diff --git a/.github/workflows/performance-monitor.yml b/.github/workflows/performance-monitor.yml
new file mode 100644
index 00000000..db34cd61
--- /dev/null
+++ b/.github/workflows/performance-monitor.yml
@@ -0,0 +1,132 @@
+name: Performance Monitor
+
+on:
+  schedule:
+    # Run weekly on Mondays at 06:00 UTC
+    - cron: "0 6 * * 1"
+  workflow_dispatch:
+    inputs:
+      iterations:
+        description: "Number of iterations per metric"
+        required: false
+        default: "5"
+
+permissions:
+  contents: read
+  issues: write
+
+jobs:
+  benchmark:
+    name: Run Performance Benchmarks
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: npm
+
+      - name: Install dependencies
+        run: npm ci
+
+      - name: Build
+        run: npm run build
+
+      - name: Install AWF locally
+        run: |
+          sudo tee /usr/local/bin/awf > /dev/null <<'WRAPPER'
+          #!/bin/bash
+          exec node "${{ github.workspace }}/dist/cli.js" "$@"
+          WRAPPER
+          sudo chmod +x /usr/local/bin/awf
+
+      - name: Run benchmarks
+        id: benchmark
+        run: |
+          npx tsx scripts/ci/benchmark-performance.ts > benchmark-results.json 2>&1 || true
+          cat benchmark-results.json
+
+      - name: Upload results
+        uses: actions/upload-artifact@v4
+        with:
+          name: benchmark-results-${{ github.sha }}
+          path: benchmark-results.json
+          retention-days: 90
+
+      - name: Check for regressions
+        id: check
+        run: |
+          if [ ! -f benchmark-results.json ]; then
+            echo "No benchmark results found"
+            exit 1
+          fi
+
+          REGRESSIONS=$(jq -r '.regressions | length' benchmark-results.json)
+          echo "regression_count=$REGRESSIONS" >> "$GITHUB_OUTPUT"
+
+          if [ "$REGRESSIONS" -gt 0 ]; then
+            echo "## ⚠️ Performance Regressions Detected" >> "$GITHUB_STEP_SUMMARY"
+            echo "" >> "$GITHUB_STEP_SUMMARY"
+            jq -r '.regressions[]' benchmark-results.json | while read -r line; do
+              echo "- $line" >> "$GITHUB_STEP_SUMMARY"
+            done
+          else
+            echo "## ✅ All Metrics Within Thresholds" >> "$GITHUB_STEP_SUMMARY"
+          fi
+
+          echo "" >> "$GITHUB_STEP_SUMMARY"
+          echo "### Results" >> "$GITHUB_STEP_SUMMARY"
+          echo "" >> "$GITHUB_STEP_SUMMARY"
+          echo "| Metric | Mean | Median | P95 | P99 | Target | Critical |" >> "$GITHUB_STEP_SUMMARY"
+          echo "|--------|------|--------|-----|-----|--------|----------|" >> "$GITHUB_STEP_SUMMARY"
+
+          jq -r '.results[] as $r | .thresholds[$r.metric] as $t |
+            "| \($r.metric) | \($r.mean)\($r.unit) | \($r.median)\($r.unit) | \($r.p95)\($r.unit) | \($r.p99)\($r.unit) | \($t.target // "N/A")\(if $t then $r.unit else "" end) | \($t.critical // "N/A")\(if $t then $r.unit else "" end) |"' \
+            benchmark-results.json >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Create regression issue
+        if: steps.check.outputs.regression_count != '0'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const fs = require('fs');
+            const report = JSON.parse(fs.readFileSync('benchmark-results.json', 'utf-8'));
+
+            const body = [
+              `## Performance Regression Detected`,
+              ``,
+              `**Commit:** ${report.commitSha}`,
+              `**Timestamp:** ${report.timestamp}`,
+              `**Iterations:** ${report.iterations}`,
+              ``,
+              `### Regressions`,
+              ``,
+              ...report.regressions.map(r => `- ${r}`),
+              ``,
