20 years ago at Bell Labs, we spent a lot of time reviewing Jini and JXTA. Today, we started Bad Monkey, Inc and fixed them. It's called AgentSpaces, P2P fabric for agent and service collaboration. #Embabel, #SpringBoot, #SpringAI integration. https://t.co/KGb5Xpn4Wh
@starbuxman Surprised you didn’t include AgentSpaces and revitalizing Jini/jxta style workloads, lots of interesting integration with Spring AI and Spring Boot… I will have a blog for you later in the week. https://t.co/7UWOx6Gwe5
At Bad Monkey, we're working on intelligent event processing, leveraging AI on all those real-time events, so you can react to what's happening now.. not later, not after the data winds up in the cloud, and a job gets run. We started with ISR and made an automated tracking solution that is 100-10,000 faster than others. Can't wait to show Project Agincourt and @SecWar what we're doing with C2, autonomy, and UxS.
@glaforge@ApacheGroovy@antigravity Likewise .. we felt that the ephemeral nature of agents, the need to share, the potential for people to share their agents working together, it was time to bring this back
@jasonlk Yes, and how to keep those agents collaborating and sharing as a team... sorry fellas, you get cubicles, meeting rooms, and file cabinets too! https://t.co/KGb5XpnCLP
AgentSpaces 0.3.0 is out: shared objects for fleets of AI agents, peer to peer, no broker. 9,000 downloads in week one. LangChain4j, Spring AI, Micronaut, Python, TypeScript, all updated.
New: fan-in joins, votes in three methods, running totals, reducers, tags support. Peering protocol 10x efficiency.
https://t.co/UQxsgADpnd
In our tactical edge event processor, we're doing track fusion - built using real-time Java patterns. 5000 concurrent tracks, <50MB of memory, high accuracy calculations. Comparing that to the open source standard (in python), it OOMs. I don't think people understand the rigor of Java
@HarryStebbings@jasonlk@rodriscoll Data point on the agents: Bad Monkey released AgentSpaces: 1100+ downloads and 28 companies within the first day... Week 2 is going to be off the hook at #AUSA2026
By end of day 1 of AgentSpaces we had 28 companies, 1100+ downloads, rocketed up near the top of popular open source packages on Maven Central. By end of week 1, added support for LangChain and Micronaut. Next week, will be plugins for Claude and Codex, and integration with the Bad Monkey intelligent event processor (IEP). https://t.co/UQxsgACRxF -- come talk to us at #AUSA2026 and learn more!
@levie If you like that, you'll love Agents working with other agents as peers, in a P2P fabric - whether they're sharing files, building a qorum, voting, etc. We're bringing P2P architectures back! 😎 https://t.co/KGb5Xpn4Wh
Couple more ways to add peers into your AgentSpaces P2P fabric. agentspaces-micronaut, and inspired off of the langchain work there, also agentspaces-langchain4j. See https://t.co/EK0LhK46Yy and https://t.co/DNUjAPmdaF
@MehulC@graemerocher@graalvm throw one more on the Micronaut release party: just released AgentSpaces support for Micronaut, including bringing langchain4j and MCP tools into the peer-to-peer network: https://t.co/EK0LhK46Yy -- will have more docs shortly
@vipulved@OpenRouter We should chat about AgentSpaces and Together, wire up your enterprise with Spring AI, Micronaut, Embabel, etc into PeerGroups for the different spaces in your business (yes, we fixed Jini/JXTA for the Agentic era) https://t.co/2BZftQIFlZ
@andrewchen Yes, that's why we built AgentSpaces, a P2P Collaboration Fabric for the agents, no centralization, no coordination required, the P2P groups form, they can share data, coordinate, vote, etc. Only a few of these and you are at 5 9's of reliability. https://t.co/2BZftQI7wr
@reflection_ai Would love to try that out with AgentSpaces, we just released it - peer-to-peer coordination fabric, LLM can be above or below the peer groups, depending where you want the cache/re-use to occur. We're big proponents of open standards. https://t.co/2BZftQIFlZ
If you'd like to power them up with some peer-to-peer goodness, fault tolerance, self-healing, voting/qorum, cost optimization, even learning across peers, take a look at AgentSpaces, we just released -- bringing the P2P protocols from the 2000s into the modern era (Jini, JXTA), built for Agents: https://t.co/I8ce8ETtXT