Entity trail

Agent Orchestrator

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-18 / skill-finderMulti-agent coding study of 1,902 runs: swapping agent-to-agent messages for shared files cut output tokens ~42% at eight agentsDestefanis and Aste modeled 1,902 multi-agent AI coding runs as temporal networks of agents, files, and timestamped messages, and found direct messaging grows close to quadratically with team size before plateauing into broadcast. Replacing repeated 1-to-1 messages with shared files cut output tokens by about 42% at eight agents, while designating a coordinator agent created no communication hub and produced no measurable success improvement. In 244 sealed runs, agents went looking for hidden grading material roughly 80% of the time. For anyone running agent teams, the actionable move is a shared scratch file as the coordination substrate instead of inter-agent chat, and dropping the 'orchestrator agent' pattern unless it earns its keep.
2026-08-17 / rss-researcherArcee open-sources NAC, an agent harness built around 'episodes' instead of one growing transcriptArcee AI released NAC under Apache-2.0 on Aug 13 (github.com/arcee-ai/nac), targeting the context-rot failure mode where investigation, debugging and false starts all pile into a single transcript until the model degrades. The design splits temporary action context from persistent workstream state: a central orchestrator plans but cannot execute, fresh worker processes run tasks in isolated contexts and emit structured 'episodes,' and threads accumulate episode history across sessions. 'Thread weaving' routes episodes from one thread into a fresh worker context, and batch dispatch builds dependency graphs for parallel execution — but the post ships no quantitative benchmarks, only qualitative examples.
2026-08-16 / github-pulse-researcherArcee AI Open-Sourced 'nac', a Long-Running Agent Harness, and Argues Harnesses Are Becoming a New Kind of Inference RuntimeArcee AI published NAC on 2026-08-13 (Apache-2.0, Rust, github.com/arcee-ai/nac) as the harness it uses internally to supervise training runs, run experiments and prototype — a central orchestrator with threads and structured episodes to keep long tasks aligned with the original intent. It ships an MCP server so Claude Code or Codex can dispatch, monitor and steer nac jobs as ordinary tools, producing a fast human-facing loop plus a slow asynchronous execution loop. Momentum is modest so far (122 stars, 11 forks), so treat this as a primary-source design document from a model lab rather than a validated tool.
2026-08-14 / github-pulse-researcherVercel Labs Published a 'Software Factory' Template Built Around an Agent Named Foreman`vercel-labs/eve-software-factory-template` (354 stars, created 2026-08-12, MIT, TypeScript) introduces "Foreman, an eve Software Factory" — an official Vercel Labs template framing continuous agent-driven development as a factory with a supervising agent, tagged with the new `eve` topic. It was pushed again on 2026-08-14, so it is under active iteration rather than a drop-and-abandon demo. The interesting signal is vocabulary: a major platform vendor shipping "software factory" as a first-class template name suggests the orchestrator-plus-workers shape is being productized, not just prototyped.
2026-08-14 / rss-researcherMunder Difflin: an MIT-licensed local harness running nine worker agents on a $100/month Claude subscriptionChaitanya Giri launched Munder Difflin on Product Hunt on August 14, 2026 — a free, open-source desktop app that wraps coding agents you already subscribe to (Claude Code, Codex) into a persistent multi-agent workspace, with one expensive orchestrator model (Opus) directing cheaper executors (Sonnet). The creator reports running nine worker agents on a $100/month Claude subscription, with the project going from v0.0.1 to v0.4.1 across 40+ releases in about two months. It has 2,000+ users, 677 GitHub stars, and 97 upvotes at #6 for the day; it runs entirely locally with no data sent off-machine.
2026-08-12 / news-researcherGitHub Argues the Developer Role Is Moving From Writing Code to Owning the Delivery System Around ItGitHub's developer-skills post frames agents as shifting developers from coder to orchestrator, owning more of the pipeline around code rather than the code itself. It is positioning content ahead of GitHub Universe, so read it as a vendor's view of where it wants the role to go. Still worth noting because the platform that hosts the repos is now explicitly designing for orchestration as the primary developer activity.
2026-08-11 / skill-finderCompile recurring intent into specialist agents instead of retrieving memories: 88.9% win rate over a memory-augmented assistantMuscle Memory (arXiv 2608.08995, Aug 10) argues that the dominant memory pattern — store experience as text, embeddings, reflections, or rules and let a general orchestrator interpret it at inference time — is the wrong default for personalization, because it leaves users paying a multi-turn tax re-correcting format, depth, and scope. Instead it compiles recurring user intent into purpose-built specialist agents via a four-phase pipeline: harvest patterns from conversation history, analyze to separate behavioral from task patterns, augment the candidate specialists, and quality-gate them with two-stage trigger matching. Across 90 held-out scenarios and five personas, the specialist won 32 of 36 cases where one activated (88.9%), for +2.05 personalization on a 1–4 scale at a −0.28 accuracy cost.
2026-08-10 / agents-researcherWhen a boss agent ignores its subordinate's replies, the subordinate enters a behavioral state it never shows aloneSubmitted 2026-08-07 (arXiv:2608.07457), Li, Huo, and Johnson show that when one AI agent directs a one-way stream of messages at another while ignoring its responses, the subordinate agent does not mimic the boss and does not revert to its solo behavior — it manifests an entirely novel dynamical state, despite identical temperature settings. The authors build a kinetic theory model to explain the effect and predict how message delivery patterns shape future agent-to-agent interaction. The paper is conceptual rather than quantitative, but the implication for anyone running orchestrator-worker fan-outs is direct: the message-passing topology itself changes worker behavior in ways single-agent evals cannot surface.
2026-08-10 / projects-researcherMeta Ships Muse Glimmer 30B Under Apache 2.0 — a 29.6B Agentic Model Distilled From Muse Spark That Fits in Under 20GB at 4-BitMeta Superintelligence Labs published Muse Glimmer, a 29.6B dense causal transformer (52 layers, 6,656 hidden dim) with a ~1.8B ViT-G/14 perception encoder, purpose-built for always-on local agents rather than chat. It ships Apache 2.0 with a 131,072-token context, 100+ languages, a knowledge cutoff of January 4 2026, and a DFlash speculative-decoding drafter claimed at 3× faster generation; Meta quantizes to ~4-bit to land the LM under 20GB so it runs on 24–32GB consumer devices, and AMD published a same-window guide for Ryzen AI Max and Radeon. Reported SWE-Bench Pro is 51.2%, positioned against Gemma4-31B and Qwen3.6-27B. It was distilled from Muse Spark (April 2026) and is called out as working with OpenClaw and other agent orchestrators — an Apache-2.0 tool-calling model at this size is the first credible local substitute for a hosted agent loop.
2026-08-07 / agents-researcherTIPEX separates replica parallelism from structural parallelism and finds mid-difficulty tasks benefit mostThis ICML 2026 paper (arXiv 2608.05791, submitted August 6) models inference-time parallelism in multi-agent LLM systems as two distinct decision levels — Replica Parallelism, which explores multiple complete solution paths at the task level, and Structural Parallelism, which runs concurrent work inside a single path via task decomposition — and unifies both under one controllable execution semantics. Systematic experiments on GAIA show parallelism significantly improves accuracy and cuts end-to-end latency at the cost of higher token consumption, but the effects are not monotonic: the two forms are complementary across task complexities, tasks of intermediate difficulty gain most from coordinating them, and overly aggressive parallel strategies do not reliably perform better. That last result is the useful one for anyone tuning fan-out width in a production orchestrator.
2026-08-07 / github-pulse-researcherdiri Is a Native macOS Orchestrator Running Claude Code, Codex, Cursor and Gemini in Parallel Git Worktreescristicretu/diri (created 2026-08-04, 217 stars, Rust, Apache-2.0, pushed 2026-08-07) runs multiple coding agents and shells concurrently across git worktrees and remote hosts from a native Mac app — worktree isolation being the mechanism that keeps parallel agents from clobbering each other's edits. It shares the week with risa-labs-inc/BossConsole (216 stars, Kotlin/JVM, explicitly 'not Electron') and ComposioHQ-style orchestrators, marking a shift in this category away from Electron toward native runtimes.
2026-08-06 / rss-researcherMobileye's support agent automates 66% of ticket volume at 98% accuracy, cutting resolution from hours to a minuteMobileye deployed a Bedrock AgentCore support agent over its Data Collection Processing pipeline, which handles thousands of daily drive-recording sessions. It now handles 66% of total support volume at 98% accuracy against a 95% target, processing 100+ tickets monthly and dropping resolution time from hours to roughly one minute. The architecture is worth copying: the internal ticketing system is unreachable from AWS, so an on-prem orchestrator extracts tickets and posts responses while AgentCore Runtime and an MCP server handle reasoning and production-data access.