+              `### Full Results`,
+              ``,
+              `| Metric | Mean | Median | P95 | P99 |`,
+              `|--------|------|--------|-----|-----|`,
+              ...report.results.map(r =>
+                `| ${r.metric} | ${r.mean}${r.unit} | ${r.median}${r.unit} | ${r.p95}${r.unit} | ${r.p99}${r.unit} |`
+              ),
+              ``,
+              `<details><summary>Raw JSON</summary>`,
+              ``,
+              '```json',
+              JSON.stringify(report, null, 2),
+              '```',
+              `</details>`,
+            ].join('\n');
+
+            await github.rest.issues.create({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              title: `Performance regression detected (${new Date().toISOString().split('T')[0]})`,
+              body,
+              labels: ['performance', 'needs-investigation'],
+            });
diff --git a/package.json b/package.json
index 1b80d7c2..e34a639a 100644
--- a/package.json
+++ b/package.json
@@ -22,7 +22,8 @@
     "prepare": "husky",
     "docs:dev": "cd docs-site && npm run dev",
     "docs:build": "cd docs-site && npm run build",
-    "docs:preview": "cd docs-site && npm run preview"
+    "docs:preview": "cd docs-site && npm run preview",
+    "benchmark": "npx tsx scripts/ci/benchmark-performance.ts"
   },
   "keywords": [
     "agentic-workflows",
diff --git a/scripts/ci/benchmark-performance.ts b/scripts/ci/benchmark-performance.ts
new file mode 100644
index 00000000..d0112cb6
--- /dev/null
+++ b/scripts/ci/benchmark-performance.ts
@@ -0,0 +1,295 @@
+#!/usr/bin/env npx tsx
+/**
+ * Performance benchmark script for AWF (Agentic Workflow Firewall).
+ *
+ * Measures key metrics:
+ *  - Container startup (cold & warm)
+ *  - Squid HTTP / HTTPS proxy latency
+ *  - Memory footprint
+ *  - Docker network creation time
+ *
+ * Outputs structured JSON with mean, median, p95, p99 per metric.
+ */
+
+import { execSync, ExecSyncOptions } from "child_process";
+
+// ── Configuration ──────────────────────────────────────────────────
+
+const ITERATIONS = 5;
+const AWF_CMD = "sudo awf";
+const ALLOWED_DOMAIN = "api.github.com";
+const CLEANUP_CMD = "sudo docker compose down -v 2>/dev/null; sudo docker rm -f awf-squid awf-agent 2>/dev/null; sudo docker network prune -f 2>/dev/null";
+
+interface BenchmarkResult {
+  metric: string;
+  unit: string;
+  values: number[];
+  mean: number;
+  median: number;
+  p95: number;
+  p99: number;
+}
+
+interface BenchmarkReport {
+  timestamp: string;
+  commitSha: string;
+  iterations: number;
+  results: BenchmarkResult[];
+  thresholds: Record<string, { target: number; critical: number }>;
+  regressions: string[];
+}
+
+// ── Thresholds (milliseconds or MB) ───────────────────────────────
+
+const THRESHOLDS: Record<string, { target: number; critical: number }> = {
+  "container_startup_cold": { target: 15000, critical: 20000 },
+  "container_startup_warm": { target: 5000, critical: 8000 },
+  "squid_https_latency": { target: 100, critical: 200 },
+  "memory_footprint_mb": { target: 500, critical: 1024 },
+  "docker_network_creation": { target: 2000, critical: 5000 },
+};
+
+// ── Helpers ────────────────────────────────────────────────────────
+
+function exec(cmd: string, opts?: ExecSyncOptions): string {
+  return execSync(cmd, { encoding: "utf-8", timeout: 120_000, ...opts }).trim();
+}
+
+function timeMs(fn: () => void): number {
+  const start = performance.now();
+  fn();
+  return Math.round(performance.now() - start);
+}
+
+function stats(values: number[]): Pick<BenchmarkResult, "mean" | "median" | "p95" | "p99"> {
+  const sorted = [...values].sort((a, b) => a - b);
