Early-Signal Detector

architecture · first seen 4 Oct 2026

multi-agent LLM system

Also: MAS, MALS, LLM-MAS, agent team, LLM swarms, agent swarm, multi-agent, LLM agent teams

Multiple LLM agents coordinating on a task by exchanging messages or states.

Multi-agent LLM systems split a task across several language-model agents that exchange messages, shared state or memory, often with assigned roles such as planner, builder or critic. The sources show the idea applied widely rather than introduced: papers on container placement, trading-agent persona selection, game scene generation, pathology slide analysis and ITOps observability, plus work on shared memory pollution and coordination overhead in agent teams. A Meta-affiliated paper proposes persistent, git-like shared work across agents. Repositories include DeepSeek's harness prototypes, local-first agent platforms and a pentest agent swarm. The weekly counts are already declining, consistent with a broad, mature topic rather than a new idea.

The open questions now sit in narrower problems such as memory governance, team composition and the cost of scaling agent counts, and those sub-topics are more likely to be the real early signals. The umbrella term itself has been mainstream since roughly 2023.

Analyst note (established), 17 Nov 2026.

Mentions per week

03619 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Rank 43 on 14 Nov 2026 · 147 mentions in 28 days · 145 independent groups · papers → code → discussion · first seen 6 weeks ago · acceleration -0.49

Earliest sources

Latest sources

Mentioned together with

model context protocolretrieval-augmented generationagent memoryprompt injectionhuman-in-the-loopReAct promptingagent harness evolutionautonomous penetration testingautomated machine learninghallucination detection