashboards/bao.json b/nix/host-modules/swarm-grafana/dashboards/bao.json
new file mode 100644
index 00000000..931f5308
--- /dev/null
+++ b/nix/host-modules/swarm-grafana/dashboards/bao.json
@@ -0,0 +1,569 @@
+{
+  "title": "hyperhive · openbao",
+  "uid": "hyperhive-bao",
+  "description": "The swarm's secret store. Two things to know before editing: the metric prefix is vault_, not openbao_ — the fork kept upstream's names, and a panel built on the name you would guess renders perfectly and shows nothing. And a SEALED store emits almost nothing: only the top row works before an operator has run `bao operator init`, everything below it is empty until then and that is the expected reading, not a broken query. Targets are labelled service.name, NOT job; there is no job label in this store. Counters are counts over the dashboard range rather than per-second rates, because a store this quiet rates to exactly zero.",
+  "editable": true,
+  "refresh": "1m",
+  "schemaVersion": 39,
+  "tags": [
+    "hyperhive",
+    "swarm",
+    "bao",
+    "secrets"
+  ],
+  "time": {
+    "from": "now-24h",
+    "to": "now"
+  },
+  "timezone": "utc",
+  "version": 1,
+  "templating": {
+    "list": []
+  },
+  "panels": [
+    {
+      "id": 1,
+      "type": "stat",
+      "title": "Seal state",
+      "description": "SEALED means the store is up but refusing every read — the state it boots into, and the state it returns to on restart unless auto-unseal is configured. The max() is load-bearing: once unsealed there are two vault_core_unsealed series and the cluster=\"\" one stays 0 forever, so a bare selector reads SEALED on a healthy store.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 0,
+        "y": 0
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "max(vault_core_unsealed{\"service.name\"=\"bao\"})",
+          "instant": true,
+          "legendFormat": "seal"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 0,
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [
+            {
+              "type": "value",
+              "options": {
+                "0": {
+                  "text": "SEALED",
+                  "color": "red",
+                  "index": 0
+                },
+                "1": {
+                  "text": "unsealed",
+                  "color": "green",
+                  "index": 1
+                }
+              }
+            }
+          ],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "red",
+                "value": null
+              },
+              {
+                "color": "green",
+                "value": 1
+              }
+            ]
+          }
+        },
+        "overrides": []
+      },
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "textMode": "value",
+        "justifyMode": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        }
+      }
+    },
+    {
+      "id": 2,
+      "type": "stat",
+      "title": "Uptime",
+      "description": "Time since the openbao process started. Read it beside the seal state: a restart re-seals the store, so a short uptime and a SEALED panel are one event rather than two problems.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 6,
+        "y": 0
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "time() - process_start_time_seconds{\"service.name\"=\"bao\"}",
+          "instant": true,
+          "legendFormat": "uptime"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "decimals": 0,
+          "color": {
+            "mode": "fixed",
+            "fixedColor": "text"
+          },
+          "mappings": []
+        },
+        "overrides": []
+      },
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "textMode": "value",
+        "justifyMode": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        }
+      }
+    },
+    {
+      "id": 3,
+      "type": "stat",
+      "title": "Goroutines",
+      "description": "A Go runtime metric, so it is present even while sealed — which makes it the panel that distinguishes a store with nothing to say from a scrape that is not arriving at all. No data here means the collector is not reaching the metrics listener.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 12,
+        "y": 0
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "go_goroutines{\"service.name\"=\"bao\"}",
+          "instant": true,
+          "legendFormat": "goroutines"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 0,
+          "color": {
+            "mode": "fixed",
+            "fixedColor": "text"
+          },
+          "mappings": []
+        },
+        "overrides": []
+      },
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "textMode": "value",
+        "justifyMode": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        }
+      }
+    },
+    {
+      "id": 4,
+      "type": "stat",
+      "title": "Requests (range)",
+      "description": "vault_core_handle_request_count over the dashboard range — every API call the store served, unseal attempts included. The summary's _count series, not the summary itself. Zero on a sealed store, and zero is the honest answer there rather than an absent panel.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 18,
+        "y": 0
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "sum(increase(vault_core_handle_request_count{\"service.name\"=\"bao\"}[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "requests"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 0,
+          "color": {
+            "mode": "fixed",
+            "fixedColor": "text"
+          },
+          "mappings": []
+        },
+        "overrides": []
+      },
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "textMode": "value",
+        "justifyMode": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        }
+      }
+    },
+    {
