diff --git a/.github/workflows/mantis-telegram-desktop-proof.yml b/.github/workflows/mantis-telegram-desktop-proof.yml
index b3b6c8569c7..2f44534206f 100644
--- a/.github/workflows/mantis-telegram-desktop-proof.yml
+++ b/.github/workflows/mantis-telegram-desktop-proof.yml
@@ -380,8 +380,9 @@ jobs:
           openai-api-key: ${{ secrets.OPENCLAW_MANTIS_AGENT_OPENAI_API_KEY || secrets.OPENAI_API_KEY }}
           prompt-file: .github/codex/prompts/mantis-telegram-desktop-proof.md
           model: ${{ vars.OPENCLAW_CI_OPENAI_MODEL_BARE }}
-          effort: high
+          effort: medium
           sandbox: danger-full-access
+          codex-args: '["-c","service_tier=\"fast\""]'
           codex-home: /tmp/mantis-codex-home-${{ github.run_id }}
           safety-strategy: unprivileged-user
           codex-user: codex
diff --git a/test/scripts/mantis-telegram-desktop-proof-workflow.test.ts b/test/scripts/mantis-telegram-desktop-proof-workflow.test.ts
index 41b430a7f2f..3182dd09cc7 100644
--- a/test/scripts/mantis-telegram-desktop-proof-workflow.test.ts
+++ b/test/scripts/mantis-telegram-desktop-proof-workflow.test.ts
@@ -149,6 +149,14 @@ describe("Mantis Telegram Desktop proof workflow", () => {
     expect(prompt).toContain("do not run\n   `pnpm qa:telegram-user:crabbox` directly");
   });
 
+  it("runs the Mantis Codex agent in fast medium-effort mode", () => {
+    const agent = workflowStep("Run Codex Mantis Telegram agent");
+
+    expect(agent.uses).toContain("openai/codex-action@");
+    expect(agent.with?.effort).toBe("medium");
+    expect(agent.with?.["codex-args"]).toBe('["-c","service_tier=\\"fast\\""]');
+  });
+
   it("derives refs from the PR instead of parsing comment prose", () => {
     const workflowText = readFileSync(WORKFLOW, "utf8");
     expect(workflowText).toContain('setOutput("baseline_ref", pr.base.sha)');