+  const n = sorted.length;
+  return {
+    mean: Math.round(sorted.reduce((a, b) => a + b, 0) / n),
+    median: sorted[Math.floor(n / 2)],
+    p95: sorted[Math.min(Math.floor(n * 0.95), n - 1)],
+    p99: sorted[Math.min(Math.floor(n * 0.99), n - 1)],
+  };
+}
+
+function cleanup(): void {
+  try {
+    execSync(CLEANUP_CMD, { stdio: "ignore", timeout: 30_000 });
+  } catch {
+    // best-effort
+  }
+}
+
+// ── Benchmarks ─────────────────────────────────────────────────────
+
+function benchmarkColdStart(): BenchmarkResult {
+  console.error("  Benchmarking cold container startup...");
+  const values: number[] = [];
+
+  for (let i = 0; i < ITERATIONS; i++) {
+    cleanup();
+    // Remove cached images to force cold pull
+    try {
+      execSync("sudo docker rmi ghcr.io/github/gh-aw-firewall/squid:latest ghcr.io/github/gh-aw-firewall/agent:latest 2>/dev/null", { stdio: "ignore", timeout: 30_000 });
+    } catch {
+      // images may not exist
+    }
+
+    const ms = timeMs(() => {
+      exec(`${AWF_CMD} --allow-domains ${ALLOWED_DOMAIN} --log-level error -- echo ok`, { stdio: "ignore" });
+    });
+    values.push(ms);
+    console.error(`    Iteration ${i + 1}/${ITERATIONS}: ${ms}ms`);
+  }
+
+  return { metric: "container_startup_cold", unit: "ms", values, ...stats(values) };
+}
+
+function benchmarkWarmStart(): BenchmarkResult {
+  console.error("  Benchmarking warm container startup...");
+  const values: number[] = [];
+
+  // Ensure images are pulled
+  cleanup();
+  try {
+    exec(`${AWF_CMD} --allow-domains ${ALLOWED_DOMAIN} --log-level error -- echo warmup`, { stdio: "ignore" });
+  } catch {
+    // warmup
+  }
+
+  for (let i = 0; i < ITERATIONS; i++) {
+    cleanup();
+    const ms = timeMs(() => {
+      exec(`${AWF_CMD} --allow-domains ${ALLOWED_DOMAIN} --log-level error -- echo ok`, { stdio: "ignore" });
+    });
+    values.push(ms);
+    console.error(`    Iteration ${i + 1}/${ITERATIONS}: ${ms}ms`);
+  }
+
+  return { metric: "container_startup_warm", unit: "ms", values, ...stats(values) };
+}
+
+function benchmarkHttpsLatency(): BenchmarkResult {
+  console.error("  Benchmarking HTTPS latency through Squid...");
+  const values: number[] = [];
+
+  for (let i = 0; i < ITERATIONS; i++) {
+    cleanup();
+    try {
+      // Use curl's time_total to measure end-to-end HTTPS request latency
+      const output = exec(
+        `${AWF_CMD} --allow-domains ${ALLOWED_DOMAIN} --log-level error -- ` +
+          `curl -fsS -o /dev/null -w '%{time_total}' https://${ALLOWED_DOMAIN}/zen`
+      );
+      const seconds = parseFloat(output);
+      if (!isNaN(seconds)) {
+        values.push(Math.round(seconds * 1000));
+      }
+    } catch {
+      console.error(`    Iteration ${i + 1}/${ITERATIONS}: failed (skipped)`);
+      continue;
+    }
+    console.error(`    Iteration ${i + 1}/${ITERATIONS}: ${values[values.length - 1]}ms`);
+  }
+
+  if (values.length === 0) {
+    values.push(0);
+  }
+
+  return { metric: "squid_https_latency", unit: "ms", values, ...stats(values) };
+}
+
+function benchmarkMemory(): BenchmarkResult {
+  console.error("  Benchmarking memory footprint...");
+  const values: number[] = [];
+
+  for (let i = 0; i < ITERATIONS; i++) {
+    cleanup();
+    // Start containers, measure memory, then stop
+    try {
+      // Run a sleep command so containers stay up, then check memory
+      const output = exec(
+        `${AWF_CMD} --allow-domains ${ALLOWED_DOMAIN} --log-level error --keep-containers -- ` +
+          `echo measuring_memory`
+      );
+      // Get memory stats for both containers
+      const squidMem = exec(