+      "id": 10,
+      "type": "stat",
+      "title": "Seal operations (range)",
+      "description": "Encrypt and decrypt calls against the seal wrapper — on this swarm a TPM through PKCS#11, so each one is a round trip to hardware. These are counters with no _total suffix. A climbing decrypt count with a flat request count means something is re-reading the barrier rather than serving clients.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 12,
+        "x": 0,
+        "y": 4
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "sum(increase(vault_seal_encrypt{\"service.name\"=\"bao\"}[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "encrypt"
+        },
+        {
+          "refId": "B",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "sum(increase(vault_seal_decrypt{\"service.name\"=\"bao\"}[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "decrypt"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 0,
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": []
+        },
+        "overrides": []
+      },
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "textMode": "value_and_name",
+        "justifyMode": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        }
+      }
+    },
+    {
+      "id": 11,
+      "type": "stat",
+      "title": "Raft peers",
+      "description": "Voting members of the store's raft cluster. This swarm runs a single node, so 1 is the expected value and anything else means a peer was added — or that this node no longer sees itself as part of the cluster it stores data for.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 12,
+        "x": 12,
+        "y": 4
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "vault_raft_peers{\"service.name\"=\"bao\"}",
+          "instant": true,
+          "legendFormat": "peers"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 0,
+          "color": {
+            "mode": "fixed",
+            "fixedColor": "text"
+          },
+          "mappings": []
+        },
+        "overrides": []
+      },
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "textMode": "value",
+        "justifyMode": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        }
+      }
+    },
+    {
+      "id": 20,
+      "type": "timeseries",
+      "title": "Request latency",
+      "description": "Quantiles openbao computes itself — a summary, so the store picked the buckets and a histogram_quantile() is neither needed nor possible here. Units are MILLISECONDS: vault's timers report ms, and reading them as seconds understates the store by three orders of magnitude.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 8
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "vault_core_handle_request{\"service.name\"=\"bao\",quantile=\"0.99\"}",
+          "legendFormat": "p99"
+        },
+        {
+          "refId": "B",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "vault_core_handle_request{\"service.name\"=\"bao\",quantile=\"0.5\"}",
+          "legendFormat": "p50"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "ms",
+          "decimals": 1,
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineWidth": 1,
+            "fillOpacity": 0,
+            "showPoints": "never",
+            "spanNulls": false,
+            "axisSoftMin": 0
+          }
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true,
+          "calcs": []
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      }
+    },
+    {
+      "id": 21,
+      "type": "timeseries",
+      "title": "Leases",
+      "description": "Outstanding leases, and the irrevocable ones beside them. The second line should sit at zero: an irrevocable lease is one the store gave up trying to revoke, so a credential somewhere outlives what issued it and no amount of waiting will clean it up.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 8
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "vault_expire_num_leases{\"service.name\"=\"bao\"}",
+          "legendFormat": "leases"
+        },
+        {
+          "refId": "B",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "vault_expire_num_irrevocable_leases{\"service.name\"=\"bao\"}",
+          "legendFormat": "irrevocable"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 0,
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineWidth": 1,
+            "fillOpacity": 10,
+            "showPoints": "never",
+            "spanNulls": false,
+            "axisSoftMin": 0
+          }
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true,
+          "calcs": []
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      }
+    },
+    {
+      "id": 30,
+      "type": "timeseries",
+      "title": "Heap in use",
+      "description": "go_memstats_heap_inuse_bytes. Present while sealed, so it is a live line even on a store that is serving nothing — a sawtooth is normal Go behaviour, a monotonic climb on an idle store is not.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "@datasourceUid@"
+      },
+      "gridPos": {
+        "h": 6,
+        "w": 24,
+        "x": 0,
+        "y": 16
+      },
+      "targets": [
+        {
+          "refId": "A",
+          "datasource": {
+            "type": "prometheus",
+            "uid": "@datasourceUid@"
+          },
+          "expr": "go_memstats_heap_inuse_bytes{\"service.name\"=\"bao\"}",
+          "legendFormat": "heap"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes",
+          "decimals": 1,
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineWidth": 1,
+            "fillOpacity": 10,
+            "showPoints": "never",
+            "spanNulls": false,
+            "axisSoftMin": 0
+          }
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true,
+          "calcs": []
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      }
+    }
+  ]