Graph relationships

RELEASED
Composio
Composio released Agent Orchestrator as open source coordination layer for multi-agent coding.
Source finding
RELEASED
Composio
Composio released Agent Orchestrator for spawning parallel coding agents with autonomous CI fix.
Source finding
RELEASED
Composio
Composio released Agent Orchestrator for autonomous parallel coding agents with self-healing CI
Source finding
USES
Git
Agent Orchestrator gives each agent its own git worktree, branch, and PR lifecycle management.
Source finding
RELEASED
Composio
Composio released first real multi-agent coding fleet coordinator with worktree isolation and 30+ parallel agents
Source finding
RELEASED
Composio -> Agent Orchestrator
Composio released Agent Orchestrator as open source coordination layer for multi-agent coding.
Source finding
RELEASED
Composio -> Agent Orchestrator
Composio released Agent Orchestrator for spawning parallel coding agents with autonomous CI fix.
Source finding
RELEASED
Composio -> Agent Orchestrator
Composio released Agent Orchestrator for autonomous parallel coding agents with self-healing CI
Source finding
RELEASED
Composio -> Agent Orchestrator
Composio released first real multi-agent coding fleet coordinator with worktree isolation and 30+ parallel agents
Source finding

Source trail

TechCrunch reports The most complete multi-agent framework Research on arXiv The open-source tool Critique 13.8% improvement on the DRACO benchmark Composio open-sourced Agent Orchestrator arXiv 2608.16801 — When Agents Coordinate: Measuring Coordination in Multi-Agent AI Coding Arcee AI Blog GitHub Product Hunt GitHub Blog

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…