+        "sudo docker stats awf-squid --no-stream --format '{{.MemUsage}}' 2>/dev/null || echo '0MiB'"
+      );
+      const agentMem = exec(
+        "sudo docker stats awf-agent --no-stream --format '{{.MemUsage}}' 2>/dev/null || echo '0MiB'"
+      );
+
+      // Parse memory values (format: "123.4MiB / 7.773GiB")
+      const parseMb = (s: string): number => {
+        const match = s.match(/([\d.]+)\s*(MiB|GiB|KiB)/i);
+        if (!match) return 0;
+        const val = parseFloat(match[1]);
+        const unit = match[2].toLowerCase();
+        if (unit === "gib") return val * 1024;
+        if (unit === "kib") return val / 1024;
+        return val;
+      };
+
+      const totalMb = Math.round(parseMb(squidMem) + parseMb(agentMem));
+      values.push(totalMb);
+      console.error(`    Iteration ${i + 1}/${ITERATIONS}: ${totalMb}MB (squid: ${squidMem}, agent: ${agentMem})`);
+    } catch {
+      console.error(`    Iteration ${i + 1}/${ITERATIONS}: failed (skipped)`);
+    }
+    cleanup();
+  }
+
+  if (values.length === 0) {
+    values.push(0);
+  }
+
+  return { metric: "memory_footprint_mb", unit: "MB", values, ...stats(values) };
+}
+
+function benchmarkNetworkCreation(): BenchmarkResult {
+  console.error("  Benchmarking Docker network creation...");
+  const values: number[] = [];
+
+  for (let i = 0; i < ITERATIONS; i++) {
+    const netName = `awf-bench-net-${i}`;
+    try {
+      execSync(`sudo docker network rm ${netName} 2>/dev/null`, { stdio: "ignore" });
+    } catch {
+      // may not exist
+    }
+    const ms = timeMs(() => {
+      exec(`sudo docker network create --subnet=172.${31 + i}.0.0/24 ${netName}`, { stdio: "ignore" });
+    });
+    values.push(ms);
+    console.error(`    Iteration ${i + 1}/${ITERATIONS}: ${ms}ms`);
+    try {
+      execSync(`sudo docker network rm ${netName} 2>/dev/null`, { stdio: "ignore" });
+    } catch {
+      // best-effort cleanup
+    }
+  }
+
+  return { metric: "docker_network_creation", unit: "ms", values, ...stats(values) };
+}
+
+// ── Main ───────────────────────────────────────────────────────────
+
+async function main(): Promise<void> {
+  const commitSha = exec("git rev-parse HEAD");
+  console.error(`AWF Performance Benchmark`);
+  console.error(`  Commit: ${commitSha}`);
+  console.error(`  Iterations: ${ITERATIONS}`);
+  console.error("");
+
+  const results: BenchmarkResult[] = [];
+
+  results.push(benchmarkNetworkCreation());
+  results.push(benchmarkWarmStart());
+  results.push(benchmarkColdStart());
+  results.push(benchmarkHttpsLatency());
+  results.push(benchmarkMemory());
+
+  // Final cleanup
+  cleanup();
+
+  // Check for regressions against critical thresholds
+  const regressions: string[] = [];
+  for (const r of results) {
+    const threshold = THRESHOLDS[r.metric];
+    if (threshold && r.p95 > threshold.critical) {
+      regressions.push(
+        `${r.metric}: p95=${r.p95}${r.unit} exceeds critical threshold of ${threshold.critical}${r.unit}`
+      );
+    }
+  }
+
+  const report: BenchmarkReport = {
+    timestamp: new Date().toISOString(),
+    commitSha,
+    iterations: ITERATIONS,
+    results,
+    thresholds: THRESHOLDS,
+    regressions,
+  };
+
+  // Output JSON to stdout
+  console.log(JSON.stringify(report, null, 2));
+
+  if (regressions.length > 0) {
+    console.error("");
+    console.error("⚠️  Performance regressions detected:");
+    for (const r of regressions) {
+      console.error(`  - ${r}`);
+    }
+    process.exit(1);
+  } else {
+    console.error("");
+    console.error("✅ All metrics within acceptable thresholds.");
+  }
+}
+
+main().catch((err) => {
+  console.error("Benchmark failed:", err);
+  process.exit(1);
+});