Orchestrators & Multi-Agent tools for MCP

Self-hosted platform where agents wake on heartbeats to claim tickets, governed by org charts, budgets, and approval gates.

Intelligent routing across multiple LLM providers with load balancing, intelligent fallbacks, rate limiting, and response caching. Achieves 40–60% token cost reduction through smart model routing (cheap models for simple tasks, capable models for complex reasoning). Essential infrastructure for harnesses operating under strict cost budgets where model selection is a per-turn decision

Agent orchestration platform -- deploy multi-agent swarms, coordinate autonomous workflows, distributed swarm intelligence, RAG integration

Cross-harness plugin marketplace that maintains one source-of-truth plugins/ directory and generates harness-native artifacts for Claude Code, Codex CLI, Cursor, OpenCode, Gemini CLI, and GitHub Copilot. It is the clearest practical example of treating reusable agent capabilities as a portable distribution format rather than ad-hoc prompt files

Task management for AI-driven development: PRD parsing, task expansion, multi-provider support and selective tool loading to save context

The leading TypeScript toolkit for building AI agents (20M+ monthly downloads, 25+ provider integrations). AI SDK 6 introduced a first-class Agent abstraction with ToolLoopAgent for production-ready tool execution loops, DevTools for local debugging, full MCP support, and type-safe UI streaming. The unified API across OpenAI, Anthropic, Google, and AWS Bedrock makes it the default choice for TypeS

OSS Zapier alternative with AI

Harness builder for deterministic AI coding workflows, combining agent steps with scripts, validation gates, approvals, and isolated git worktrees. Claude Code, Codex, and more.

The Open Runtime for AI Agents

Astro's TypeScript-native agent harness that treats an agent as a function composed of model, sandbox, skills, tools, and MCP servers. The deploy-anywhere runtime (Node, Cloudflare Workers, GitHub Actions, Daytona) and first-class durability make it a practical reference for building autonomous agents rather than chatbots

Local-first desktop workspace for AI coding agents with persistent projects and sessions, model switching, Plan/Goal modes, permission controls, plugins, MCP, and multi-agent orchestration

可私有部署的多租户知识智能体平台：统一 RAG、知识图谱、多智能体、MCP/Skills、沙盒与权限管理。Yuxi = Cloud Agents + Knowledge RAG, Self-hosted knowledge agent platform for RAG, knowledge graphs and multi-agent workflows.

Token-efficient microkernel agent with an on-device SquillaRouter that dispatches each turn to the cheapest capable model, a unified turn loop across CLI/Web/chat, and persistent memory with on-device embeddings. Demonstrates that harness-level routing and loop optimization — not just model scale — are the binding levers for intelligence density

Self-hosted orchestration dashboard for agent task dispatch, multi-agent workflow coordination, and spend monitoring across gateways. The zero-external-dependency design (SQLite, single pnpm start) makes it the most practical open-source control plane for teams that need governance and cost visibility without building infrastructure from scratch

Next-generation personal AI assistant powered by LLM, RAG and agent loops. Supports computer-use, browser-use, coding agent, 30+ model providers, visual workflow builder, and MCP tools. Self-hostable with admin dashboard

Cross-platform local MCP capabilities for ChatGPT with Chrome integration, Goal, Compact & Resume, and durable multi-agent workflows.

Platform for building and running AI agents across the full lifecycle — build, test, deploy, monitor, improve. Hosts your agents, runs their tools, and coordinates them over an event mesh, with sessions, artifacts, auth, evaluators, and observability built in.

Open-source platform for building AI agent networks with multi-protocol support (WebSocket, gRPC, HTTP, MCP, A2A) and multi-agent orchestration

by Patrick Ellis - Three review workflows developed at an AI-native startup: automated PR code review (dual-loop, GitHub Actions plus slash command), security review against OWASP Top 10 with severity-classified findings, and the widely-copied design-review agent that drives Playwright MCP to check UI/UX and accessibility against a live preview. Each ships as a slash command, a subagent, and a wor

Connect any AI copilot or autonomous agent to your enterprise tools — through a single, secure MCP/Agent gateway with built-in identity, access control, and full observability.

