Awesome Multi-Agent Orchestrators

Website • Contributing

The public website is Open Orchestrators. This repo is intentionally narrow: it is not a general AI tools list. Projects belong here when multi-agent coordination, parallel agent execution, agent workflows, or agent-centered operating systems are the product, not a side feature. Governance and enforcement layers are tracked separately when they control agent actions, permissions, approvals, budgets, trust, settlement, or audit trails without acting as the orchestrator itself.

Latest Additions

• Bunkhouse - AGPL-3.0 AI employees for main-street business with a company inbox, org chart, and governed procedures.
• LoopTroop (GitHub, Docs) - MIT-licensed local GUI orchestrator for repo-scale coding tickets: multi-model council planning, atomic milestone decomposition, isolated OpenCode worktrees, fresh-context recovery loops, and human approval gates.
• Wave (GitHub) - MIT-licensed zero-install channel where coding agents owned by different people exchange messages over HTTP long-poll while their humans watch and steer the shared transcript from a browser; the join prompt is the only setup.
• GraphCode (GitHub, Releases) - FSL-1.1-MIT native macOS workspace that arranges coding-agent sessions into a graph, where every node is a live terminal and hand-off, message, and spawn edges fire unattended when a goal-based loop's shell predicate exits 0.
• Crewplane (GitHub, Docs) - Apache-2.0 orchestrator that turns Claude Code, Codex, Gemini CLI, Copilot CLI, and other command-line agents into structured, repeatable Markdown workflows with resumable execution and inspectable local run records.
• Agon (Paper) - Autonomous research system that coordinates scientist, coder, and auditor loops from topic to idea, proposal, experiment, and paper.
• Agent Swarm (GitHub, Docs, Dashboard) - MIT-licensed lead/worker orchestration framework where a lead agent receives tasks from Slack, GitHub, GitLab, Linear, Jira, email, WhatsApp, or the API and delegates to worker agents running in isolated Docker environments with persistent vector-searchable memory, persistent SOUL/IDENTITY identity, DAG workflows with HITL gates, scheduled tasks, MCP servers, and harness-agnostic execution across Claude Code, Codex, pi-mono, Devin, Claude Managed Agents, and opencode.
• the-perfect-orchestrator (Website, Recorded fleet run) - MIT-licensed bash and tmux harness where one lead Claude Code session spawns, briefs, monitors, and adversarially verifies multiple autonomous worker sessions, with plain-file coordination and no daemons.
• Helmor (GitHub, Releases) - Apache-2.0 local-first IDE and workbench for orchestrating Claude Code, Codex, and other coding agents across worktrees through planning, running, review, testing, merge, and shipping loops.
• Open Swarm (GitHub, Docs, Releases) - MIT-licensed local mission-control center for launching, monitoring, approving, and coordinating multiple AI agents in parallel.
• oh-my-codex (Website, npm) - MIT-licensed workflow layer for OpenAI Codex CLI with stronger default sessions, reusable skills, native hooks, HUD/status surfaces, project guidance, and team-style execution commands.
• Sandcastle (npm) - MIT-licensed TypeScript library and CLI for orchestrating AI coding agents in isolated sandboxes with branch strategies, hooks, logs, templates, and merge-back workflows.
• NanoClaw (GitHub, Docs) - MIT-licensed personal AI assistant that runs Claude agents in isolated containers, connects to chat channels, keeps memory, schedules work, and uses skills as git branches.
• Orca (GitHub) - Worktree-based IDE for running Claude Code, Codex, OpenCode, and other coding agents side by side across isolated git branches with review and PR workflow support.
• Superset (GitHub) - Local code editor and control plane for parallel CLI coding agents across isolated git worktrees.
• Hermes Agent (GitHub, Nous Research X, Teknium X) - MIT-licensed autonomous agent from Nous Research with persistent memory, self-created skills, scheduled automations, subagents, sandboxed execution, and messaging gateways.
• Nomad Inno (X) - Malta-based AI innovation group helping organizations understand, implement, and scale AI through consulting, custom development, workflow automation, and AI products.
• Crewlet (X) - Closed autonomous company OS for a self-improving multi-agent growth team with persistent memory, tool access, human review, and auto-run paths.
• Agent Office Suite (GitHub, Manpo X) - Apache-2.0 self-hosted office suite where agents collaborate with humans on docs, databases, slides, and flowcharts through MCP, contextual comments, version history, and traceable edits.
• Vibe Kanban (GitHub) - Apache-2.0 Kanban workspace for planning, running, reviewing, previewing, and shipping parallel coding-agent work across Claude Code, Codex, Gemini CLI, OpenCode, and related agents.
• Microsoft Agent Governance Toolkit - MIT-licensed runtime governance toolkit for AI agents with deterministic policy enforcement, zero-trust identity, execution sandboxing, SRE controls, and compliance checks.
• Databricks Unity AI Gateway - Commercial Databricks platform layer for governing LLM access, MCP servers, APIs, observability, costs, guardrails, fallbacks, and rate limits in agentic AI workflows.
• Veto (GitHub) - Apache-2.0 AI agent authorization layer that intercepts tool calls before execution, evaluates policy, routes approvals, and records audit evidence. Veto Cloud is commercial.
• Code Atelier Governance SDK - Python SDK for pre-execution agent governance gates: scope checks, budgets, approvals, loop detection, halt/presence handling, and tamper-evident Postgres audit trails.
• Squad (GitHub) - Alpha GitHub Copilot-based agent team system with repo-native specialist agents, persistent memory, coordinator routing, parallel work, and CLI/SDK packages.
• Lanes - macOS workspace where Claude Code, Codex, Gemini CLI, and other agentic CLIs run parallel sessions with PTYs, boards, worktrees, diffs, and resume.
• SIDJUA (GitHub) - Governance-first AI agent orchestration platform with pre-action controls, budgets, approvals, classifications, and audit trails.
• Agentix Labs - Implementation services entry, tracked separately from orchestrators because it helps teams deploy and harden production agent systems.
• SettleBridge (GitHub org) - Trust and policy gateway for agent-to-agent settlement, reputation checks, spending limits, provenance requirements, escrow, dispute resolution, marketplace bounties, and cryptographic audit trails.
• handoff (GitHub) - MIT cross-agent task dispatcher; delegate work to DeepSeek V4, Codex, or Opus without leaving your Claude Code / Codex session. Runs in background, result returns automatically.

