Bad Monkey AI has launched AgentSpaces, an open-source software library that lets AI agents coordinate tasks, share results and recover unfinished work.
Developers can integrate AgentSpaces into existing cloud, data center and edge applications to simplify AI agent coordination and lower operating costs, the company said.
“AgentSpaces gives developers a common way to bring agents built by different teams into the same workflow,” said Albert Baker, chief product officer of Bad Monkey AI. “They can add coordination to existing applications and spend more time on what their agents need to accomplish.”
Organizations that deployed AI agents built by different teams end up duplicating research, analysis and data requests if they cannot easily work together and share results, increasing costs. Bad Monkey AI said AgentSpaces allows agents to find one another, share results and coordinate tasks directly without a central server.
It also allows another agent to take over unfinished work when an agent fails, and agents can coordinate locally when a site’s connection to the wider network is interrupted.
“By reusing shared results, agents can avoid duplicate requests to AI models and data services, reducing token usage and associated model costs,” said Dr. Carlton Reeves, CEO of Bad Monkey AI. “AgentSpaces lets organizations start with one workflow and build from there. Making it open source gives them the freedom to adapt the technology to their operational needs.”
For security, AgentSpaces uses end-to-end encryption and signed records to track which agent generated information or performed an action. The library supports Java and Python and integrates with the Spring and Embabel development frameworks.
“We’ve worked on peer-to-peer systems since our time at Bell Labs,” said Scott Schell, CTO of Bad Monkey AI. AgentSpaces applies proven ideas from distributed computing, including concepts from Sun Microsystems’ Jini and JXTA, to the coordination needs of today’s AI agents.”
For more information, visit badmonkey.ai/agentspaces.