(3.1k ⭐) - A comprehensive collection of specialized AI subagents for Claude Code, designed to enhance development workflows with domain-specific expertise

A practical, open-source guide to mastering WorkBuddy through real-world workflows.开源的 WorkBuddy 实战蓝皮书：教程、真实工作流、Skills、MCP、自动化与多智能体实践。

Open-source AI teammates in Slack, Teams, Discord, Telegram and Mattermost. Self-host it, with any agent harness and any model.

Leverage Notte Web AI agents & cloud browser sessions for scalable browser automation & scraping workflows

Deterministic scheduler for 40+ CLI coding agents running in parallel git worktrees with an HMAC-signed audit chain, signed agent cards, and per-artefact lineage. The zero-LLM coordination loop and tamper-evident audit trail make it the only open-source orchestrator designed for compliance-sensitive agent fleets

Native-Rust agent harness platform with four surfaces (GUI, CLI, web, one-shot), multi-provider routing, and sovereign-by-design architecture. The most complete open-source reference for a locally-run, offline-capable harness that unifies MCP, skills, AGENTS.md, hooks, and session resumption in a single binary

Open source autonomous AI penetration testing platform where Markdown methodology agents orchestrate 80+ offensive security tools through MCP controlled execution with agentic reasoning, keeping an evidence trail per finding. Model agnostic, tuned for Claude Opus

Coordinate Claude, Codex and Cursor windows with shared tasks, acknowledgements, prechecks, parallel tests, source-bound proofs and independent review. Built-in skills recall lessons, simplify code and forecast work. Track progress, decisions and evidence in a bilingual local dashboard, with MCP plugins. Browse skills and verified security audits.

(974 ⭐) - Delegate tasks to Codex GPT 5.2 directly from within Claude Code

An open-source, PyTorch-like runtime for dynamic multi-agent and multi-session workflows.

(607 ⭐) - An agent orchestration harness for Claude Code. It coordinates multiple AI agents in parallel, persists memory across sessions, and routes your intent to the cheapest execution path automatically

Self-hosted, open-source AI browser agent that browses, clicks, types and reads real web pages from plain-English instructions; runs locally on a patched Firefox engine and ships as an MCP server and a local web UI

Local-first shared work-state MCP server (npx -y @concord-ai/concord-mcp) that lets coding agents coordinate on one repo through presence, file claims, messaging, handoffs and review packets. Works with Gemini CLI alongside Claude Code, Codex, Cursor and other agents. TypeScript, stdio, MIT

Attributed, resolvable artifact references for agent handoffs: a ~30-byte token replaces pasted context, letting each consumer resolve only the projection or slice it needs under byte budgets while propagating corrections to every holder. The first concrete primitive that turns multi-agent handoffs from a context-copy problem into a reference-and-projection problem

HunterCode · Community Edition · 腾讯 WorkBuddy 金融版的开源本地替代方案 · open-source, local alternative to Tencent WorkBuddy Finance Edition · 多智能体投研终端 A股/港股/美股 · 推理在你机器上跑 · 仓位不出你硬盘 · BYOK · Apache 2.0 · docker compose 自托管

Workflow orchestration with Codex and other CLI backends; successor to the archived Claude Code Workflow

白箱AGI架构探索：元认知（自我认知循环）、持续学习（知识飞轮）、世界模型（条件空间+语义时空图）、自我改进（自举纪律）、零LLM白箱管线与可审计信任护栏。

Coding agent built on Pi, with a CLI and a desktop app, where a small, fast judge model (Jev, a local judge or any LLM) makes the routine calls at more than 30 decision points, such as which chunks of a long tool output enter the context, whether a rule-flagged command was asked for, and whether a fetched page or MCP result carries instructions aimed at the model; every verdict is logged locally

TypeScript AI SDK and agent framework with zero dependencies. Build production AI agents and multi-agent swarms: durable execution, long-term memory, hybrid RAG, MCP client/server, A2A, human-in-the-loop approvals, CodeAct sandboxes, realtime voice. One streaming API, 29 providers: Claude, GPT, Gemini, Grok, DeepSeek, Ollama. Node, Bun, Deno, edge.

