diff --git a/package.json b/package.json
index 1e5e677..d0da4ea 100644
--- a/package.json
+++ b/package.json
@@ -4,6 +4,8 @@
   "description": "ExpressJS Performance Working Group",
   "scripts": {
     "setup": "npm i",
+    "local-server": "expf local-server",
+    "local-load": "expf local-load",
     "test": "echo \"Error: no test specified\" && exit 1",
     "load": "expf load",
     "test:load": "expf load --test=@expressjs/perf-load-example",
diff --git a/packages/cli/bin/expf.mjs b/packages/cli/bin/expf.mjs
index 6e0e78e..914a950 100644
--- a/packages/cli/bin/expf.mjs
+++ b/packages/cli/bin/expf.mjs
@@ -78,6 +78,20 @@ switch (positionals[2]) {
       console.error(e);
     }
     break;
+  case 'local-server':
+    try {
+      await (await import('../local-server.mjs')).default(values);
+    } catch (e) {
+      console.error(e);
+    }
+    break;
+  case 'local-load':
+    try {
+      await (await import('../local-load.mjs')).default(values);
+    } catch (e) {
+      console.error(e);
+    }
+    break;
   default:
     console.log(`
 Express Performance Testing CLI
diff --git a/packages/cli/load.mjs b/packages/cli/load.mjs
index e3f68ea..6f0a58a 100644
--- a/packages/cli/load.mjs
+++ b/packages/cli/load.mjs
@@ -2,6 +2,7 @@ import { normalize, join, dirname } from 'node:path';
 import { writeFile, mkdir } from 'node:fs/promises';
 import nv from '@pkgjs/nv';
 
+// TODO: Add description to each flag
 export function help (opts = {}) {
   return `$ expf load [flags]
 
diff --git a/packages/cli/local-load.mjs b/packages/cli/local-load.mjs
new file mode 100644
index 0000000..a10ab91
--- /dev/null
+++ b/packages/cli/local-load.mjs
@@ -0,0 +1,179 @@
+import { spawnSync, spawn } from 'node:child_process';
+import { availableParallelism } from 'node:os';
+import { styleText } from 'node:util';
+
+// TODO: add options to load different servers
+export function help () {
+  return `$ expf local-bench
+
+  Start Autocannon load to local HTTP server
+`
+}
+
+class Wrk2Benchmarker {
+  constructor() {
+    this.name = 'wrk2';
+    this.executable = 'wrk2';
+    const result = spawnSync(this.executable, ['-h']);
+    this.present = !(result.error && result.error.code === 'ENOENT');
+  }
+
+  create(options) {
+    const duration = typeof options.duration === 'number' ?
+      Math.max(options.duration, 1) :
+      options.duration;
+    const scheme = options.scheme || 'http';
+    const args = [
+      '-d', duration,
+      '-c', options.connections,
+      '-R', options.rate,
+      '--latency',
+      '-t', Math.min(options.connections, availableParallelism() || 8),
+      `${scheme}://127.0.0.1:${options.port}${options.path}`,
+    ];
+    for (const field in options.headers) {
+      args.push('-H', `${field}: ${options.headers[field]}`);
+    }
+    const child = spawn(this.executable, args);
+    return child;
+  }
+
+  processResults(output) {
+    // Capture only the Latency Distribution block (until a blank line or "Detailed Percentile spectrum:")
+    const blockRe =
+      /Latency Distribution \(HdrHistogram - Recorded Latency\)\s*\n([\s\S]*?)(?:\n\s*\n|^ {0,2}Detailed Percentile spectrum:|\Z)/m;
+
+    const m = output.match(blockRe);
+    if (!m) return undefined;
+
+    const lines = m[1].trim().split('\n');
+
+    // e.g.: " 50.000%  780.00us" or " 90.000%    1.23ms"
+    const lineRe = /^\s*(\d{1,3}\.\d{3})%\s+([0-9.]+)\s*(ms|us)\s*$/;
+
+    const points = [];
+    for (const line of lines) {
+      const lm = line.match(lineRe);
+      if (!lm) continue;
+      const pct = parseFloat(lm[1]);          // e.g. 99.900
+      const val = parseFloat(lm[2]);          // numeric value
+      const unit = lm[3];                     // "ms" | "us"
+      const valueMs = unit === 'us' ? val / 1000 : val;
+      if (Number.isFinite(pct) && Number.isFinite(valueMs)) {
+        points.push({ percentile: pct, ms: valueMs });
+      }
+    }
+
+    return points.length ? points : undefined;
+  }
+}
+
+class AutocannonBenchmarker {
