COLLECTION / OPEN-SOURCE
Open-source AI tools
Products with public source code you can self-host, audit and modify — no gatekeepers.
148 tools
Immich
Immich community
Status not recorded
A self-hosted photo and video library with mobile backup, albums, face grouping and contextual search; storage, server security and recovery are yours to operate.
Meilisearch
Meilisearch
Status not recorded
A search API with full-text, semantic and hybrid retrieval: self-host the MIT Community Edition or evaluate separately licensed Enterprise features and managed Cloud.
vLLM
vLLM Project
Status not recorded
An Apache-2.0 inference and serving engine with batching, hardware-specific backends and documented API interfaces; model and deployment compatibility need separate checks.
AUTOMATIC1111
AUTOMATIC1111
Status not recorded
A self-hosted Stable Diffusion interface with text/image generation, inpainting and configurable workflows.
Cline
Cline Bot
Status not recorded
An open-source coding agent with editor, terminal and desktop interfaces, configurable tool approvals and a choice of model providers.
ComfyUI
Comfy Org
Status not recorded
Node-based media workflows with JSON graph exchange and custom extensions; local open-source software, paid cloud execution and external model calls have separate requirements and costs.
Faster-Whisper
SYSTRAN
Status not recorded
Faster-Whisper reimplements Whisper inference with CTranslate2, offering CPU/GPU quantization, batch transcription and optional word timestamps.
Fooocus
lllyasviel
Status not recorded
An SDXL-based local image interface with image prompts, variation and inpainting, now in limited bug-fix support.
llama.cpp
ggml (community)
Status not recorded
A C/C++ inference engine for supported GGUF models, with CLI, server and hardware-specific backends for local or separately managed deployment.
News Feed Eradicator
Jordan West
Status not recorded
A browser extension that replaces supported social feeds with a quote while retaining other site functions; it is a configurable browsing aid, with version-specific license terms.
Ollama
Ollama
Status not recorded
A local model runner and API with optional paid cloud models; the MIT runner, downloaded models and hosted-service terms have separate costs and permissions.
PydanticAI
Pydantic Services Inc.
Status not recorded
A Python AI SDK for agents with typed tools, dependency injection and validated structured outputs. The MIT library is separate from model-provider costs and optional Logfire or Gateway services.
Qdrant
Qdrant
Status not recorded
A vector database with dense/sparse hybrid queries and payload filtering, available as an Apache-2.0 engine and separately contracted managed, hybrid or private cloud.
Typesense
Typesense
Status not recorded
A search server with typo tolerance, facets, geo and vector features; GPLv3 server software and Typesense Cloud have separate licensing and operating costs.
Upscayl
Upscayl community
Status not recorded
Image upscaling with a free desktop application and a separate paid cloud service.
Whisper
OpenAI
Status not recorded
Whisper is OpenAI's downloadable speech-recognition code and models for multilingual transcription and translation into English, with local deployment options.
Activepieces
Activepieces
Status not recorded
Workflow automation with integrations, branches, loops and AI steps; MIT core and separately licensed enterprise features, with cloud and self-hosted options.
Agno
Agno Inc.
Status not recorded
An agent SDK and FastAPI runtime with a separate control plane for operating agents, teams and workflows.
Appsmith
Appsmith
Status not recorded
Internal app builder with widgets, database/API queries and JavaScript; Apache-2.0 Community Edition and separate commercial editions.
Chroma
Chroma
Status not recorded
A retrieval database for documents, embeddings and metadata, with local, self-hosted and separately billed Cloud deployments.
DeepSeek R1
DeepSeek
Status not recorded
DeepSeek's original downloadable R1 reasoning model and related distilled checkpoints, with model-specific licenses and self-managed computing costs.
Goose
Agentic AI Foundation
Status not recorded
Apache-2.0 local AI agent with desktop, CLI and API interfaces, using configured models and MCP extensions.
Gradio
Hugging Face (Gradio)
Status not recorded
An Apache-licensed Python UI library for functions and models, with optional public share links and separate hosting choices.