Save 94% on AI coding tokens. Index your codebase, agents search instead of reading files. Works with Claude Code, Codex, Copilot, Cursor, Gemini CLI. Local MCP server, free, open source.

Local MIT CLI that keeps Claude Code and Codex as live peers in one session for mid-turn review, task split, and quota-boundary handoff.

Open-source control layer for AI coding agents. Adds hard budget caps, verifier gates, explicit stop reasons, and run receipts around Claude Code, Codex, and Ralph-style autonomous coding workflows

Open-source, self-hostable alternative to ChatGPT and NotebookLM — agents, skills, MCP, and cited knowledge in one AI workstation. Your models, your infrastructure. Free to use GPT 6: https://SourceWeft.com/

Run a hive mind of agents to focus on a specific goal

A programmable tool and agent runtime for Pi

Desktop voice and chat assistant for Windows, macOS, and Linux that runs CLI coding agents such as Claude Code, Codex, and Gemini CLI side by side in Git worktrees, with memory, scheduled routines, and any single model key or local models via Ollama.

Serverless control plane on Lambda with a dedicated EC2 worker per session, triggered by issue comments, assignments, and PR reviews.

Open-source desktop coding agent with role-based sub-agents, a model per role, and a built-in editor, terminal, and Git workspace; publishes its Terminal-Bench 2.1 runs against Codex CLI

The operational superset of Pi Coding Agent — everything Pi, plus observability, governance, recovery, evaluation and multi-agent orchestration. Pi Coding Agent 的运维超集——完整继承 Pi 的核心能力，并进一步扩展可观测性、治理、故障恢复、评估和多 Agent 编排。

Self-hosted "company OS" where agents work in departments, delegate to each other, and keep working unwatched on your Anthropic/OpenAI subscriptions — company map, per-agent dashboards, artifacts, meetings, and terminals. Source-available under FSL-1.1.

Official source for the sandbase-mcp skill - discover, inspect, and invoke 2,000+ AI models and APIs through a local MCP bridge with explicit schema and cost checks (Apache-2.0)

Coordinates Claude Code, Codex, and Gemini CLI on one shared repo through a research-plan-delegate-adversarial-verify-deploy pipeline, with file locking, knowledge capture, and drift detection. Single Rust binary.

(127 ⭐) - A curated collection of 100+ sub-agent prompts and MCP servers for Claude Code

Structured multi-agent debate in the terminal

NixOS-based AI operating system with multi-agent routing (Opus agent for full system access, Sonnet agent for mobile/concise). Modular runtime swaps Claude Code and OpenClaw drivers per-agent via SIGHUP — no SSH or rebuild. 91 typed MCP tools across 9 Rust daemons, hash-chained audit ledger, atomic NixOS rollback, P2P encrypted mesh (Noise_XX + ML-KEM-768 hybrid PQ), encrypted credential store (AE

UIUC's open-source specification and execution language for LLM-agent workflows: declarative YAML with typed steps, branching, loops, and explicit state management, backed by a Docker sandbox with 50+ MCP tools, checkpointing, and trajectory logging. A concrete reference for turning ad-hoc agent loops into version-controlled, reproducible harness artifacts

Local-first AI workforce platform — named agents with roles, personas, tools, memory, and per-agent MCP servers that self-organize through task delegation. Any agent can assign work to another; the scheduler wakes coworkers when dependencies clear

(67 ⭐) - A multi-agent orchestration system that enables Claude Code, Codex CLI, Amazon Q Developer, and Gemini Code Assist to collaborate bidirectionally through intelligent task routing and consensus-based decision making

Cross-platform skill format. Write once, use in Codex and Claude Code

Niche agents: accessibility auditor, i18n extractor, performance profiler. Fills gaps VoltAgent doesn't cover

Run Codex and Claude Code in tandem on the same codebase

Headless coordination over MCP: claim work, detect file conflicts, heartbeat, and hand off across sessions.

