Best tools for building multi-agent systems (2026)
Frameworks for teams of agents that hand work to each other, split tasks and check each other's output. We list 78 tools for building multi-agent systems below, each with a link-checked page, pricing as the maker states it and alternatives.
Updated
78 tools: 65 open source, 3 source-available, 5 free or freemium, 3 paid, usage-based or enterprise, 2 pricing on the maker's site.
Agent frameworks and SDKs (26)
BeeAI Framework
Framework for building production-ready agents and multi-agent systems in Python or TypeScript.
Added
LangGraph4j
Java library for building stateful multi-agent applications as graphs, working with LangChain4j and Spring AI.
Added
Jido
Elixir framework for building supervised, fault-tolerant agents and multi-agent systems on the BEAM.
Added
KaibanJS
JavaScript-native framework for building and managing multi-agent systems with a Kanban-style board for agent tasks.
Added
Pydantic Deep Agents
Self-hosted terminal AI assistant and Python framework on Pydantic AI, with tool calling, sandboxed execution, multi-agent teams and checkpoints.
Added
txtai
All-in-one Python framework for semantic search, LLM orchestration and agents built on an embeddings database of vector, graph and relational indexes.
Added
UFO
Microsoft's desktop agent framework that operates Windows apps and orchestrates multi-device tasks as task graphs.
Added
Agency Swarm
Multi-agent orchestration framework built on the OpenAI Agents SDK for creating collaborating agent "agencies" with defined roles.
Added
agentUniverse
Multi-agent framework with reusable collaboration patterns for building LLM agent applications.
Added
Neuron AI
PHP framework for creating and orchestrating AI agents with LLMs, tools, vector stores and memory.
Added
AutoGen
Microsoft's framework for multi-agent conversations, now archived and no longer maintained; Microsoft points new users to Microsoft Agent Framework.
Added
CrewAI
Framework for role-based multi-agent crews.
Added
Agno
Python framework and runtime for building, running and managing multi-agent systems with memory, knowledge and tools.
Added
AgentScope
Framework for building and running transparent multi-agent applications with visual debugging.
Added
LangChain Deep Agents
Library for long-running agents with subagents and skills; 0.7.23 (Oct 7) adds pinned_skills.
Added
CAMEL
CAMEL, one of the first multi-agent frameworks, for building role-playing agents and large agent societies.
Added
Swarms
Enterprise multi-agent orchestration framework.
Added
AG2
AG2, the community continuation of AutoGen, an open-source framework for multi-agent systems.
Added
Langroid
Python framework for multi-agent programming with LLMs.
Added
LazyLLM
Low-code framework for building multi-agent LLM applications.
Added
MetaGPT
Multi-agent framework that assigns roles like product manager and engineer to LLM agents to build software.
Added
Haystack
Open-source orchestration framework for building production LLM apps, RAG and agents.
Added
Aden Hive
Multi-agent harness for production AI.
Added
Omnigent
Open-source agent framework and meta-harness that orchestrates Claude Code, Codex, Cursor and other agents together.
Added
PraisonAI
Multi-agent framework for building a self-improving AI workforce with little code.
Added
BotSharp
Multi-agent framework for .NET.
Added
Running many agents in parallel (11)
LayerFS
LayerFS is a filesystem built for running many coding agents in parallel. Each agent mounts a complete, writable workspace on top of shared committed history, does its work, commits what it changed and goes away, so compute only exists while a command runs. Every tool call becomes a commit on a branch, same-file clashes between agents are refused rather than silently merged, each line records the step that wrote it, and any commit can be forked at no cost for recursive exploration or RL training. Agents on different computers share one serverless store with chunk-level deduplication.
Added
ZCode
ZCode is Z.ai's open-source AI coding workspace and official harness for GLM-5.3, with a desktop app, web interface and terminal agent that run multiple agents on long goals, plus control from WeChat, Feishu or Telegram.
Added
Vicoa
Vicoa is an agent-native IDE for running Claude Code, Codex, Gemini and 40+ other coding agents in parallel, each on its own git worktree, from desktop, web or phone, with notifications when an agent needs you.
Added
golutra
Multi-agent workspace that turns Codex, Claude Code and OpenClaw CLIs into one orchestrated team with parallel execution and task tracking.
Added
Ruflo
Open-source multi-agent meta-harness for Claude Code, Codex, and MCP clients; Oct 13 release adds Mods (mission workflows) and a single control plane with human approval.
