diff --git a/agent/config.yml b/agent/config.yml
index 1fa944f..0cde0ac 100644
--- a/agent/config.yml
+++ b/agent/config.yml
@@ -55,10 +55,15 @@ autolearn:
   enabled: true
   autoContinue: true
 compaction:
+  # Server-side compaction always runs on frank-medium: the endpoint is
+  # frank-medium's /v1/chat/completions, and every model in models.yml
+  # carries compactionModel: llamacpp (dankserver)/frank-medium so the
+  # POST body's model id stays valid from any session model.
+  remoteEndpoint: http://192.168.0.235:7770/v1/chat/completions
   methodOrder:
+    - remote
     - shake
     - soft
-    - remote
     - handoff
     - snapcompact
   idleThresholdTokens: 100000
@@ -91,6 +96,12 @@ display:
   showTokenUsage: true
 task:
   showResolvedModelBadge: true
+  # Cap ACTIVE subagents at 2 (a semaphore on concurrency); total
+  # instantiations over a session are uncapped.
+  maxConcurrency: 2
+  agentModelOverrides:
+    # scout/research runs on frank-small at xhigh thinking (the smol role)
+    scout: "@smol"
 snapcompact:
   systemPrompt: none
   toolResults: false
diff --git a/agent/models.yml b/agent/models.yml
index d72a733..47bc926 100644
--- a/agent/models.yml
+++ b/agent/models.yml
@@ -1,4 +1,6 @@
-# Imported from ~/.pi/agent/models.json (pi agent settings)
+# OMP agent model registry — single source of truth (OMP is the primary
+# harness; there is no second copy to keep in sync).
+# Conversion notes from the pi-era import:
 # - apiKey "EMPTY" -> auth: none (keyless local endpoint)
 # - thinkingLevelMap {max: xhigh} -> compat.reasoningEffortMap
 providers:
@@ -57,6 +59,11 @@ providers:
     models:
       - id: frank-large
         name: "frank-large (dankclaw-gpu)"
+        # Server-side compaction always runs on frank-medium (see
+        # compaction.remoteEndpoint in config.yml); this steers the
+        # compaction model candidate so the wire model id stays valid
+        # on the frank-medium endpoint even from a frank-large session.
+        compactionModel: llamacpp (dankserver)/frank-medium
         # halogen-flash-server. The model's native window is 262,144, but this
         # deployment caps a conversation at 200,192 — exactly half of its
         # 400,384-position KV pool, so two 200k conversations are resident at
@@ -65,7 +72,13 @@ providers:
         contextWindow: 200192
         # HALOGEN_MAX_TOKENS_CAP: above this a request is a 400 rather than a
         # silent truncation, and thinking counts against the budget.
-        maxTokens: 65536
+        # maxTokens must also survive the harness compaction gap: OMP compacts at
+        # 70% of contextWindow = 140,134, so a pre-compaction prompt can legally
+        # reach that height. 140,134 + 57,344 = 197,478 <= 200,192, which closes
+        # the dead zone that 400'd a request on 2026-09-17 (prompt 138,898 +
+        # budget 64,000 = 202,898 > 200,192: above the server's fit check, below
+        # the compaction trigger). Any budget > 60,058 reopens the zone.
+        maxTokens: 57344
         reasoning: true
         # Text only as of 2026-09-16: the vision tower moved to frank-small and
         # this server refuses image content parts.
@@ -96,6 +109,10 @@ providers:
     models:
       - id: frank-small
         name: "frank-small (dankclaw-3090)"
+        # See frank-large: server-side compaction always runs on
+        # frank-medium, so the wire model id stays valid from a
+        # frank-small session.
+        compactionModel: llamacpp (dankserver)/frank-medium
         # Dankclaw's RTX 3090: mainline llama.cpp serving
         # HauhauCS/Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced at Q4_K_M.
         # The MTP head ships as a separate assistant sidecar