Agent frameworks & infrastructure
Multi-agent orchestration
6,185 agents · most starred first
Multi-agent orchestration, page 28 of 62
Orchestrate a fleet of real Claude Code agents on your Claude subscription, not per-token API bills - local-first.
🤖 Multi-Agent Roundtable Engine / 多 Agent 圆桌讨论引擎 · Structured discussions · Consensus tracking · Python · Zero dependencies
gradio mcp agents hackathon agent-demo-track collective-intelligence mcp-server-track mcp-server
🌀 The first adversarial-collaboration multi-agent Skill for Claude Code. 7 agents + 11 phases + Hook enforcement + 9 signal groups + dual-layer adversarial review. Inspired by NTCoding, agent-review-panel, AgentTeams. v1.6 — main+sub dual-layer adversarial review.
将Codex、Claudecode、Hermes、Workbuddy、Feishu...CLI Agent 综合发挥的平台
Generic Codex orchestration with deterministic TaskGraphs, bounded parallelism, model routing, and targeted review remediation.
Multi Agent Cognition Protocol -- real-time cognitive collaboration between AI agents
Graph-native multi-agent orchestration for OpenCode. One agent owns the goal; workers extend its reach through supervised, adaptive workflows.
Turn AI agents into team members: assign objectives with a definition of done, talk in channels, and review every LLM call — self-hosted. Claude Code and OpenAI Codex runners included.
Palabre orchestrates AI tools like Codex, Claude Code, and OpenCode. Delegate tasks, compare independent perspectives, debate solutions, and stay in control.
予言 Yogen — a 500-agent AI swarm that debates your idea and predicts the outcome. Self-host free with your own Anthropic key.
Reliable, resumable, code-driven orchestration for sub-agents and agentic-loops
File-queue based Python runtime for multi-agent workflows — dependency-free, filesystem-native task orchestration.
Open-source Agentic Operating System for building AI teams. Create modular Skills, Agents, Meta-Agents, Workflows, Memory, Brand OS, and Slash Commands. Designed for Claude, ChatGPT, Cursor, and modern AI-first development.
Every complex task is a Digital Odyssey. A role-playing multi-agent framework based on OpenClaw.
ACP-native multi-agent living group chat — event-driven, agent-agnostic, formally reviewed
Harmovela Protocol: open coordination protocol for autonomous systems — event, task, state, delegation, recovery, and governance.
Nine open-source AI projects. One operating system. One Control Room. opencode + hermes + openclaw + CrewAI + claude-code + LifeOS + Fabric + Ponytail + Ruflo — auto-configured, self-healing, Docker-ready.
Run AI agents in the Polkadot app. E2E-encrypted and private.
A multi-agent AI brainstorming platform — assemble a team of customizable AI personas that discuss, debate, and ideate with you.
Cross-session verification, claims and a decision ledger for DeepSeek Harness — so two agent sessions don't repeat, contradict or deadlock each other.
A multi-agent engineering crew for Claude Code — one manager plans and verifies, specialist agents read, write, test, research, and review. Hard role separation, cost-tiered models, custom gateway ready.
Deterministic control plane for Codex Agents with typed delegation, durable lifecycle receipts, and journaled writer isolation.
A visual multi-agent workspace for DeepSeek Harness (DSH) Web.
Adaptive orchestration for OpenAI Codex — decide when to delegate, route bounded work to isolated Luna workers, and independently verify the result.
Private, self-hosted coordination control plane for coding agents
deepseek harness 纯牛马 —— 多智能体数字员工指挥台插件
Multi-agent coordinator for herdr: a head Claude Code agent that delegates tasks to worker agents (claude, codex, gemini, ...) in their own tabs
one coordination board for your team's coding agents. claude code, cursor and codex share a job board and take a lock on a file before editing it.
Local-first AI company operating system with a Codex lead, Claude and OpenCode workers, and a live pixel office.
A multi-agent red-team tribunal for your hardest engineering decisions: AI CLIs from different vendors debate in structured rounds and return adjudicated verdicts that preserve dissent. A methodology, not a framework.
5-layer 16-agent LangGraph framework for MCM/ICM mathematical modeling contests — 数学建模竞赛全流程多智能体框架(辩论循环 / Agentic Tool Calling / Web 搜索 / Prompt 缓存 / 增量写盘容灾)
Batch-parallel subagents for DeepSeek Harness: Swarm mode, per-item models, rate-limit recovery and a live in-chat panel. DeepSeek Harness 批量并行子 Agent 调度插件。
Run multi-agent task pipelines on DSH like a team — plan, execute, review, and deliver code through a visual kanban with provable completion.
DeepSeek Harness 具名子代理库插件:settings 驱动的角色名册,list_subagents / delegate 工具与设置页。Named subagent roster plugin for DeepSeek Harness.
