# Open Source AI Agent Tools

> Source: https://theagentsindex.com/collections/open-source-ai-agent-tools (curated, quality-gated, re-verified)
> Curated by: The Agents Index Editorial, Research desk

We checked the license behind all 126 tools in our index and found 33 that call themselves open source in some form, but only 26 clear the bar of a clean, unrestricted license with no strings attached. This collection is those 26: the tools you can fork today, self-host forever, and keep running even if the vendor disappears, with nothing held back behind a separate commercial tier.

The other seven "open" tools don't make this list, and that's deliberate: n8n and Dify both hedge their openness with a license condition their own docs tell you to check (fair-code, added-conditions Apache), Mastra and OpenHands are open at the core but gate their enterprise controls behind a separate paid license, Sim is open-source at the code level (29.4k GitHub stars) but sells self-hosting itself only on its Enterprise tier, and Warp and Claude Agent SDK each open-source only one half of the product: Warp's terminal client (AGPL-3.0/MIT) while its Oz cloud-agent platform stays closed, and the Agent SDK's two wrapper packages (MIT) while the Claude Code CLI binary they spawn as a subprocess is never published as source. Junie isn't part of that "hedged open" group either. It's simply closed-source, with no public repository at all. Real, legitimate business models, just not a "no caveats" self-hosting story. See our full license census for the complete breakdown, caveats included.

This list isn't the same set as our [self-hostable collection](/collections/ai-agent-tools-you-can-self-host/), even though the two overlap heavily. 24 of these 26 tools also appear there. Cartesia and Kilo Code are the two exceptions. Cartesia's Apache-2.0 Line SDK is unconditionally open but its voice runtime only ever runs on Cartesia's own infrastructure. Kilo Code was added to this collection on 2026-08-18, the same day it was published, and the self-hostable collection has not yet been re-read to include it. That collection also counts 13 tools that don't clear this one's bar: Mastra, Claude Agent SDK, n8n and Dify carry a license condition on an otherwise-open core, and Augment Code, Factory, CodeRabbit, Bland AI, Deepgram Voice Agent API, Windsurf, Warp, OpenHands and Junie are closed-source products that still sell a real self-hosted or on-prem path. If what you actually need is "can I run this myself," check that collection instead.