+  constructor() {
+    const shell = (process.platform === 'win32');
+    this.name = 'autocannon';
+    this.opts = { shell };
+    this.executable = shell ? 'autocannon.cmd' : 'autocannon';
+    const result = spawnSync(this.executable, ['-h'], this.opts);
+    if (shell) {
+      this.present = (result.status === 0);
+    } else {
+      this.present = !(result.error && result.error.code === 'ENOENT');
+    }
+  }
+
+  create(options) {
+    const args = [
+      '-d', options.duration,
+      '-c', options.connections,
+      '-j',
+      '-n',
+    ];
+    for (const field in options.headers) {
+      if (this.opts.shell) {
+        args.push('-H', `'${field}=${options.headers[field]}'`);
+      } else {
+        args.push('-H', `${field}=${options.headers[field]}`);
+      }
+    }
+    const scheme = options.scheme || 'http';
+    args.push(`${scheme}://127.0.0.1:${options.port}${options.path}`);
+    const child = spawn(this.executable, args, this.opts);
+    return child;
+  }
+
+  processResults(output) {
+    let result;
+    try {
+      result = JSON.parse(output);
+    } catch {
+      return undefined;
+    }
+    if (!result || !result.requests || !result.requests.average) {
+      return undefined;
+    }
+    return result.requests.average;
+  }
+}
+
+function runBenchmarker(instance, options) {
+  console.log(styleText(['blue'], `Running ${instance.name} load...`));
+  return new Promise((resolve, reject) => {
+    const proc = instance.create(options);
+    proc.stderr.pipe(process.stderr);
+
+    let stdout = '';
+    proc.stdout.setEncoding('utf8');
+    proc.stdout.on('data', (chunk) => stdout += chunk);
+
+    proc.once('close', (code) => {
+      if (code) {
+        let error_message = `${instance.name} failed with ${code}.`;
+        if (stdout !== '') {
+          error_message += ` Output: ${stdout}`;
+        }
+        reject(new Error(error_message), code);
+        return;
+      }
+
+      const result = instance.processResults(stdout);
+      resolve(result);
+    });
+  })
+
+}
+
+// TODO: accept args to decide which load tool use
+// For now, use both.
+export async function startLoad () {
+  const autocannon = new AutocannonBenchmarker();
+  if (!autocannon.present) {
+    console.log(styleText(['bold', 'yellow'], 'Autocannon not found. Please install it with `npm i -g autocannon`'));
+  } else {
+    const result = await runBenchmarker(autocannon, {
+      duration: 30,
+      connections: 100,
+      port: 3000,
+      path: '/',
+    });
+    console.log('Result Autocannon:', result)
+  }
+
+  const wrk2 = new Wrk2Benchmarker();
+  if (!wrk2.present) {
+    console.log(styleText(['bold', 'yellow'], 'Wrk2 not found. Please install it'));
+  } else {
+    const result = await runBenchmarker(wrk2, {
+      duration: 30,
+      connections: 100,
+      rate: 2000,
+      port: 3000,
+      path: '/'
+    });
+    console.log('Result Wrk2:', result);
+  }
+}
+
+export default async function main () {
+  await startLoad();
+}
diff --git a/packages/cli/local-server.mjs b/packages/cli/local-server.mjs
new file mode 100644
index 0000000..d11f229
--- /dev/null
+++ b/packages/cli/local-server.mjs
@@ -0,0 +1,39 @@
+import { join } from 'node:path';
+import { spawn } from 'node:child_process';
+
+// TODO: add options to load different servers
+export function help () {
+  return `$ expf local-server
+
+  Start Express local HTTP server
+`
+}
+
+export function startServer () {
+  const server = spawn(
+    process.execPath,
+    [
+      join(import.meta.dirname, '../../servers/node-http/index.mjs'),
+    ],
+    {
+      env: {
+        ...process.env,
+        PORT: 3000
+      },
+      stdio: ['inherit', 'inherit', 'inherit']
+    }
+  );
+
+  return new Promise((resolve, reject) => {
+    server.on('error', reject);
+    server.on('close', resolve);
+
+    process.on('beforeExit', () => {
+      server.close();
+    });
+  })
+}
+
+export default async function main (_opts = {}) {
+  await startServer();
+}