Collected molecules will appear here. Add from search or explore.
A framework for General Computer Control (GCC) that enables AI agents to interact with any software or game via visual input and keyboard/mouse control, featuring long-term memory and skill curation.
stars
2,490
forks
261
Cradle is a high-signal project from BAAI (Beijing Academy of Artificial Intelligence) that tackles the 'General Computer Control' (GCC) problem. With nearly 2,500 stars and significant fork activity, it has established itself as a leading open-source research framework for vision-based agents. Its primary moat lies in its 'Skill Library' and 'Self-Improvement' loops, which allow the agent to move beyond simple script execution to long-term task mastery (demonstrated in complex environments like Red Dead Redemption 2). However, the defensibility is under extreme pressure from frontier labs; Anthropic’s 'Computer Use' API and OpenAI’s rumored 'Operator' project provide native model capabilities that threaten to make high-level scaffolding frameworks like Cradle redundant. While Cradle offers better modularity and transparency than closed-source APIs, the 'Platform Domination Risk' is high because OS owners (Microsoft, Apple) are likely to integrate similar capabilities directly into the kernel or UI layer. The 6-month displacement horizon reflects the rapid pace at which VLM-native UI control is evolving, potentially moving from external mouse/keyboard simulation to direct semantic UI tree manipulation.
TECH STACK
INTEGRATION
library_import
READINESS