diff --git a/app/test/config.test.js b/app/test/config.test.js
index 7139dae..942acc3 100644
--- a/app/test/config.test.js
+++ b/app/test/config.test.js
@@ -3,7 +3,7 @@ import assert from 'node:assert/strict';
 import { getLLMConfig, getOpenCodeHttpsAgent } from '../config.js';
 
 const ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER',
+  'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL',
 ];
 
 let saved = {};
@@ -20,26 +20,35 @@ afterEach(() => {
 
 describe('getLLMConfig', () => {
   it('returns null provider when no env vars set', () => {
-    const cfg = getLLMConfig();
+    const cfg = getLLMConfig({ commandExistsFn: () => false });
     assert.equal(cfg.provider, null);
     assert.deepEqual(cfg.apiKeys, []);
+    assert.equal(cfg.command, null);
   });
 
-  it('detects opencode server with defaults', () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    const cfg = getLLMConfig();
-    assert.equal(cfg.provider, 'opencode');
-    assert.deepEqual(cfg.apiKeys, ['opencode']);
-    assert.equal(cfg.baseURL, 'http://opencode.local:4096');
-    assert.equal(cfg.model, 'gemini-2.5-flash');
+  it('detects the first installed assistant CLI with defaults', () => {
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'claude' });
+    assert.equal(cfg.provider, 'claude');
+    assert.deepEqual(cfg.apiKeys, ['claude']);
+    assert.equal(cfg.baseURL, null);
+    assert.equal(cfg.command, 'claude');
+    assert.equal(cfg.model, 'sonnet');
   });
 
-  it('detects opencode server with custom model', () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
+  it('uses MODEL for the selected assistant CLI', () => {
+    process.env.MODEL = 'gpt-5-mini';
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'codex' });
+    assert.equal(cfg.provider, 'codex');
+    assert.equal(cfg.command, 'codex');
+    assert.equal(cfg.model, 'gpt-5-mini');
+  });
+
+  it('can force a CLI with AI_ASSISTANT_CLI', () => {
+    process.env.AI_ASSISTANT_CLI = 'opencode';
     process.env.OPENCODE_MODEL = 'google/gemini-2.5-pro';
-    const cfg = getLLMConfig();
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'codex' || command === 'opencode' });
     assert.equal(cfg.provider, 'opencode');
-    assert.equal(cfg.baseURL, 'http://opencode.local:4096');
+    assert.equal(cfg.command, 'opencode');
     assert.equal(cfg.model, 'google/gemini-2.5-pro');
   });
 
diff --git a/app/test/llm.test.js b/app/test/llm.test.js
index ffb3e1e..5d7bd5c 100644
--- a/app/test/llm.test.js
+++ b/app/test/llm.test.js
@@ -1,133 +1,96 @@
 import { describe, it, beforeEach, afterEach, mock } from 'node:test';
 import assert from 'node:assert/strict';
-import axios from 'axios';
+import { mkdtemp, writeFile, chmod, rm, readFile } from 'fs/promises';
+import { tmpdir } from 'os';
+import { join } from 'path';
 import { extractBalancedJSON, extractJSONText } from '../llm.js';
 
 const ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER', 'OPENCODE_RETRY_ATTEMPTS',
+  'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL', 'PATH', 'AI_ASSISTANT_TIMEOUT_MS', 'AI_ASSISTANT_MAX_BUFFER',
+  'FAKE_AI_STDOUT', 'FAKE_AI_STDERR', 'FAKE_AI_EXIT', 'FAKE_AI_STDIN_PATH', 'FAKE_AI_ARGS_PATH',
 ];
 
 let saved = {};
+let tempDir;
 beforeEach(() => {
   saved = {};
   for (const k of ENV_KEYS) { saved[k] = process.env[k]; delete process.env[k]; }
+  tempDir = null;
 });
-afterEach(() => {
+afterEach(async () => {
   for (const k of ENV_KEYS) {
     if (saved[k] === undefined) delete process.env[k];
     else process.env[k] = saved[k];
   }
+  if (tempDir) await rm(tempDir, { recursive: true, force: true });
   mock.restoreAll();
 });
 
