In this article i've proposed a simple way to work on multiple tasks in parallel using 1 OpenClaw agent.
The trick is to create multiple groups in Telegram, add you agent there with require_mention=false and work in them.
This way, context for each session is stored separately and you don't need to way for a result on one task to work on another one, while memory is still unified, which allows to cross reference conversations.
we consider this a separate issue that should be addressed using special approaches. however, the proposed technique generally tends to decrease the probability of something being lost in the middle by minimizing the total size of the retrieved documents.
couldn't agree more. It's also correct for other fields: design, analytics, marketing, etc. When cost of implementation goes to zero, what you do and why is the most important thing.
The trick is to create multiple groups in Telegram, add you agent there with require_mention=false and work in them.
This way, context for each session is stored separately and you don't need to way for a result on one task to work on another one, while memory is still unified, which allows to cross reference conversations.