MIT-licensed platform for always-on AI agents. As a software factory for Claude Code, a crew of sessions (an orchestrator, builders and a reviewer) takes planned projects from tickets to merged pull requests; as an AI SRE, triggers over your systems' events wake an agent with the correlated timeline. Ships an MCP server, a Claude Code plugin and agent skills

Runs project-board tasks in persistent sandboxes, with manager-agent delegation and deliverable review. Supports Claude Code, Codex, OpenCode, and other configured runtimes.

by Kevin Hartman - Spec-first, test-driven agent framework: runs the build loop against live Lakebase Postgres branches under a deterministic orchestrator, with human-approval gates

Self-hosted platform where persistent bots with tools, MCP servers, and long-term memory hand work to each other in shared threads and can stop to ask a human before a sensitive action

Self-hosted server that runs a team of personal AI agents, each with its own persona and private memory, a shared long-term wiki the agents consolidate in background "dream" cycles, and conversations that switch between Claude, ChatGPT/Codex and OpenAI-compatible local models

All-in-one agent harness for Codex CLI - Boss meta-orchestrator, 400+ agents, 200+ skills, 3 MCP servers

Intelligent orchestration platform routing tasks to Claude, Gemini, Codex, and OpenCode via Budget→TOPSIS→LinUCB composite router. 30 MCP tools (orchestrate, consensus voting, research pipelines), 12 expert agents, 17 skills, plugin-native /agents + hooks. Install: /plugin marketplace add williamzujkowski/nexus-agents

by MuhammadTalhaMT - Relays messages between two Claude Code sessions on separate machines using the Channels API, with delivery tracking and per-message reply budget. Pairs sessions belonging to different accounts, which native cross-session messaging does not. Although this sort of functionality is now a platform feature, it's always cool to see people who have built their own solutions

Rust engine that escrows a verified pre-commit inverse before an agent's tool effect (filesystem mutation or MCP call) lands, or refuses the effect if no inverse can be verified; issues signed, offline-verifiable DSSE receipts and Merkle tile logs. Apache-2.0, alpha (v0.1.1)

Self-hosted web workspace that runs Gemini CLI alongside OpenCode, Codex, Claude Code and other coding CLIs as side-by-side panes in persistent rooms, with MCP servers and managed browser panes. Install: npx --yes run-spaceapp@latest install. Source-available (BSL 1.1)

Gemini CLI extension, MCP server and CLI over the session transcripts Gemini CLI, Claude Code, Codex, Copilot CLI, Cursor, Grok and OpenCode already write. List and read other live sessions, hand off compact state, send into a specific chat, or spawn a headless session that replies through a shared inbox. Install: gemini extensions install https://github.com/abnegate/magents. Rust, MIT

Desktop and CLI workspace with teammate conversations, bounded subagents, browser and file tools, MCP integrations, and persistent project memory

Architecture pattern for autonomous agents using Claude Cowork, Notion, and MCP. Persistent multi-domain agents with no custom infrastructure or code

Secure local multi-agent runtime that plans tasks, delegates execution to isolated workers, and exposes tools, MCP, scheduling, and a private dashboard for autonomous operations

Self-hosted coordination for agent fleets: shared semantic memory, tasks, agent-to-agent messaging, session handoffs and cross-agent knowledge synthesis

Skill router and context picker for coding agents: a local model (via Ollama, offline by default) routes each prompt to the right skill via hybrid BM25 + embedding retrieval, cross-encoder rerank, and an LLM judge, then assembles a minimal context bundle for the execution model. Built-in prompt-injection gate scans every retrieved document, fail-closed. Ships an MCP stdio server and an OpenAI-comp

Sync MCP server configs from Claude Code to Gemini CLI, Codex CLI, OpenCode, Kiro, Cursor. One command, all agents

Autonomous agent for short-video production: parses natural-language requests, plans the edit, then executes it — AI script, neural voiceover, CLIP-verified stock footage, karaoke captions, and publishing to YouTube, TikTok, Instagram and X

Runs Gemini CLI alongside your other coding agents (Claude Code, Codex, Qwen, Vibe) in parallel and read-only over your diff, then synthesizes their reviews into one ship / no-ship verdict. A cross-model review panel

Gemini CLI extension + MCP server that lets Gemini CLI, Claude Code, Codex and Cursor sessions on one machine search each other's history and message each other live (send/wait/reply, channels, file claims). Local SQLite, no daemon

A curated catalogue of awesome agentic AI patterns