-function mockOpenCodeResponse(content) {
-  let calls = 0;
-  mock.method(axios, 'post', async () => {
-    calls += 1;
-    if (calls === 1) return { data: { id: 'ses_test' } };
-    return { data: { parts: [{ type: 'text', text: content }] } };
-  });
+async function installFakeCLI(command = 'codex') {
+  tempDir = await mkdtemp(join(tmpdir(), 'ai-cli-test-'));
+  const script = join(tempDir, command);
+  await writeFile(script, `#!/bin/sh
+if [ -n "$FAKE_AI_ARGS_PATH" ]; then printf '%s\\n' "$*" > "$FAKE_AI_ARGS_PATH"; fi
+if [ -n "$FAKE_AI_STDIN_PATH" ]; then cat > "$FAKE_AI_STDIN_PATH"; else cat >/dev/null; fi
+if [ -n "$FAKE_AI_STDERR" ]; then printf '%s' "$FAKE_AI_STDERR" >&2; fi
+if [ -n "$FAKE_AI_STDOUT" ]; then printf '%s' "$FAKE_AI_STDOUT"; fi
+exit "\${FAKE_AI_EXIT:-0}"
+`);
+  await chmod(script, 0o755);
+  process.env.PATH = `${tempDir}:${saved.PATH || ''}`;
+  return { stdinPath: join(tempDir, 'stdin.txt'), argsPath: join(tempDir, 'args.txt') };
 }
 
-describe('chat - OpenCode', async () => {
+describe('chat - assistant CLI', async () => {
   const { chat } = await import('../llm.js');
 
-  it('uses OpenCode server session API', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    process.env.OPENCODE_MODEL = 'gemini-2.5-flash';
-    const calls = [];
-    mock.method(axios, 'post', async (url, payload, opts) => {
-      calls.push({ url, payload, headers: opts.headers });
-      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
-      return { data: { parts: [{ type: 'text', text: 'opencode response' }] } };
-    });
+  it('runs the detected CLI with MODEL and sends the prompts through stdin', async () => {
+    const { stdinPath, argsPath } = await installFakeCLI('codex');
+    process.env.MODEL = 'gpt-5-mini';
+    process.env.FAKE_AI_STDOUT = 'cli response';
+    process.env.FAKE_AI_STDIN_PATH = stdinPath;
+    process.env.FAKE_AI_ARGS_PATH = argsPath;
 
     const result = await chat('sys', 'user');
 
-    assert.equal(result, 'opencode response');
-    assert.equal(calls[0].url, 'http://opencode.local:4096/session');
-    assert.deepEqual(calls[0].payload.model, { providerID: 'google', id: 'gemini-2.5-flash' });
-    assert.equal(calls[1].url, 'http://opencode.local:4096/session/ses_test/message');
-    assert.deepEqual(calls[1].payload.model, { providerID: 'google', modelID: 'gemini-2.5-flash' });
-    assert.equal(calls[1].payload.system, 'sys');
-    assert.deepEqual(calls[1].payload.parts, [{ type: 'text', text: 'user' }]);
-    assert.equal(calls[1].headers['Authorization'], undefined);
+    assert.equal(result, 'cli response');
+    assert.match(await readFile(argsPath, 'utf8'), /exec --model gpt-5-mini/);
+    const prompt = await readFile(stdinPath, 'utf8');
+    assert.match(prompt, /<system>\nsys\n<\/system>/);
+    assert.match(prompt, /<user>\nuser\n<\/user>/);
   });
 
-  it('passes an insecure https agent to OpenCode by default', async () => {
-    process.env.OPENCODE_BASE_URL = 'https://opencode.local:4096';
-    const agents = [];
-    mock.method(axios, 'post', async (url, _payload, opts) => {
-      agents.push(opts.httpsAgent);
-      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
-      return { data: { parts: [{ type: 'text', text: 'ok' }] } };
-    });
-
-    await chat('sys', 'user');
-
-    assert.equal(agents.length, 2);
-    assert.equal(agents[0].options.rejectUnauthorized, false);
-    assert.equal(agents[1].options.rejectUnauthorized, false);
-  });
-
-  it('extracts text from OpenCode message parts', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    let calls = 0;
-    mock.method(axios, 'post', async () => {
-      calls += 1;
-      if (calls === 1) return { data: { id: 'ses_test' } };
-      return { data: { parts: [{ type: 'text', text: 'hello' }, { type: 'text', text: ' world' }] } };
-    });
+  it('can force opencode with AI_ASSISTANT_CLI', async () => {
+    const { argsPath } = await installFakeCLI('opencode');
+    process.env.AI_ASSISTANT_CLI = 'opencode';
+    process.env.MODEL = 'google/gemini-2.5-pro';
+    process.env.FAKE_AI_STDOUT = 'ok';
+    process.env.FAKE_AI_ARGS_PATH = argsPath;
 
     const result = await chat('sys', 'user');
 
-    assert.equal(result, 'hello world');
+    assert.equal(result, 'ok');
+    assert.match(await readFile(argsPath, 'utf8'), /run --model google\/gemini-2.5-pro/);
   });
 