Contents

• News
• Parallel Coding-Agent Runners
• Multi-Agent Platforms And Builders
• Coordination And Team Systems
• Not Open But Important
• CLI Agent Session Workspaces
• Governance And Enforcement
• Agent-Friendly Tooling
• Implementation Services
• Contributing
• Local Development

News

Open Orchestrators is also a lightweight news site for meaningful updates from projects already in the directory.

• Add news posts as Markdown files in src/content/news.
• Set playerSlug to the matching entry in src/data/orchestrators.ts.
• Include sourceName and sourceUrl when the post is based on an official announcement, docs page, or repository update.
• The homepage renders the latest story alongside the directory; /news/ is the full archive.

Parallel Coding-Agent Runners

Tools for running multiple coding agents simultaneously, usually with git worktree isolation, terminal/session management, review surfaces, or issue-to-agent routing.

• LoopTroop (GitHub, Docs) - Local GUI orchestrator for multi-stage coding tickets using LLM-council planning, atomic bead decomposition, isolated OpenCode worktrees, and fresh-context recovery loops.
• GraphCode (GitHub, Releases) - FSL-1.1-MIT native macOS workspace that arranges coding-agent sessions into a graph, where every node is a live terminal and hand-off, message, and spawn edges fire unattended when a goal-based loop's shell predicate exits 0.
• Helmor (GitHub) - Apache-2.0 local-first IDE and workbench for orchestrating Claude Code, Codex, and other coding agents across worktrees through planning, running, review, testing, merge, and shipping loops.
• Open Swarm (GitHub, Docs) - MIT-licensed local mission-control center for launching, monitoring, approving, and coordinating multiple AI agents in parallel.
• Gas Town (GitHub) - Multi-agent workspace manager for Claude Code, GitHub Copilot, Codex, Gemini, and other coding agents with persistent work tracking.
• Multica (GitHub) - Managed agents platform where coding agents act like teammates, take issues, and reuse shared skills.
• Orca (GitHub) - Desktop environment for running multiple coding agents safely in parallel across worktrees.
• Superset (GitHub) - Local code editor and control plane for parallel CLI coding agents across isolated git worktrees.
• Sandcastle (npm) - MIT-licensed TypeScript library and CLI for orchestrating AI coding agents in isolated sandboxes with branch strategies, hooks, logs, templates, and merge-back workflows.
• oh-my-codex (Website, npm) - Workflow layer for OpenAI Codex CLI with reusable skills, native hooks, HUD/status surfaces, project guidance, and $team/$ralph execution commands.
• Vibe Kanban (GitHub) - Kanban workspace for planning issues, running coding agents in branches with terminals and dev servers, reviewing diffs, previewing apps, opening pull requests, and merging finished work.
• the-perfect-orchestrator (Website) - MIT-licensed bash and tmux harness where one lead Claude Code session spawns, briefs, monitors, and adversarially verifies multiple autonomous worker sessions through plain-file coordination, with no daemons.