Invoke
Invoke
Active
A self-hosted image application with a layered canvas, node workflows and model management; the former vendor cloud has closed.
Jan
Menlo Research
Status not recorded
An Apache-licensed desktop AI app for local models or configured cloud providers, with assistants, MCP connections and a local API.
Langflow
Langflow / DataStax
Status not recorded
A Python-based visual builder for AI application flows, with editable components, APIs and MCP integration.
Langfuse
Langfuse / ClickHouse
Status not recorded
An AI engineering platform for traces, prompts, datasets and evaluations, with an MIT core, commercial extensions and managed Cloud.
LeechBlock
James Anderson
Status not recorded
A free browser extension for scheduled website blocking, time allowances and lockdowns. Its documented controls use rules rather than verified AI features.
LibreChat
LibreChat
Status not recorded
An MIT-licensed self-hosted web chat platform for configured local or remote models, with agents, files and multi-user controls.
LiteLLM
BerriAI
Status not recorded
A model-routing SDK and self-hosted AI gateway with virtual keys, budgets, fallbacks and request-cost records; Enterprise adds licensed features.
LlamaIndex
LlamaIndex Inc.
Status not recorded
A published framework for connecting, indexing and querying data in LLM applications, separate from LlamaParse's hosted document services.
Mem0
Mem0 Inc.
Status not recorded
A memory layer for AI applications, offered as a managed API and an Apache-licensed library; hosting, features, model costs and data terms differ.
MetaGPT
FoundationAgents
Status not recorded
A multi-agent framework that organizes model-backed software roles and structured handoffs through configurable procedures.
Milvus
Zilliz
Status not recorded
An Apache-licensed vector database with Lite, Standalone and Distributed deployments, supporting filtered, hybrid and full-text retrieval.
Copilot for Obsidian
Brevilabs LLC
Status not recorded
An Obsidian plugin for vault-based agents, projects and note assistance, with configurable external, local or paid hosted AI routes.
pgvector
pgvector community
Status not recorded
A PostgreSQL extension for exact and approximate vector search alongside relational data, with HNSW and IVFFlat indexes.
PhotoPrism
PhotoPrism community
Status not recorded
A self-hosted photo library with AI-assisted organization and optional external models; Community code, paid features and hosting have separate conditions.
Promptfoo
Promptfoo
Status not recorded
Open-source CLI and library for configured LLM evaluations and red-team tests, with CI integration and separate commercial platform plans.
PyTorch
PyTorch Foundation
Status not recorded
A Python tensor and neural-network framework with automatic differentiation, CPU/GPU execution and distributed-training components.
Qwen 3
Alibaba
Status not recorded
Alibaba's original 2025 Qwen3 open-weight family with thinking/non-thinking modes; distinct from Qwen Chat and later Qwen generations.
SelfControl
Charlie Stigler
Status not recorded
A free macOS website and mail-server blocker using timed lists. It is a conventional productivity utility rather than an AI assistant.
smolagents
Hugging Face
Status not recorded
Python agent library offering code actions, JSON tool calling and configurable model, tool and execution backends.
Stagehand
Browserbase
Status not recorded
A browser-agent SDK mixing natural-language act/extract/observe operations with code, using local browsers or separately billed Browserbase hosting.
Streamlit
Snowflake (Streamlit)
Status not recorded
A Python framework for interactive data, dashboard and chat apps, with separate Community Cloud and Snowflake deployment options.
Warp
Warp
Status not recorded
Warp combines a terminal, coding agents and cloud automation, with an open-source client.
WhisperX
WhisperX community
Status not recorded
WhisperX combines Whisper-based transcription, forced alignment for word timestamps and optional speaker diarization in a self-managed pipeline.
Zed
Zed Industries
Status not recorded
Zed is an open-source editor with collaboration and optional AI services; hosted inference has separate conditions.
AnythingLLM
Mintplex Labs
Status not recorded
The all-in-one local RAG app: drop documents in, chat with them across any model.
AutoGen
Microsoft
Status not recorded
Microsoft's multi-agent framework in community-managed maintenance mode, with AgentChat, Core and prototyping tools.