-  it('throws an error when OpenCode fails instead of exiting the process', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_RETRY_ATTEMPTS = '1';
-    mock.method(axios, 'post', async () => {
-      const err = new Error('Request failed with status code 500');
-      err.response = { status: 500, data: { error: 'provider overloaded' } };
-      throw err;
-    });
+  it('throws an error when the CLI fails instead of exiting the process', async () => {
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_EXIT = '2';
+    process.env.FAKE_AI_STDERR = 'provider overloaded';
     const exitMock = mock.method(process, 'exit', () => { throw new Error('exit should not be called'); });
 
-    await assert.rejects(() => chat('sys', 'user'), /HTTP 500.*provider overloaded/);
+    await assert.rejects(() => chat('sys', 'user'), /provider overloaded/);
 
     assert.equal(exitMock.mock.calls.length, 0);
   });
-
-  it('retries transient OpenCode failures before returning content', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_RETRY_ATTEMPTS = '2';
-    let messageAttempts = 0;
-    mock.method(axios, 'post', async (url) => {
-      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
-      messageAttempts += 1;
-      if (messageAttempts === 1) {
-        const err = new Error('Request failed with status code 500');
-        err.response = { status: 500, data: { error: 'temporary failure' } };
-        throw err;
-      }
-      return { data: { parts: [{ type: 'text', text: 'ok after retry' }] } };
-    });
-
-    const result = await chat('sys', 'user');
-
-    assert.equal(result, 'ok after retry');
-    assert.equal(messageAttempts, 2);
-  });
 });
 
 describe('chatJSON', async () => {
   const { chatJSON } = await import('../llm.js');
 
   it('parses plain JSON response', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('[{"level":"critical"}]');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '[{"level":"critical"}]';
 
     const result = await chatJSON('sys', 'user');
 
@@ -135,8 +98,8 @@ describe('chatJSON', async () => {
   });
 
   it('strips markdown code block before parsing', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('```json\n[{"level":"info"}]\n```');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '```json\n[{"level":"info"}]\n```';
 
     const result = await chatJSON('sys', 'user');
 
@@ -144,8 +107,8 @@ describe('chatJSON', async () => {
   });
 
   it('extracts JSON array from surrounding prose', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('**Reviewing findings**\n\n[{"level":"warning","suggestion":"x"}]\n\nDone.');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '**Reviewing findings**\n\n[{"level":"warning","suggestion":"x"}]\n\nDone.';
 
     const result = await chatJSON('sys', 'user');
 
@@ -153,8 +116,8 @@ describe('chatJSON', async () => {
   });
 
   it('extracts JSON object from surrounding prose', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('**Begin Combine**\n{"merged_text":"repo block\\n\\nsource block"}');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '**Begin Combine**\n{"merged_text":"repo block\\n\\nsource block"}';
 
     const result = await chatJSON('sys', 'user');
 
@@ -162,8 +125,8 @@ describe('chatJSON', async () => {
   });
 
   it('returns [] when JSON is invalid', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('not json');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = 'not json';
 
     const result = await chatJSON('sys', 'user');
 
diff --git a/app/test/preflight.test.js b/app/test/preflight.test.js
index 8d02e5e..4f6e4c6 100644
--- a/app/test/preflight.test.js
+++ b/app/test/preflight.test.js
@@ -1,21 +1,38 @@
 import { describe, it, afterEach, mock } from 'node:test';
 import assert from 'node:assert/strict';
 import axios from 'axios';
+import { mkdtemp, writeFile, chmod, rm } from 'fs/promises';
+import { tmpdir } from 'os';
+import { join } from 'path';
 import { checkRequiredEnv, verifyGiteaToken, verifyCommentToken, verifyLLM, runPreflight } from '../preflight.js';
 
 const LLM_ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER',
+  'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL', 'PATH',
 ];
+const ORIGINAL_PATH = process.env.PATH;
 
 function clearLLMEnv() {
   for (const k of LLM_ENV_KEYS) delete process.env[k];
 }
 
-afterEach(() => {
+let tempDir;
+
+afterEach(async () => {
   mock.restoreAll();
   clearLLMEnv();
+  process.env.PATH = ORIGINAL_PATH;
+  if (tempDir) await rm(tempDir, { recursive: true, force: true });
+  tempDir = null;
 });
 
