diff --git a/README.md b/README.md
index 2616d5d..4f33277 100644
--- a/README.md
+++ b/README.md
@@ -37,6 +37,7 @@ Save hours of research time—get straight to what matters.
 | Benchmark | Link | Year | Highlight | Data Size | Metric | Measurement | Platform |
 | --- | --- | --- | --- | --- | --- | --- | --- |
 | MiniWoB++ | https://github.com/Farama-Foundation/miniwob-plusplus | 2017 | Evaluates agents on basic web interactions like clicking, typing, and form navigation. | 100 web interaction tasks | Task Success Rate | Element Match | Web |
+| ClawBench | https://github.com/TIGER-AI-Lab/ClawBench | 2026 | Tests whether agents can complete everyday online tasks (purchases, bookings, job applications, email) on live production websites, using a Chrome-extension + CDP layer that intercepts only the final write request for safe end-to-end runs. | 153 tasks across 144 live websites, 15 categories | Task Success Rate | Execution-based (blocked final submission) | Web |
 | RUSS | https://github.com/xnancy/russ | 2021 | Uses ThingTalk for mapping natural language to web actions, enabling precise web-based task execution in real HTML environments. | 741 instructions | Task Success Rate | Text Match, Element Match | Web |
 | WebShop | https://webshop-pnlp.github.io | 2022 | Simulates e-commerce navigation with real-world products, challenging agents with instruction comprehension, multi-page navigation, and strategic exploration. | 12,087 instructions | Task Success Rate, Step Success Rate* | Text Match | Web |
 | Mind2Web | https://github.com/OSU-NLP-Group/Mind2Web | 2023 | Tests adaptability on real-world, dynamic websites across domains. | 2,000 tasks | Step Success Rate, Task Success Rate | Element Match, Action Match | Web |