diff --git a/lib/crates/fabro-agent/tests/it/parity_matrix.rs b/lib/crates/fabro-agent/tests/it/parity_matrix.rs
index 808d8b71d..6b60d1cf7 100644
--- a/lib/crates/fabro-agent/tests/it/parity_matrix.rs
+++ b/lib/crates/fabro-agent/tests/it/parity_matrix.rs
@@ -17,6 +17,7 @@ use fabro_auth::EnvCredentialSource;
 use fabro_llm::client::Client;
 use fabro_llm::provider::ProviderAdapter;
 use fabro_llm::providers::{OpenAiAdapter, OpenAiCompatibleAdapter};
+use fabro_model::catalog::{LlmCatalogSettings, ProviderCatalogSettings};
 use fabro_model::{Catalog, ModelHandle, ProviderId};
 use fabro_test::{TwinScenario, TwinScenarios, TwinToolCall, twin_openai};
 use tokio::sync::Mutex as AsyncMutex;
@@ -184,7 +185,20 @@ fn make_openai_compatible_twin_session(
     twin: &OpenAiTwinOptions,
 ) -> Session {
     let client = make_openai_compatible_twin_client(&provider, twin);
-    let catalog = Arc::new(Catalog::from_builtin().expect("default catalog should build"));
+    // LiteLLM is opt-in in the built-in catalog. Enable the provider in this
+    // twin fixture so the profile can resolve the same OpenAI-compatible
+    // codec that the manually registered adapter uses.
+    let mut settings = LlmCatalogSettings::default();
+    settings
+        .providers
+        .insert(provider.to_string(), ProviderCatalogSettings {
+            enabled: Some(true),
+            ..ProviderCatalogSettings::default()
+        });
+    let catalog = Arc::new(
+        Catalog::from_builtin_with_overrides(&settings)
+            .expect("OpenAI-compatible twin catalog should build"),
+    );
     let profile: Arc<dyn AgentProfile> = Arc::new(
         OpenAiProfile::new(model)
             .with_provider_id(provider)
diff --git a/test/twin/openai/docs/compatibility-matrix.md b/test/twin/openai/docs/compatibility-matrix.md
index 6c6588a2f..6fad3d925 100644
--- a/test/twin/openai/docs/compatibility-matrix.md
+++ b/test/twin/openai/docs/compatibility-matrix.md
@@ -30,6 +30,7 @@ Supported `/v1/responses` fields:
 Supported `/v1/chat/completions` fields:
 
 - bearer auth
+- `max_tokens`
 - `stream`
 - `tools`
 - `tool_choice`
diff --git a/test/twin/openai/src/openai/models.rs b/test/twin/openai/src/openai/models.rs
index 06d99e0e7..abd2e47e7 100644
--- a/test/twin/openai/src/openai/models.rs
+++ b/test/twin/openai/src/openai/models.rs
@@ -531,6 +531,7 @@ pub fn normalize_whitespace(input: &str) -> String {
 pub struct ChatCompletionsRequest {
     pub model:           String,
     pub messages:        Vec<ChatMessage>,
+    pub max_tokens:      Option<i64>,
     #[serde(default)]
     pub stream:          bool,
     pub tools:           Option<Vec<Value>>,
diff --git a/test/twin/openai/tests/chat_completions_contract.rs b/test/twin/openai/tests/chat_completions_contract.rs
index 965d361ff..db1ace170 100644
--- a/test/twin/openai/tests/chat_completions_contract.rs
+++ b/test/twin/openai/tests/chat_completions_contract.rs
@@ -69,6 +69,7 @@ async fn chat_completions_accepts_supported_openai_compatible_fields() {
                 },
                 { "role": "user", "content": "structured chat" }
             ],
+            "max_tokens": 128,
             "stream": false,
             "tools": [{ "type": "function", "function": { "name": "lookup" } }],
             "tool_choice": "auto",