+async function installFakeCLI(command = 'codex') {
+  tempDir = await mkdtemp(join(tmpdir(), 'preflight-cli-test-'));
+  const script = join(tempDir, command);
+  await writeFile(script, '#!/bin/sh\nexit 0\n');
+  await chmod(script, 0o755);
+  process.env.PATH = tempDir;
+}
+
 describe('checkRequiredEnv', () => {
   it('reports all three missing when nothing provided', () => {
     const result = checkRequiredEnv({ token: '', repo: '', pr: '' });
@@ -102,81 +119,40 @@ describe('verifyCommentToken', () => {
 });
 
 describe('verifyLLM', () => {
-  it('fails when OpenCode is not configured', async () => {
+  it('fails when no supported assistant CLI is detected', async () => {
     clearLLMEnv();
+    process.env.AI_ASSISTANT_CLI = 'no-such-ai-cli';
+    process.env.PATH = '';
 
     const result = await verifyLLM();
 
     assert.equal(result.ok, false);
-    assert.match(result.error, /OPENCODE_BASE_URL/);
+    assert.match(result.error, /AI 助理 CLI/);
   });
 
-  it('checks OpenCode server provider and model', async () => {
+  it('passes when a supported assistant CLI is detected', async () => {
     clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    process.env.OPENCODE_MODEL = 'gemini-2.5-flash';
-    const urls = [];
-    mock.method(axios, 'get', async (url) => {
-      urls.push(url);
-      if (url.endsWith('/global/health')) return { data: { healthy: true, version: '1.17.7' } };
-      return { data: { providers: [{ id: 'google', models: { 'gemini-2.5-flash': { id: 'gemini-2.5-flash' } } }] } };
-    });
+    await installFakeCLI('codex');
+    process.env.AI_ASSISTANT_CLI = 'codex';
+    process.env.MODEL = 'gpt-5-mini';
 
     const result = await verifyLLM();
 
     assert.equal(result.ok, true);
-    assert.equal(result.provider, 'opencode');
-    assert.deepEqual(urls, ['http://opencode.local:4096/global/health', 'http://opencode.local:4096/config/providers']);
+    assert.equal(result.provider, 'codex');
+    assert.equal(result.command, 'codex');
+    assert.equal(result.model, 'gpt-5-mini');
   });
 
-  it('fails when configured provider is missing', async () => {
+  it('fails when a requested CLI is not installed', async () => {
     clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    mock.method(axios, 'get', async (url) => {
-      if (url.endsWith('/global/health')) return { data: { healthy: true } };
-      return { data: { providers: [{ id: 'anthropic', models: {} }] } };
-    });
+    process.env.AI_ASSISTANT_CLI = 'missing-cli';
+    process.env.PATH = '';
 
     const result = await verifyLLM();
 
     assert.equal(result.ok, false);
-    assert.match(result.error, /未設定 provider=google/);
-  });
-
-  it('fails when configured model is missing', async () => {
-    clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    process.env.OPENCODE_MODEL = 'gemini-2.5-pro';
-    mock.method(axios, 'get', async (url) => {
-      if (url.endsWith('/global/health')) return { data: { healthy: true } };
-      return { data: { providers: [{ id: 'google', models: { 'gemini-2.5-flash': { id: 'gemini-2.5-flash' } } }] } };
-    });
-
-    const result = await verifyLLM();
-
-    assert.equal(result.ok, false);
-    assert.match(result.error, /未列出 model=gemini-2.5-pro/);
-  });
-
-  it('passes an insecure https agent by default', async () => {
-    clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'https://opencode.local:4096';
-    const agents = [];
-    mock.method(axios, 'get', async (url, opts) => {
-      agents.push(opts.httpsAgent);
-      if (url.endsWith('/global/health')) return { data: { healthy: true } };
-      return { data: { providers: [{ id: 'google', models: { 'gemini-2.5-flash': { id: 'gemini-2.5-flash' } } }] } };
-    });
-
-    const result = await verifyLLM();
-
-    assert.equal(result.ok, true);
-    assert.equal(agents.length, 2);
-    assert.equal(agents[0].options.rejectUnauthorized, false);
-    assert.equal(agents[1].options.rejectUnauthorized, false);
+    assert.match(result.error, /AI 助理 CLI/);
   });
 
 });
@@ -188,7 +164,7 @@ describe('runPreflight', () => {
       verifyToken: async () => ({ ok: true }),
       verifyComment: async () => ({ ok: true }),
       verifyRemote: () => ({ ok: true }),
-      verifyLLMFn: async () => ({ ok: true, provider: 'opencode' }),
+      verifyLLMFn: async () => ({ ok: true, provider: 'codex' }),
       ...overrides,
     };
   }
@@ -239,7 +215,7 @@ describe('runPreflight', () => {
 
   it('returns false when LLM verification fails', async () => {
     const result = await runPreflight('/ws', makeDeps({
-      verifyLLMFn: async () => ({ ok: false, error: 'OpenCode server 驗證失敗' }),
+      verifyLLMFn: async () => ({ ok: false, error: 'AI 助理 CLI 驗證失敗' }),
     }));
     assert.equal(result, false);
   });