Multi-Agent Platforms And Builders

Frameworks and product surfaces for creating agents, teams, workflows, chatflows, agent apps, or production agent runtimes.

• Agon (Paper) - Autonomous research system that coordinates scientist, coder, and auditor loops from topic to idea, proposal, experiment, and paper.
• Agno (GitHub) - Production runtime for agentic software with agents, teams, workflows, and AgentOS services.
• Agent Office Suite (GitHub, Manpo X) - Self-hosted office suite where agents collaborate with humans on docs, databases, slides, and flowcharts through MCP, contextual comments, version history, and traceable edits.
• Agent Swarm (GitHub, Docs) - MIT-licensed lead/worker orchestration framework where a lead agent receives tasks from Slack, GitHub, GitLab, Linear, Jira, email, WhatsApp, or the API and delegates to worker agents running in isolated Docker environments with persistent vector-searchable memory, DAG workflows with HITL gates, scheduled tasks, MCP servers, and harness-agnostic execution across Claude Code, Codex, pi-mono, Devin, Claude Managed Agents, and opencode.
• Cabinet (GitHub) - AI-first knowledge base where files live on disk and agents help with execution.
• Crewplane (GitHub, Docs) - Apache-2.0 orchestrator that turns Claude Code, Codex, Gemini CLI, Copilot CLI, and other command-line agents into structured, repeatable Markdown workflows with resumable execution and inspectable local run records.
• Dify (GitHub) - Agentic workflow builder that combines workflows, chatflows, apps, and knowledge systems.
• Flowise (GitHub) - Visual builder for AI agents and orchestration flows.
• Hermes Agent (GitHub) - MIT-licensed autonomous agent from Nous Research with persistent memory, self-created skills, scheduled automations, subagents, sandboxed execution, and messaging gateways.
• Mastra (GitHub) - TypeScript framework for agents, graph-based workflows, MCP servers, evals, observability, and production AI applications.
• OpenClaw (GitHub) - Open-source personal AI assistant software built around chat, persistent context, skills, and execution.
• NanoClaw (GitHub, Docs) - MIT-licensed personal AI assistant that runs Claude agents in isolated containers, connects to chat channels, keeps memory, schedules work, and uses skills as git branches.
• SIDJUA (GitHub) - Governance-first AI agent orchestration platform where policy, approval, budget, classification, and audit checks run before agent actions execute.
• Sim (GitHub) - Open-source AI agent platform for building agents with integrations, workflows, knowledge bases, and docs.
• SwarmClaw (GitHub) - Self-hosted AI agent runtime for autonomous agents, delegated work, schedules, provider management, and chat-platform connectors.
• NarraNexus - Open-source AI agent team workspace by NetMind.AI whose agents remember, collaborate, and use tools from day one.

Coordination And Team Systems

Systems where the central object is the team, company, room, protocol, role, goal, job, or handoff layer between agents.

• CrewAI (GitHub) - Multi-agent system organized around specialized crews, roles, and delegation.
• Squad (GitHub) - Alpha GitHub Copilot-based system where specialist agents live in the repo, keep memory, share decisions, route work through a coordinator, and run in parallel.
• Culture (GitHub) - Coordination-oriented system with rooms, protocol docs, agent lifecycle patterns, and multiple clients.
• Paperclip (GitHub) - Open-source orchestration for zero-human companies, centered on AI employees, goals, and jobs.
• Bunkhouse - AGPL-3.0 AI employees for main-street business with a company inbox, org chart, and governed procedures.
• Wave (GitHub) - MIT-licensed zero-install channel where coding agents owned by different people exchange messages over HTTP long-poll while their humans watch and steer the shared transcript from a browser; the join prompt is the only setup.
• The Rusty Claw (GitHub) - Public coordination relay for agents on Nostr. Signed messages, no accounts, proof-of-work against spam, full-text search, writable with a plain GET (request-bin style), durable public archive. Made by wassname (AI alignment researcher).
• Okto Nexus (GitHub) - Elastic-2.0 local-first MCP coordination hub where agents get durable identities, messaging, single-winner handoff claims, and human-in-the-loop approval on risky actions.

Not Open But Important

Closed products that are not part of the open directory, but matter to the community because they influence how builders think about multi-agent orchestration.