+}
diff --git a/nix/host-modules/swarm-otel.nix b/nix/host-modules/swarm-otel.nix
index e4e40e17..2f5cdbea 100644
--- a/nix/host-modules/swarm-otel.nix
+++ b/nix/host-modules/swarm-otel.nix
@@ -91,6 +91,46 @@ let
   # error rather than a warning nobody reads.
   parsePublished = url: builtins.match "https://([^/]+)(/.*)" url;
 
+  # A loopback target may carry a path and a query as well as `host:port`,
+  # because not every exporter serves `/metrics`: openbao 404s there and
+  # answers on `/v1/sys/metrics?format=prometheus`. Split for the same reason
+  # the published one is — one declaration, three prometheus fields.
+  #
+  # ⚠️ The query may not ride along in `metrics_path`: prometheus percent-
+  # encodes the `?`, so the request goes to a path that does not exist. It has
+  # to become `params`, whose values are lists.
+  #
+  # `null` when the shape is wrong, which the assertion below reports by name.
+  parseScrape = t: builtins.match "([^/?]+)(/[^?]*)?(\\?(.+))?" t;
+
+  queryParams =
+    query:
+    lib.listToAttrs (
+      map (
+        kv:
+        let
+          parts = lib.splitString "=" kv;
+        in
+        lib.nameValuePair (lib.head parts) [ (lib.concatStringsSep "=" (lib.tail parts)) ]
+      ) (lib.splitString "&" query)
+    );
+
+  # Both optional fields are omitted rather than defaulted, so a plain
+  # `host:port` renders the config it rendered before this grammar existed.
+  loopbackScrapeConfig =
+    job: target:
+    let
+      parts = parseScrape target;
+      path = lib.elemAt parts 1;
+      query = lib.elemAt parts 3;
+    in
+    {
+      job_name = job;
+      static_configs = [ { targets = [ (lib.elemAt parts 0) ]; } ];
+    }
+    // lib.optionalAttrs (path != null) { metrics_path = path; }
+    // lib.optionalAttrs (query != null) { params = queryParams query; };
+
   # The client secret takes three names, and the reason is `DynamicUser`.
   #
   # Upstream's collector unit runs with `DynamicUser = true`, and the
@@ -332,6 +372,12 @@ in
         Prometheus exposition endpoints this collector scrapes, as
         `<job name> = "<host>:<port>"`.
 
+        A path and query may follow the port —
+        `"127.0.0.1:8202/v1/sys/metrics?format=prometheus"` — for an
+        exporter that does not serve `/metrics`. Both are optional and
+        omitted when absent, so a bare `host:port` is scraped exactly as
+        before.
+
         **A service declares its own entry, from its own module, under its
         own `enable`.** Do not assemble the list here: an entry then exists
         only where the service that named it runs, so a target is never
@@ -711,6 +757,23 @@ in
           token, and plain http would put it on the wire in clear text.
         '';
       }
+      {
+        # Same reason the published one is asserted: a value the grammar
+        # rejects surfaces as `attempt to call elemAt on null` from inside
+        # the renderer, naming neither the option nor the offending value.