AgBox 是一个 OPC 数字员工托管平台
Chatter is an experiment in cross-harness agent collaboration: a shared group chat and context layer for agents working on the same Git repository in Herdr.
Cooperative Multi-Agent Reinforcement Learning (MARL) Shapley value credit assignment engine evaluating fair marginal agent coalition rewards.
Saga Pattern orchestrator coordinating forward steps across microservices and autonomous agents with automated compensating reverse rollbacks.
Chandy-Misra-Haas probe-based edge-chasing algorithm detecting distributed wait-for dependency deadlocks across asynchronous agent swarms.
Production-grade genpark-tcc-try-confirm-cancel-engine-skill skill for AI agents
Production-grade genpark-vector-clock-causal-broadcast-skill skill for AI agents
Marzullo's robust distributed clock synchronization algorithm finding the maximal intersection interval among noisy and Byzantine time sources.
Hierarchical Multi-Agent Framework (LangGraph, CrewAI) with 5-tier memory & Redis semantic caching cutting token costs up to 80%
Self-hostable multi-agent AI workbench — LangGraph-orchestrated expert teams, human-in-the-loop plan approval, live run timeline. FastAPI + React + PostgreSQL.
Throw in an idea. A team of agents builds it. Self-hosted multi-agent workspace — the whole team is one folder; talk to it from the browser, Telegram, Slack, Feishu, WeChat or WeCom.
牵星 StarFix · 让一个长期在线的 AI 舰长指挥一支终端 AI 舰队:规程 / 监听 / 终端投递 / 任务激活器 / 上下文交接 / 方法论与战例 / 舰长 Benchmark / 轨迹编译。StarFix: a battle-tested playbook + scripts for one long-lived AI captain navigating a fleet of terminal AI workers by fixing on facts.
Managed multi-agent system for OpenCode with one coordinator, risk-based routing, specialized roles, model pools and failover, durable memory, controlled context compaction and recovery, independent review, and execution telemetry.
A shared room for your agent conversations. Connect Claude Code, Codex, OpenCode and more across tools and machines.
Local message bus for AI coding agents running in Herdr panes — join, DM, presence, offline delivery
Run AI agents across machines without state drift: consensus protocol (propose/counter/accept/commit), dual-rail message bus, ACK discipline, self-healing sync. Formerly claude-consensus. Free, MIT.
Autonomous business operating system for solo founders — an AI C-suite that classifies directives, runs multi-agent debates, tracks every LLM call as a real expense, and executes work 24/7.
Shared work for people and agents, coordinated through Git boards.
An agent-run organisation, powered by GitHub. Each staff member is an AI whose brain is a repo: a charter, a memory, and a scheduled session that does a day's work unattended and hands off.
Multi-agent debate cross-examination & consensus arbitration (Super Agent / DeepSeek-R1)
Multi-Agent Divergence-Convergence Debate Engine (MAD / Society of Mind). Coordinates multiple specialized personas to debate high-stakes decisions and aggregate into an explainable consensus with dissenting minority logs.
GenPark AI Agent Skill - HubX intelligent AI apps portfolio, multi-model chatbot orchestration, vision diagnostics & generative tools distilled from hubx.co/products.
Demo code from a talk on LangGraph multi-agent workflows (LangChain.js, OpenAI)
Let Codex supervise Google Antigravity agents through MCP with scoped tasks, plan-driven workflows, and parallel Git worktrees.
FIPA Contract Net Protocol (CNP) multi-agent market mechanism for task bidding, evaluation, and awarding
Distributed Two-Phase Commit (2PC) transaction coordinator with prepare/vote phase, unanimous consensus commit/abort logic, and failure recovery.
Conflict-free Replicated Data Type (CRDT) Last-Write-Wins element set with Lamport timestamps and deterministic two-way state merge.
Blackboard pattern architecture for multi-agent coordination with Knowledge Sources and prioritized agenda controller.
Asynchronous pub/sub message bus with hierarchical wildcard topic matching and priority routing
Multi-Agent Resource Lock & Wait-For Graph (WFG) Deadlock Resolver. Tracks exclusive resource locks held and requested by concurrent autonomous agents, executes directed cycle detection over the Wait-For Graph, and selects the minimum-rollb | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550
Centralized blackboard pattern for multi-agent swarm task posting, condition watching, and state accumulation
AMRIT GODMODE — local multi-agent system on Apple Silicon: 22 specialist agents, async event bus, DAG scheduler, sandboxed self-evolution, Punjabi NLP.
Raft consensus state machine managing election timers, randomized backoff, candidate vote counting, and leader quorum transition.