Beam
Beam Cloud
Status not recorded
Hosted CPU/GPU functions, endpoints, queues and sandboxes, with separately licensed beta9 self-hosting and a BYOC control-plane option.
BentoML
BentoML
Status not recorded
A Python model-serving library that packages code, dependencies and model artifacts into Bentos and container images; managed deployment has separate terms.
CAMEL-AI
CAMEL-AI.org
Status not recorded
CAMEL-AI is an open-source Python framework for agents, multi-agent workflows and research.
CrewAI
CrewAI Inc.
Status not recorded
Python agent framework with role-based Crews and event-driven Flows; AMP is a separate managed platform.
DeepEval
Confident AI
Status not recorded
Apache-licensed Python LLM evaluation framework; the Confident AI reporting and tracing platform is separate.
Dust
Dust SAS
Status not recorded
Workspace AI agents that combine instructions, company data and tools, with seat and credit plans; hosted terms and repository licensing are separate.
E2B
E2B
Status not recorded
Isolated cloud machines for agent code and tools, with separate managed Cloud, Enterprise BYOC and Apache-licensed code paths; paused state can persist.
Firecrawl
Firecrawl
Status not recorded
Web-data API for scraping, crawling and search with Markdown or structured outputs; AGPL core and additional hosted capabilities.
Graphify
Graphify
Status not recorded
Code knowledge graphs with a local AST-based CLI and hosted plans; document/media semantic processing depends on the configured model.
Haystack
deepset
Status not recorded
Python framework for retrieval, agent and RAG pipelines; deepset hosted products are separate.
Helicone
Helicone
Status not recorded
LLM gateway and observability with traces, cost metrics and Apache-licensed self-hosting code; hosted services remain live in maintenance mode after joining Mintlify.
Hypothesis
Hypothesis
Status not recorded
Annotate the live web: public and private notes on any page — Instapaper's more annotation-native cousin for people who want the highlight to stay on the URL.
KoboldCpp
LostRuins (community)
Status not recorded
One-file GGUF inference with roleplay heritage: stories, adventures, vision models — Ollama's storytelling cousin.
Label Studio
HumanSignal
Status not recorded
Open-source platform for labelling and reviewing data across text, images, audio and video, including agent traces and RLHF workflows.
LanceDB
LanceDB
Status not recorded
The embedded vector database: serverless retrieval on object storage, zero servers to babysit.
LangChain
LangChain Inc.
Status not recorded
MIT Python agent library with model/tool integrations; LangSmith tracing and deployment are separate services.
LangGraph
LangChain Inc.
Status not recorded
MIT graph orchestration library for stateful agent workflows, with configurable persistence and human oversight.
LangWatch
LangWatch
Status not recorded
LLM observability and evals from Europe: trace agents, score quality, guard production — GDPR-native.
Letta
Letta Inc.
Status not recorded
Stateful agent product with a current Letta Code harness and optional local backend; the old V1 API server is archived.
LlamaFactory
LlamaFactory
Status not recorded
Apache-2.0 toolkit for configurable language-model fine-tuning, with CLI/WebUI workflows; model weights and training data have separate licenses.
LocalAI
mudler
Status not recorded
The OpenAI-compatible API you run yourself: drop in GGUF or other local backends, keep the same SDK, and never send the prompt off the machine.
MLflow
Databricks
Status not recorded
Open-source platform for tracking experiments, evaluating and registering models, and tracing LLM applications and agents.
MonkeyCode
Chaitin Tech
Status not recorded
Development platform with self-hosted environment, model and task management; managed online terms need separate confirmation.
nanobot
HKUDS
Status not recorded
HKUDS self-hosted Python agent with a WebUI, scheduled tasks, memory and chat integrations; models and external channels are configured separately.
Omnivore
Omnivore
Status not recorded
Open-source read-it-later: highlights, labels and a library you host yourself. The hosted app was discontinued and omnivore.app is no longer the project’s site — it now resolves to a domain-sale page — so what survives is the code, which was still receiving commits on 2026-09-19.
pyannote
pyannoteAI
Status not recorded
Python speaker-diarization toolkit with separately licensed community models and paid cloud/on-premises offerings.