+        assertion = lib.all (t: parseScrape t != null) (lib.attrValues cfg.scrapeTargets);
+        message = ''
+          services.hyperhive.swarm.otel.scrapeTargets has ${
+            lib.concatMapStringsSep ", " (kv: "${kv.name} = \"${kv.value}\"") (
+              lib.filter (kv: parseScrape kv.value == null) (lib.attrsToList cfg.scrapeTargets)
+            )
+          }, which is not of the form <host>:<port>[/<path>][?<k>=<v>].
+
+          A trailing `?` with nothing after it is the usual cause. This
+          option is loopback-and-unauthenticated by contract, so it takes
+          no scheme and no credential.
+        '';
+      }
       {
         # A job name used by BOTH scrape options. Within one option this
         # cannot happen — the module system refuses two definitions of the
@@ -927,10 +990,7 @@ in
                 # and deploys perfectly while adding no data.
                 // lib.optionalAttrs (cfg.scrapeTargets != { } || cfg.publishedScrapeTargets != { }) {
                   prometheus.config.scrape_configs =
-                    lib.mapAttrsToList (job: target: {
-                      job_name = job;
-                      static_configs = [ { targets = [ target ]; } ];
-                    }) cfg.scrapeTargets
+                    lib.mapAttrsToList loopbackScrapeConfig cfg.scrapeTargets
                     # ⚠️ `++`, so the two kinds of target land in ONE list —
                     # which is exactly why a job name may not appear in both
                     # options. A list concatenation does not resolve a
diff --git a/nix/module-eval.nix b/nix/module-eval.nix
index 4028a376..b3cb38ab 100644
--- a/nix/module-eval.nix
+++ b/nix/module-eval.nix
@@ -118,6 +118,10 @@ let
   baoWithCollector = hive {
     deploy.bao.enable = true;
     deploy.swarm-otel.enable = true;
+    # A second job declared as bare `host:port`, so the pair of cases below
+    # reads one rendered scrape list: the store's entry carries a path, this
+    # one carries none.
+    swarm.otel.scrapeTargets.plain = "127.0.0.1:9999";
   };
   baoNoCollector = hive {
     deploy.bao.enable = true;
@@ -133,6 +137,14 @@ let
   # look there rather than at the host's service set.
   baoUnits = machine: machine.containers.swarm-bao.config.systemd.services;
 
+  # The scrape list prometheus is handed, not the option a service declared:
+  # the address, the path and the query are one string on the way in and three
+  # fields on the way out, and only the second shape is what gets requested.
+  scrapeJob =
+    machine: job:
+    lib.findFirst (c: c.job_name == job) null
+      machine.containers.swarm-otel.config.services.opentelemetry-collector.settings.receivers.prometheus.config.scrape_configs;
+
   # A priority collision is a property of the *option*, not
   # of the merged value's interior — nix throws the moment the value is
   # demanded at all, so `seq`-ing each `serviceConfig` value to WHNF is
@@ -385,6 +397,28 @@ let
         in
         s.listener ? metrics && (s.telemetry.prometheus_retention_time or "0s") != "0s";
     }
+    {
+      # openbao serves no `/metrics` at all, so a scrape of the default path
+      # 404s: the store looks like a dead exporter, and every panel built on
+      # it renders empty rather than erroring.
+      name = "the store's scrape asks for the path openbao serves";
+      ok =
+        let
+          j = scrapeJob baoWithCollector "bao";
+        in
+        (j.metrics_path or "") == "/v1/sys/metrics" && (j.params.format or [ ]) == [ "prometheus" ];
+    }
+    {
+      # Presence control for the case above: both fields are omitted rather
+      # than defaulted, so a target declared as bare `host:port` renders what
+      # it rendered before the path grammar existed.
+      name = "a target with no path renders neither metrics_path nor params";
+      ok =
+        let
+          j = scrapeJob baoWithCollector "plain";
+        in
+        j != null && !(j ? metrics_path) && !(j ? params);
+    }
     {
       # Absence arm. Unauthenticated by design, so it must not exist where
       # nothing reads it.