diff --git a/.gitignore b/.gitignore
index d84365bb07..0ab773c153 100644
--- a/.gitignore
+++ b/.gitignore
@@ -25,3 +25,7 @@ cli/tests/.test_coverage/
 # MacOS generated files
 .DS_Store
 .DS_Store?
+
+# Flamegraphs
+/flamebench*.svg
+/flamegraph*.svg
diff --git a/tools/README.md b/tools/README.md
index c408e78c5f..7bdf6f77a1 100644
--- a/tools/README.md
+++ b/tools/README.md
@@ -30,3 +30,32 @@ executable
 ```sh
 cargo run -- run --allow-read --allow-write --allow-run --unstable ./tools/<script>
 ```
+
+## flamebench.js
+
+`flamebench.js` facilitates profiling and generating flamegraphs from
+benchmarks.
+
+General usage:
+
+```
+❯ ./tools/flamebench.js
+flamebench <bench_name> [bench_filter]
+
+Available benches:
+op_baseline
+ser
+de
+```
+
+To profile the `op_baseline` bench, run `./tools/flamebench.js op_baseline`,
+this will run all 3 benches in `op_baseline.
+
+Often when profiling/optimizing, you'll want to focus on a specific sub-bench,
+`flamebench` supports a bench/test filter arg like the regular cargo commands.
+So you can simply run `./tools/flamebench.js op_baseline bench_op_async` or
+`./tools/flamebench.js op_baseline bench_op_nop` to profile specific benches.
+
+Tip: the `[bench_filter]` argument doesn't have to be an exact bench name, you
+can use a shorthand or a partial match to profile a group of benches, e.g:
+`./tools/flamebench.js de v8`
diff --git a/tools/flamebench.js b/tools/flamebench.js
new file mode 100755
index 0000000000..bd409c86c4
--- /dev/null
+++ b/tools/flamebench.js
@@ -0,0 +1,119 @@
+#!/usr/bin/env -S deno run --unstable --allow-read --allow-run
+// Copyright 2018-2021 the Deno authors. All rights reserved. MIT license.
+import { join, ROOT_PATH as ROOT } from "./util.js";
+
+async function bashOut(subcmd) {
+  const p = Deno.run({
+    cmd: ["bash", "-c", subcmd],
+    stdout: "piped",
+    stderr: "null",
+  });
+
+  // Check for failure
+  const { success } = await p.status();
+  if (!success) {
+    throw new Error("subcmd failed");
+  }
+  // Gather output
+  const output = new TextDecoder().decode(await p.output());
+  // Cleanup
+  p.close();
+
+  return output.trim();
+}
+
+async function bashThrough(subcmd, opts = {}) {
+  const p = Deno.run({ ...opts, cmd: ["bash", "-c", subcmd] });
+
+  // Exit process on failure
+  const { success, code } = await p.status();
+  if (!success) {
+    Deno.exit(code);
+  }
+  // Cleanup
+  p.close();
+}
+
+async function availableBenches() {
+  // TODO(AaronO): maybe reimplement with fs.walk
+  // it's important to prune the walked tree so this is fast (<50ms)
+  const prunedDirs = ["third_party", ".git", "target", "docs", "test_util"];
+  const pruneExpr = prunedDirs.map((d) => `-path ${ROOT}/${d}`).join(" -o ");
+  return (await bashOut(`
+  find ${ROOT} -type d \
+      \\( ${pruneExpr} \\) \
+      -prune -false -o  \
+      -path "${ROOT}/*/benches/*" -type f -name "*.rs" \
+      | xargs basename | cut -f1 -d.
+  `)).split("\n");
+}
+
+function latestBenchBin(name) {
+  return bashOut(`ls -t "${ROOT}/target/release/deps/${name}"* | head -n 1`);
+}
+
+function runFlamegraph(benchBin, benchFilter, outputFile) {
+  return bashThrough(
+    `sudo -E flamegraph -o ${outputFile} ${benchBin} ${benchFilter ?? ""}`,
+    // Set $PROFILING env so benches can improve their flamegraphs
+    { env: { "PROFILING": "1" } },
+  );
+}
+
+async function binExists(bin) {
+  try {
+    await bashOut(`which ${bin}`);
+    return true;
+  } catch (_) {
+    return false;
+  }
+}
+
+async function main() {
+  const { 0: benchName, 1: benchFilter } = Deno.args;
+  // Print usage if no bench specified
+  if (!benchName) {
+    console.log("flamebench.js <bench_name> [bench_filter]");
+    // Also show available benches
+    console.log("\nAvailable benches:");
+    const benches = await availableBenches();
+    console.log(benches.join("\n"));
+    return Deno.exit(1);
+  }
+
+  // List available benches, hoping we don't have any benches called "ls" :D
+  if (benchName === "ls") {
+    const benches = await availableBenches();
+    console.log(benches.join("\n"));
+    return;
+  }
+
+  // Ensure flamegraph is installed
+  if (!await binExists("flamegraph")) {
+    console.log(
+      "flamegraph (https://github.com/flamegraph-rs/flamegraph) not found, please run:",
+    );
+    console.log();
+    console.log("cargo install flamegraph");
+    return Deno.exit(1);
+  }
+
+  // Build bench with frame pointers
+  await bashThrough(
+    `RUSTFLAGS='-C force-frame-pointers=y' cargo build --release --bench ${benchName}`,
+  );
+
+  // Get the freshly built bench binary
+  const benchBin = await latestBenchBin(benchName);
+
+  // Run flamegraph
+  const outputFile = join(ROOT, "flamebench.svg");
+  await runFlamegraph(benchBin, benchFilter, outputFile);
+
+  // Open flamegraph (in your browser / SVG viewer)
+  if (await binExists("open")) {
+    await bashThrough(`open ${outputFile}`);
+  }
+}
+// Run
+await main();