Ray
Anyscale
Status not recorded
Distributed Python as a library: data, training, tuning, serving — Kubernetes' more Python-native cousin.
Scrapling
Scrapling
Status not recorded
BSD-3-Clause Python scraping framework with selectors, HTTP/browser fetchers, sessions, crawling and MCP integration; website-content rights remain separate.
Semantic Kernel
Microsoft
Status not recorded
Microsoft MIT SDK for integrating models and plugins; assess Agent Framework migration separately for new work.
SGLang
LMSYS
Status not recorded
The serving engine with RadixAttention: shared prefixes cached — vLLM's more prefix-aware cousin.
Tabby
TabbyML
Status not recorded
Self-hosted code assistant with model and IDE integrations; Apache2 core and licensed enterprise components have separate terms.
ToolJet
ToolJet
Status not recorded
Lighter open internal tools: drag, queries, a panel you can host — Appsmith's more start-native cousin.
Triton Inference Server
NVIDIA
Status not recorded
NVIDIA's inference workhorse: squeeze every GPU with batching and multi-model serving.
Ultralytics
Ultralytics
Status not recorded
YOLO vision software and models for training, prediction and export, alongside a separately billed cloud annotation/training/deployment platform.
Wan
Alibaba
Status not recorded
Alibaba's video model family: public Wan2.2 code/weights for self-hosting and separately billed cloud models. Capabilities, hardware, licenses and fees must be checked per release and service.
Wan 2.1
Alibaba
Status not recorded
The downloadable Wan2.1 video-model release, with separate text, image and first/last-frame checkpoints under Apache-2.0; hardware and content rights still require review.
Weaviate
Weaviate (SeMI)
Status not recorded
The open-source vector database: self-hostable similarity search with hybrid (dense+sparse) retrieval, built-in vectorization modules and GraphQL access — the developer-controlled alternative to managed stores.
Aider
Aider
Status not recorded
Apache2 terminal pair-programming tool that edits a repository, with configurable Git commits and model endpoints.
Botpress
Botpress
Status not recorded
The open-ish studio for chat agents: visual flows, an LLM layer, and a path to self-host when the conversation cannot live only in someone else's cloud.
Browser Use
Browser Use
Status not recorded
Browser automation library with local or cloud browsers and a separate hosted agent API.
claude-mem
claude-mem
Status not recorded
Agent memory engine that stores searchable observations, with optional hosted observation and cloud sync.
DeerFlow
ByteDance
Status not recorded
Self-hosted agent harness with subagents, skills, memory and configurable tool execution.
Evidently AI
Evidently AI
Status not recorded
Open-source evaluation and observability for ML and LLM systems, covering data drift, predictive quality and LLM output checks.
FastGPT
labring (FastGPT)
Status not recorded
The open-source knowledge-base platform: RAG pipelines, workflow orchestration, and a visual builder in one.
Gemini CLI
Status not recorded
Gemini CLI is an open-source terminal agent whose model access, charges and data terms depend on authentication; personal Google-login retirement must be checked.
Gorilla
UC Berkeley
Status not recorded
Gorilla publishes function-calling research, model releases and BFCL evaluation; code, specific model rights and benchmark scope must be assessed separately.
GPT4All
Nomic AI
Status not recorded
Nomic's local chat app: a one-click desktop installer for open models, offline RAG over your files, and an ecosystem that showed up when 'run an LLM at home' was still a dare.
Headroom
Headroom Labs
Status not recorded
Local LLM context compression through an SDK or proxy, with retrievable cached originals.
HunyuanVideo
Tencent
Status not recorded
Tencent's open-weights video generation model: one of the first open releases to rival closed leaders on motion realism and cinematic camera work, with 13B parameters you can run yourself.
Khoj
Khoj Labs
Status not recorded
An AGPL-licensed assistant for searching and chatting with notes and documents, available through self-hosting or a separately governed cloud service.
LTX 2.5
Lightricks
Status not recorded
Lightricks' open foundation models for video and audio generation, positioned with licensing terms aimed at developers.
