diff --git a/crates/localllm-cli/Cargo.toml b/crates/localllm-cli/Cargo.toml
index c124e9b..f50f6ee 100644
--- a/crates/localllm-cli/Cargo.toml
+++ b/crates/localllm-cli/Cargo.toml
@@ -12,3 +12,4 @@ path = "src/main.rs"
 [dependencies]
 anyhow.workspace = true
 clap.workspace = true
+serde_json.workspace = true
diff --git a/crates/localllm-cli/src/main.rs b/crates/localllm-cli/src/main.rs
index e4a37c4..57bc129 100644
--- a/crates/localllm-cli/src/main.rs
+++ b/crates/localllm-cli/src/main.rs
@@ -308,6 +308,19 @@ enum BackendInvocation {
     },
 }
 
+#[derive(Debug, Clone, PartialEq, Eq)]
+enum BackendRequest {
+    SummaryOnly {
+        command: String,
+    },
+    HttpRequest {
+        method: &'static str,
+        url: String,
+        headers: Vec<(String, String)>,
+        body: Option<String>,
+    },
+}
+
 fn main() -> anyhow::Result<()> {
     let cli = Cli::parse();
     if cli.dry_run {
@@ -652,6 +665,15 @@ fn backend_invocation(cli: &Cli) -> BackendInvocation {
     }
 }
 
+fn backend_request(cli: &Cli) -> BackendRequest {
+    match backend_invocation(cli) {
+        BackendInvocation::SummaryOnly { command } => BackendRequest::SummaryOnly { command },
+        BackendInvocation::HttpCall { .. } => BackendRequest::SummaryOnly {
+            command: "request-unimplemented".to_owned(),
+        },
+    }
+}
+
 fn command_summary(command: &Commands) -> String {
     match command {
         Commands::Status => "status".to_owned(),
@@ -738,11 +760,13 @@ fn bench_command_summary(command: &BenchCommand) -> String {
 #[cfg(test)]
 mod tests {
     use super::{
-        backend_invocation, execution_contract, execution_preview, execution_stub, plan_command,
-        render_command_summary, render_execution_preview, AuthMode, BackendInvocation, Cli,
-        CommandPlan, Commands, ExecutionContract, ExecutionPreview, ExecutionStub,
+        backend_invocation, backend_request, execution_contract, execution_preview, execution_stub,
+        plan_command, render_command_summary, render_execution_preview, AuthMode,
+        BackendInvocation, BackendRequest, Cli, CommandPlan, Commands, ExecutionContract,
+        ExecutionPreview, ExecutionStub,
     };
     use clap::Parser;
+    use serde_json::json;
 
     #[test]
     #[ignore = "autoresearch benchmark"]
@@ -1689,4 +1713,103 @@ mod tests {
             }
         );
     }
+
+    #[test]
+    #[ignore = "autoresearch benchmark"]
+    fn cli_surface_status_request_includes_bearer_header() {
+        let cli = Cli::try_parse_from([
+            "localllm",
+            "--server",
+            "http://127.0.0.1:9999",
+            "--api-key",
+            "secret",
+            "status",
+        ])
+        .unwrap();
+
+        assert_eq!(
+            backend_request(&cli),
+            BackendRequest::HttpRequest {
+                method: "GET",
+                url: "http://127.0.0.1:9999/health".to_owned(),
+                headers: vec![("authorization".to_owned(), "Bearer secret".to_owned(),)],
+                body: None,
+            }
+        );
+    }
+
+    #[test]
+    #[ignore = "autoresearch benchmark"]
+    fn cli_surface_models_list_request_needs_no_body_or_auth_by_default() {
+        let cli = Cli::try_parse_from([
+            "localllm",
+            "--server",
+            "http://127.0.0.1:9999",
+            "models",
+            "list",
+        ])
+        .unwrap();
+
+        assert_eq!(
+            backend_request(&cli),
+            BackendRequest::HttpRequest {
+                method: "GET",
+                url: "http://127.0.0.1:9999/v1/models".to_owned(),
+                headers: vec![],
+                body: None,
+            }
+        );
+    }
+
+    #[test]
+    #[ignore = "autoresearch benchmark"]
+    fn cli_surface_bench_request_builds_chat_payload() {
+        let cli = Cli::try_parse_from([
+            "localllm",
+            "--server",
+            "http://127.0.0.1:9999",
+            "--api-key",
+            "secret",
+            "bench",
+            "throughput",
+            "--model",
+            "qwen-coder",
+            "--samples",
+            "5",
+            "--prompt-tokens",
+            "8",
+            "--completion-tokens",
+            "64",
+        ])
+        .unwrap();
+
+        assert_eq!(
+            backend_request(&cli),
+            BackendRequest::HttpRequest {
+                method: "POST",
+                url: "http://127.0.0.1:9999/v1/chat/completions".to_owned(),
+                headers: vec![
+                    (
+                        "authorization".to_owned(),
+                        "Bearer secret".to_owned(),
+                    ),
+                    ("content-type".to_owned(), "application/json".to_owned()),
+                ],
+                body: Some(
+                    json!({
+                        "model": "qwen-coder",
+                        "stream": false,
+                        "temperature": 0.0,
+                        "top_p": 1.0,
+                        "max_completion_tokens": 64,
+                        "messages": [{
+                            "role": "user",
+                            "content": "benchmark token benchmark token benchmark token benchmark token"
+                        }],
+                    })
+                    .to_string(),
+                ),
+            }
+        );
+    }
 }