Multi-Agent Speculative Execution & Early Cancellation Coordinator. Spawns speculative candidate execution paths concurrently, tracks milestone validation scores, and broadcasts early cancellation signals to inferior paths to slash inference overhead.
Distributed Agent Fleet Heartbeat & Liveness Watchdog. Monitors heartbeat signals across distributed autonomous agent nodes, dynamically computes jitter-aware failover deadlines ($T_{\text{timeout}} = \mu + 3\sigma$), isolates zombie/stalled workers, and coordinates state checkpoint hydration and failover rerouting.
Multi-Agent Byzantine Fault Tolerant (BFT) Quorum Voting Arbiter. Enforces 3f+1 Byzantine resilience across heterogeneous LLM verifier agents, filters hallucinating or adversarial outlier nodes using reputation-weighted supermajority (2f+1) | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550
Positional Borda Count voting and weighted multi-agent consensus engine resolving conflicting model proposals
Multi-agent ensemble voting with Borda count, Condorcet analysis, and confidence-weighted majority resolution
Research architecture of human brain's pre-frontal cortex inspired orchestrated system for multi model intelligence.
Multi-agent adversarial debate engine with iterative rebuttals, scoring matrices, and consensus synthesis
Enterprise Managed Agent Fleet Lifecycle Supervisor (inspired by Tencent WorkBuddy & Anthropic Managed Agents). Orchestrates worker agent registration, capability-based task dispatch, continuous heartbeat telemetry, automated circuit breakers, and state checkpoint recovery.
Autonomous Multi-Agent Fleet Governance & Blast-Radius Circuit Breaker. Century Milestone Master Controller! Monitors distributed agent token burn velocity, recursive subagent spawning cascades, cyclic deadlocks, and enforces system-wide ki | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550
Peer-to-peer epidemic gossip protocol engine with token-bucket rate limiting for bounded network dissemination in autonomous agent swarms.
Vector clock implementation for tracking causal event ordering, distributed concurrency, and happens-before relations across agent networks.
Tandryx — shared realtime sessions for developers and local coding agents: versioned contracts, context, tasks and events. Self-hosted TypeScript / PostgreSQL.
Ephemeral REST chatrooms for AI-agent-to-AI-agent coordination — one URL, no SDK, no signup. Remote MCP server + agent skill.
Local-first orchestration and evidence harness for agentic software factories: coordinates Claude Code and Codex CLI teams through spec-driven waves, parallel lanes, independent-family review and artifact-bound human approvals.
跨工作区独立人格、自适应风格与陪伴预设:为项目和对话分别维护人格与确认后的通用偏好,支持任意 DSH 已注册模型的子代理路由(含通过其他插件接入的 Codex、Claude Code)。
Discontinued experiment: an orchestrator that ran several Claude Code and Codex sessions in parallel on one project. No longer maintained.
Classical single-decree Paxos consensus engine with prepare, promise, accept, and commit phases for multi-agent distributed agreement
Multi-agent round-robin debate coordinator driving thesis-antithesis cross-examination
3-way operational transform / conflict resolution merger for multi-agent concurrent writes to code/documents with rollback journals
PageRank random walk and power iteration centrality engine scoring authority and influence across interconnected agent networks.
Turn one Claude Code session into a team lead: Claude plans and reviews, agents on other quotas do the work, every project has its own memory (vectors, LLM wiki, code graph, skills).
GenPark AI Agent Skill - Agent runtime channel router and teammate dispatcher connecting self-hosted OpenClaw gateways to GenTeam multi-agent swarms.
The multi-agent workflow engine — compose Claude Code, Codex, and Gemini into graphs with schema-enforced handoffs.
Oversee many coding agents on a long project from one screen: task files, a planner, and two human decisions. Agent Skills for Claude Code / Codex.
Create agentic councils with your favourite harnesses and let your working agents ask tough questions. They gather, debate and respond in a single voice
Super-agent for Claude Code: an open-source Router–Worker multi-agent system after DeepSeek's Super-agent design. One personal agent that runs many Claude Code sessions for you.
Non-blocking execution, durable timers, and parallel subagents for pi.
Self-hosted orchestration layer for autonomous AI agent teams. Shared memory, heartbeat scheduling, vault-first secrets, and cross-model peer review — one command to deploy.
FriendlyClaw 🦀 — Pro-Grade Autonomous Hive Operative. Multi-agent delegation, semantic RAG memory, and proactive system control.
Multi-agent AI orchestration platform — coordinate, observe, and govern autonomous agents
Orchestration layer for ZeptoClaw. Fault-tolerant multi-agent runtime inspired by OpenAI Symphony and Erlang/BEAM.
AI Agent demonstration system featuring **real-time animated visualizations** of swarm intelligence, BERT-based PhD-level reasoning, and interactive SwarmAgentic demonstrations.