Added
OpenSwarm
AI-first 'operating system' where swarms of agents run apps and browsers in parallel.
Added
Oh My Claudecode
Multi-agent orchestration layer for Claude Code that coordinates teams of specialized agents.
Added
Paseo
Orchestrate multiple coding agents from desktop and mobile.
Added
Superset
Agentic IDE for orchestrating many coding agents in parallel.
Added
Agent Orchestrator
Runs and supervises teams of coding agents from planning to merge.
Added
Maestro
Command center for orchestrating multiple coding agents.
Added
Workflow automation (9)
Sage
Sage is an MIT-licensed framework for building configurable, self-improving AI agents from reusable agent packages, tools, Skills and MCP services. It supports multi-agent workflows and multi-round self-improvement, keeping a version history of agent packages so you choose which version to adopt.
Added
Dynamiq
Python orchestration framework for building agentic and RAG applications from reusable nodes and workflows.
Added
Jasper
Marketing AI platform that orchestrates agents to run end-to-end marketing workflows such as content, campaigns and brand-consistent copy.
Added
UiPath
Enterprise automation platform that orchestrates AI agents, software robots and people across business processes with security and compliance controls.
Added
OpenAI Agents SDK (Python)
OpenAI's Python SDK for building multi-agent workflows; v0.23.0/0.23.1 released Oct 1-2.
Added
Eigent
Open-source cowork desktop app that runs multi-agent workflows locally.
Added
Microsoft Agent Framework
Microsoft's framework for building and orchestrating agents and multi-agent workflows in Python and .NET.
Added
Orkes
Orchestration platform built on Conductor for durable workflows and AI agents.
Added
Dagu
Self-hostable workflow orchestrator with a simple definition format and web UI, usable for agent jobs.
Added
Research and reports (7)
Grok Bot Jev Router
Grok Bot Jev Router is an open-source Python reference router that puts TypeSafe Jev in front of Grok Bot as a cheap decision layer. Before expensive research, browser, retry or subagent work, Jev classifies the request and returns an explicit action such as reuse a cached result, stop a failing retry, cap research, run a deterministic step, or ask a human. It starts in shadow mode that only logs decisions, has an active mode for a Grok Bot skill that honors them, and includes a kill switch. It never sends, pays or deletes anything itself.
Added
open-kritt
A self-hosted AI security research tool that orchestrates agents to hunt for vulnerabilities in a codebase and then validate and rank what they find; managed scans are also offered.
Added
AutoDev
Multi-agent development platform built on Kotlin Multiplatform for research, coding, review and data queries across IntelliJ, VS Code, CLI and more.
Added
Paper2Agent
Multi-agent system that turns a research paper and its code into a working MCP server and agent skills you can query and run.
Added
Google Deep Review
Google Research's multi-agent system that checks research papers, design docs, and legal or financial documents for logical gaps and unsupported claims.
Added
DeepCode
Open agentic coding system that turns papers and specs into code with multi-agent orchestration.
Added
Crewdle
Hire autonomous agents and teams for support, sales, bookkeeping and research, paid per task.
Added
More tools (25)
Cumora
Cumora is a team chat app where AI agents are members of the workspace rather than a chat box you prompt. Each agent gets a persona, its own private memory and files, a status, and a schedule on which it wakes up, reads the room and decides whether to message someone, post an idea or pull a small group together. Agents can message each other, and you can read those side conversations. People join by email or link, agents can be grouped into teams and given projects, and the same live workspace runs on desktop, in the browser and on a phone. It is free during an invite-only preview.
Added
Zero
An open-source terminal coding agent written in Go that works with 30+ model providers, supports MCP, skills and subagents, and keeps all sessions on your own disk.
Added
AgentConnect
An open-source, multi-agent workspace where you @-mention any agent wherever work happens, and agents work alongside your team and each other.
Added
Reason Network
A small peer-to-peer daemon that links coding-agent sessions across your team's computers, so Claude Code, Codex, Cursor, Gemini CLI and other agents can see who is working on what and message each other directly, with no server.
Added
Agent Tincan
Agent Tincan lets your AI agents (Claude Code, Codex, Gemini, Hermes, ChatGPT and others) ask each other to do things and hand back results through a small relay on your own Tailscale network, with no open ports or API keys between agents.
Added
Agentlas
A local-first agent OS that keeps specialist agents in a hub and spins up a temporary orchestrator for each task, with reusable methods and your choice of model and host.