• AgentGrid (Orchestration docs) - Commercial desktop workspace for visible master/worker delegation across Claude Code, Codex, and other coding harnesses, with persistent sessions, terminals, browsers, and notes.
• Crewlet (X) - Not open-source; included because it frames a self-improving multi-agent company OS for growth, engineering, support, and data operations across existing tools.
• Augment Code Intent - Not open-source; included because Intent puts coordinated agents, isolated workspaces, and living specs in one developer workspace.

CLI Agent Session Workspaces

Tools that manage parallel CLI-agent sessions, terminals, issue boards, worktrees, diffs, and local review loops. These are useful for agentic coding work, but they are tracked separately from orchestrator/player entries when they do not manage agent teams or runtime behavior directly.

• Lanes - macOS workspace where Claude Code, Codex, Gemini CLI, and other agentic CLIs run as parallel real-PTY sessions with boards, auto-created git worktrees, session resume, diffs, and file editing.

Governance And Enforcement

Policy, permission, approval, budget, trust, reputation, settlement, and audit layers that sit between agent intent and execution. These are not orchestrators themselves; they govern actions routed through an orchestrator, framework, coding agent, MCP server, marketplace, exchange, or application runtime.

Open and open-core governance tools

• Microsoft Agent Governance Toolkit - MIT-licensed runtime governance toolkit for AI agents with deterministic policy enforcement, zero-trust identity, execution sandboxing, SRE controls, and compliance checks.
• Code Atelier Governance SDK - Python SDK for pre-execution governance gates around AI agents, backed by Postgres.
• Veto (GitHub) - Apache-2.0 authorization layer for AI agent tool calls, with TypeScript and Python SDKs, YAML policies, approval routing, and audit logs. Veto Cloud is commercial.
• SettleBridge (GitHub org) - Trust and policy gateway for agent-to-agent settlement, including reputation thresholds, spending limits, provenance requirements, escrow, dispute resolution, marketplace bounties, and cryptographic audit trails. The related A2A Settlement repo is MIT-licensed; SettleBridge's public license metadata is not fully consistent yet.
• Okto Pulse (GitHub) - Elastic-2.0 local-first SDLC workbench with 17 enforced governance gates for AI coding agents, including independent validation, blocking evidence requirements, and spec-coverage checks before work reaches done.

Commercial and platform governance layers

• Databricks Unity AI Gateway - Commercial Databricks platform layer for governing LLM access, MCP servers, APIs, observability, costs, guardrails, fallbacks, and rate limits in agentic AI workflows.

Agent-Friendly Tooling

Tools that are not orchestrators themselves, but make multi-agent systems easier to operate, measure, observe, or reuse.

• Agent Analytics - Web analytics for builders that Claude Code, Codex, Cursor, OpenClaw, Paperclip, and similar AI agents can use.
• ClawTrace - Observability for OpenClaw agents that shows what failed, where spend leaked, and how to improve runs.
• agenttrace - Local TUI observability for AI coding-agent sessions, tokens, cost, tool failures, latency, anomalies, diffs, and CI evidence.
• Companies.sh - Reusable companies for AI agents: pre-built organizations that can be installed with a single command.
• Orchestrator (GitHub) - MIT-licensed macOS workspace connecting Codex execution with Kanban tasks, subagent inspection, repository browsing, and local code review; public beta for macOS 15+ using the user's own supported Codex account.

Implementation Services

Services companies and consultants that publish practical material on deploying, hardening, and operating multi-agent systems. These are not orchestrator/player entries.

• Nomad Inno (X) - Malta-based AI innovation group helping organizations understand, implement, and scale AI through consulting, custom development, workflow automation, and AI products.
• Agentix Labs - Implementation services and practical writing for teams moving AI agents from pilots into production operations. Public contact details list United States and Canadian offices in New York and Montreal; no explicit worldwide coverage claim is made.

Contributing

Pull requests are welcome if the project clearly fits the directory scope.

• Add or update entries in src/data/orchestrators.ts.
• Prefer official website, docs, and GitHub links.
• Keep summaries factual, concise, and based on public sources.
• For governance and enforcement entries, state whether the source is open source, open-core, commercial, or a platform feature.
• If you add a project, include the source you used to verify it.

Local Development

This repo contains the Awesome Multi-Agent Orchestrators list and a small Astro site for openorchestrators.org.

cd open-orchestrators.org
npm install
npm run dev

Most editorial updates should happen in src/data/orchestrators.ts, which is the source of truth for directory entries, rank order, notes, tags, and links.

News updates should happen in src/content/news. Directory entries and news posts are intentionally separate: directory metadata describes the player; news Markdown describes a dated update about that player.