| Name | What it is | Why it's here | Pricing | API |
| --- | --- | --- | --- | --- |
| [LangChain](https://theagentsindex.com/langchain.md) | Open source agent frameworks plus LangSmith, the platform for tracing, evaluating and deploying agents. | The most widely adopted agent framework; MIT-licensed core, free to self-host with no vendor lock-in. | — | Yes |
| [CrewAI](https://theagentsindex.com/crewai.md) | Open-source multi-agent framework with a managed platform for deploying and governing agent workflows. | MIT-licensed multi-agent orchestration framework, reached a stable 1.0 general-availability release in October 2025. | — | Yes |
| [AutoGen](https://theagentsindex.com/microsoft-autogen.md) | Microsoft's open-source Python framework for building single- and multi-agent applications; now in maintenance mode. | Microsoft's MIT-licensed multi-agent framework: free to self-host, pay only your own model provider. | — | Yes |
| [Microsoft Agent Framework](https://theagentsindex.com/microsoft-agent-framework.md) | Open-source SDK for AI agents and multi-agent workflows across .NET, Python and Go. | Microsoft's MIT-licensed successor to AutoGen and Semantic Kernel: free to self-host across .NET, Python and Go, or an optional Azure-billed managed layer on top. | — | Yes |
| [LlamaIndex](https://theagentsindex.com/llamaindex.md) | Open-source agent framework plus a credit-metered document parsing, extraction and indexing platform. | MIT-licensed core framework for building agents over your own data; the paid LlamaCloud parsing layer is optional. | — | Yes |
| [LangGraph](https://theagentsindex.com/langgraph.md) | Open-source orchestration framework and runtime for stateful, long-running LLM agents, from LangChain. | MIT-licensed agent runtime from the LangChain team, free to self-host; LangSmith and LangGraph Platform are optional hosted layers on top, not a gate on the library itself. | — | Yes |
| [OpenAI Agents SDK (Python)](https://theagentsindex.com/openai-agents-sdk.md) | Open-source Python framework for building agents with tools, handoffs, guardrails, sandboxes and tracing. | MIT-licensed and free to self-host: OpenAI's own lightweight SDK for building agentic workflows. | — | Yes |
| [Pydantic AI](https://theagentsindex.com/pydantic-ai.md) | Typed Python agent framework: any model, any surface, with a harness for long-running coding and research agents. | MIT-licensed, type-safe Python framework from the team behind Pydantic: free to run anywhere. | — | Yes |
| [Letta](https://theagentsindex.com/letta.md) | Stateful agent harness, cloud and SDK for agents with persistent, git-versioned memory. | Apache-2.0 licensed agent framework with persistent memory; self-host free or opt into managed Letta Cloud later. | — | Yes |
| [Cline](https://theagentsindex.com/cline.md) | Open-source coding agent for VS Code, JetBrains, the terminal and CI, running on whichever model you bring. | Apache-2.0 licensed VS Code, JetBrains and CLI agent: only Enterprise SSO and support sit behind a paid tier. | — | Yes |
| [Aider](https://theagentsindex.com/aider.md) | Open-source terminal AI pair programmer that edits and commits code in your local git repo. | Apache-2.0 licensed terminal coding agent, free forever: bring your own model API key or run a local model. | — | Yes |
| [Codex CLI](https://theagentsindex.com/codex-cli.md) | OpenAI's terminal coding agent for inspecting, editing and running repository code from the shell. | Apache-2.0 licensed CLI is fully open, even though the OpenAI models it calls are not. | — | Yes |
| [OpenCode](https://theagentsindex.com/opencode.md) | Open source AI coding agent for the terminal, desktop, browser and IDE, running any model you connect. | MIT-licensed terminal coding agent with 188,976 GitHub stars: one of the most active open-source projects in the category. | — | Yes |
| [zot](https://theagentsindex.com/zot.md) | Single-binary terminal coding agent harness in Go: five built-in tools, most providers, extensions in any language. | MIT-licensed single Go binary, no signup or Docker required: bring your own API key or run local Ollama models. | — | Yes |
| [Juggler](https://theagentsindex.com/juggler.md) | Visual workbench for AI coding agents: branching conversation trees, inspectable model transactions, local or remote. | AGPLv3 for the main app, Apache-2.0 for its extension SDK: a stronger copyleft commitment than most rivals. | — | No |
| [LiveKit Agents](https://theagentsindex.com/livekit-agents.md) | Open-source Python and Node.js framework for voice AI agents, with LiveKit Cloud hosting, telephony and observability. | Apache-2.0 licensed voice-agent framework: one of only two clean open-source picks in the voice-agents category, alongside the Apache-2.0 Line SDK that Cartesia ships. | — | Yes |
| [GPT Researcher](https://theagentsindex.com/gpt-researcher.md) | Open-source autonomous research agent that writes cited long-form reports from web and local sources. | Apache-2.0 licensed: the only clean open-source pick in the research-agents category. | — | Yes |
| [Zed](https://theagentsindex.com/zed.md) | Rust code editor with an Agent Panel running Zed's own agent, ACP external agents, or terminal agent CLIs. | GPL-3.0-or-later core (Apache-2.0 for select components): the first full IDE in this collection, rather than a CLI tool or editor extension. | — | No |
| [Google Agent Development Kit (ADK)](https://theagentsindex.com/google-adk.md) | Open-source framework for building, evaluating and deploying AI agents in Python, TypeScript, Go, Java or Kotlin. | Apache-2.0 licensed across all four of its language SDKs (Python, Go, Java, TypeScript): Google's own official agent framework, free to self-host anywhere. | — | Yes |
| [Botpress](https://theagentsindex.com/botpress.md) | AI agent platform with a visual studio, a code-first TypeScript ADK and a helpdesk, priced per conversation. | MIT-licensed at the core with no license caveat, so it belongs on this list as squarely as any framework here; the managed cloud is an option alongside self-hosting rather than the only way to run it. | — | Yes |
| [Cartesia](https://theagentsindex.com/cartesia.md) | Real-time speech models and a managed runtime for production phone and web voice agents. | Line, its voice-agent SDK, is Apache-2.0 with no caveat: the second open-source entry in the voice-agents category alongside LiveKit Agents; the Sonic/Ink speech models it calls, and the Line runtime itself, stay proprietary and Cartesia-hosted. | — | Yes |
| [Agno](https://theagentsindex.com/agno.md) | Python SDK, open-source AgentOS runtime and hosted control plane for building and operating agents. | Apache-2.0 licensed Python framework, formerly Phidata, with built-in memory, knowledge and multi-agent Teams; the paid AgentOS runtime is a separate BYOC product, not a gate on the framework itself. | — | Yes |
| [AgentsKit.js](https://theagentsindex.com/agentskit.md) | TypeScript foundation for AI agents: runtime, tools, memory, RAG, model adapters and headless chat UI bindings. | MIT-licensed TypeScript foundation for building agents: the Registry's ready-made agent factories are copied straight into your own codebase rather than pulled from a hosted runtime. A separate hosted product, AgentsKit OS, exists in enterprise-only early access with no self-serve signup, but the open-source core remains the only way most teams will run it. | — | Yes |
| [OpenClaw](https://theagentsindex.com/openclaw.md) | Self-hosted gateway that connects chat apps like WhatsApp, Telegram and Slack to AI agents on your own machine. | MIT-licensed personal AI agent that installs and runs on your own machine: the second agent platform in this collection with a clean, unrestricted license after Botpress, but with no managed cloud tier alongside it at all. | — | Yes |
| [goose](https://theagentsindex.com/goose.md) | Open-source local AI agent — desktop app, CLI and ACP server — running on the LLM provider you bring. | Apache-2.0 licensed and now stewarded by the Linux Foundation's Agentic AI Foundation: free to self-host as a desktop app, CLI or embeddable API, with no managed cloud tier and no enterprise carve-out on the code. | — | Yes |
| [Kilo Code](https://theagentsindex.com/kilo-code.md) | Open source AI coding agent for VS Code, JetBrains, CLI and cloud, with 500+ models at provider rates. | MIT-licensed end to end and explicitly built to be self-hosted instead of relying on Kilo's own hosted Gateway: the extension, CLI and cloud-agent code all carry the same license, with no enterprise carve-out. | — | Yes |

## How this list is built

A tool earns a place here only if its entire self-hostable codebase carries a clean, unrestricted OSI-style license (MIT, Apache-2.0, GPL-3.0-or-later, or AGPLv3) with no source-available carve-out, no fair-code condition, and no proprietary enterprise layer gating the part you'd actually run. Verified against each tool's own GitHub license file or our sourced listing research, not a vendor's marketing claim of "open source."

We distinguish an open core from the complete runnable product: a hosted control plane, closed runtime, source-available restriction, fair-code clause, or paid license for self-hosting excludes the listing even when part of its repository is public. Each refresh rechecks the license identifier, repository scope, self-hosting path and whether the licensed code is sufficient to operate the product independently of the vendor.