Added
luvus
Mission control for AI coding agents: run, watch, resume and orchestrate Claude Code, Codex, Copilot, opencode and others from one terminal with a live view of each agent.
Added
Pragor
Coordination and governance layer for teams of people and AI agents, not an autonomous agent: a shared board with tasks, directed messages, approvals, files and an attributable activity record, where agents connect over REST or project MCP access.
Added
ChatDev
Zero-code multi-agent platform from OpenBMB, grown from a software-development agent company into a general multi-agent orchestration system.
Added
Orkas
Local-first multi-agent desktop app where a commander model coordinates specialist agents and your installed coding CLIs.
Added
Agent Resource Market
Eellos LLC's API-first beta where AI agents discover and publish resource offers, join structured discussions and post contextual reviews. Exchanges are simulated with demo credits (payments are disabled) and registration is capacity-limited.
Added
DevOpsGPT
Multi-agent system that combines LLMs with DevOps tools to turn natural-language requirements into working software on existing code.
Added
Kiro Crew
Kiro's open-source, persistent development workspace that runs unattended and recurring coding tasks locally or on your own hardware.
Added
Codeg
Multi-agent coding workspace that brings sessions from Claude Code, Codex, OpenCode and other agents into one app so they can work together.
Added
CLI Agent Orchestrator
AWS Labs tool that coordinates several coding CLIs such as Claude Code, Kiro and Codex as a team, each in its own tmux session.
Added
Tensorlake
Serverless runtime that gives agents MicroVM sandboxes and durable orchestration for long-running background jobs.
Added
Bubbaloop
Kornia's open-source single-binary agent runtime with an MCP server that orchestrates cameras, sensors and robots on Jetson, Raspberry Pi or Linux through chat.
Added
PentAGI
Autonomous multi-agent system for running penetration testing tasks.
Added
Oh My OpenCode Slim
Lean multi-agent suite for OpenCode that mixes models and delegates tasks automatically.
Added
SuperPlane
Open-source platform for orchestrating one-shot engineering work with agents.
Added
Flyte
Type-safe orchestration for agents, ML pipelines and inference on Kubernetes.
Added
Open Multi-Agent
Self-hosted TypeScript agent runtime with durable approvals and verifiable run records.
Added
Archestra
Enterprise AI platform with guardrails, an MCP registry, gateway and orchestrator.
Added
Ralph Orchestrator
Orchestrator for autonomous agent loops using the Ralph Wiggum technique.
Added
Inworld
Platform and runtime for real-time voice AI and interactive characters, including text-to-speech and agent orchestration for consumer apps.
Added
GitHub repos (71)
| Repo | GitHub stars | Licence | Last push |
|---|---|---|---|
| Ruflo ↗ | 74,609 | MIT | 2026-10-21 |
| MetaGPT ↗ | 70,885 | MIT | 2026-01-21 |
| AutoGen ↗ | 61,395 | MIT for code (LICENSE-CODE); CC-BY-4.0 for documentation | 2026-10-16 |
| CrewAI ↗ | 59,675 | MIT | 2026-10-21 |
| Agno ↗ | 42,769 | Apache License 2.0 | 2026-10-21 |
| Oh My Claudecode ↗ | 39,972 | MIT | 2026-10-21 |
| ChatDev ↗ | 34,545 | Apache License 2.0 | 2026-10-16 |
| AgentScope ↗ | 33,376 | Apache License 2.0 | 2026-10-21 |
| Orkes ↗ | 32,319 | Apache License 2.0 | 2026-10-19 |
| PentAGI ↗ | 30,728 | MIT | 2026-10-09 |
| LangChain Deep Agents ↗ | 30,236 | MIT | 2026-10-21 |
| OpenAI Agents SDK (Python) ↗ | 30,023 | MIT | 2026-10-19 |
| Haystack ↗ | 26,748 | Apache License 2.0 | 2026-10-21 |
| Paseo ↗ | 21,143 | Apache-2.0 (third-party components under their own licences) | 2026-10-21 |
| CAMEL ↗ | 17,911 | Apache License 2.0 | 2026-10-19 |
Top 15 of 71 by stars. Stars and licence come from each repo, checked when the listing was last verified.
Related tasks
Popular with: AI engineers · Related searches: Multi-agent frameworks
Tools are matched to tasks by category, use cases and what each tool says it does. Missing one? Submit it free.