Web Agents Python Packages

agentlab

AgentLab: An open-source framework for developing, testing, and benchmarking web agents on diverse tasks, designed for scalability and reproducibility.

3K 574 112

uground-demo-test

[ICLR'25 Oral] UGround: Universal GUI Visual Grounding for GUI Agents

1K 312 18

clawbench-eval

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

1K 162 10

claw-harness

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

1K 162 10

clawbench-harness

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

1K 162 10

nail-clawbench

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

1K 162 10

openclawbench

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

1K 162 10

clawbench-cli

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

773 162 10

claw-ai

ClawBench: Can AI Agents Complete Everyday Online Tasks? (alias of claw-bench)

252 162 10

task-harness

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

246 162 10

claw-eval

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

235 162 10

claw-agent

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

226 162 10

mcq-bench

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

216 162 10

harness-hub

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

215 162 10

life-bench

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

212 162 10

nail-eval

ClawBench: Can AI Agents Complete Everyday Online Tasks? (alias of claw-bench)

211 162 10

r2-harness

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

210 162 10

r2agent

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

210 162 10

nail-agent

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

209 162 10

nail-group

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

208 162 10

doomarena-taubench

DoomArena is a Framework for Testing AI Agents Against Evolving Security Threats

208 58 6

nail-bench

ClawBench: Can AI Agents Complete Everyday Online Tasks? (alias of claw-bench)

207 162 10

harnessos

Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.

206 162 10

everyday-bench

ClawBench: Can AI Agents Complete Everyday Online Tasks? (alias of claw-bench)

202 162 10