LTX Video
Lightricks
Status not recorded
The open-source real-time video model: LTX generates frames faster than playback on consumer GPUs.
MiniCPM
ModelBest
Status not recorded
An openly licensed language-model family for local and resource-constrained deployments; current MiniCPM5 artifacts and separate multimodal releases require artifact-specific checks.
OpenHands
All Hands AI
Status not recorded
OpenHands now separates Agent Canvas, agent SDK/server and commercial hosting; deployment permissions and each component's terms need review.
PaddleOCR
PaddlePaddle
Status not recorded
OCR and document-processing toolkit with separate text-recognition, layout/table and vision-language pipelines; model and language support vary.
Pathway
Pathway
Status not recorded
Realtime data pipelines for LLM apps: live sources stay in sync instead of nightly batch.
RAGFlow
InfiniFlow
Status not recorded
Open-source RAG engine and agent platform built around document understanding and hybrid search, with visual agent workflows.
Rasa
Rasa
Status not recorded
Open-source conversational framework that grew CALM: flows you own, LLMs you plug in, and a self-host path for assistants that cannot live only in a vendor canvas.
RTK
RTK
Status not recorded
Local command-output filters for coding agents, with per-agent hooks, usage statistics and separately licensed optional pipeline/proxy components.
Supervision
Roboflow
Status not recorded
Python utilities for visualizing model predictions, tracking/counting objects and converting datasets; it requires a separate detection model.
Text Generation Inference
Hugging Face
Status not recorded
Hugging Face's serving layer: open weights in, OpenAI-shaped LLM endpoints out, tuned for big batches.
Text Generation WebUI
oobabooga (community)
Status not recorded
The Gradio cockpit for local models: chat, characters, LoRAs, every backend — LM Studio's hacker cousin.
Unsloth
Unsloth
Status not recorded
Open-source framework and desktop app for fine-tuning and running open models locally, with memory-efficient training.
Vanna
Vanna AI
Status not recorded
Text-to-SQL that learns your schema: ask questions in English, get queries trained on your DDL.
Vespa
Vespa.ai
Status not recorded
Yahoo's search engine turned platform: billion-document ranking with on-the-fly ML inference.
Xinference
Xorbits
Status not recorded
Distributed inference for LLMs and beyond: serve models across a cluster with a UI — Ollama's more cluster-native cousin.
Zep
Zep AI
Status not recorded
Long-term memory for agents via temporal knowledge graphs: Graphiti remembers facts with expiry dates.
Axolotl
Axolotl AI
Status not recorded
Open-source fine-tuning tool for LLMs, configured by YAML instead of code, with recipes for common training setups.
Bolt.diy
Bolt.diy community
Status not recorded
bolt.diy is a community-maintained, MIT-licensed web-development workspace with selectable local or cloud models; its runtime, model and deployment services have separate conditions.
Chrome DevTools MCP
Google Chrome
Status not recorded
Gives coding agents a real Chrome instance through MCP or the CLI, so they can inspect pages, debug and run audits.
Context7
Upstash
Status not recorded
Version-specific library documentation and code examples, served into AI coding tools so they stop inventing APIs.
InternLM
Shanghai AI Laboratory
Status not recorded
A downloadable language-model family whose documented InternLM3-8B-Instruct artifact offers normal and thinking modes; named code and weights declare Apache2.
MinerU
OpenDataLab
Status not recorded
Document parsing for AI: PDFs, Office files and images into structured Markdown, with layout, formula and table handling.
OpenLLM
BentoML
Status not recorded
Open models as production endpoints: BentoML's packaging around open weights.
Petals
Petals community (BigScience)
Status not recorded
Run giant open models by borrowing bits of other people's GPUs: BitTorrent-style distributed inference.
Ponytail
Ponytail
Status not recorded
Pushes coding agents toward the smallest working change: standard library over custom code, native over dependencies, one line over fifty.
Ragas
Exploding Gradients
Status not recorded
Open-source framework for evaluating LLM and RAG applications: automatic metrics, synthetic test data and production monitoring.
SuperAGI
SuperAGI Cloud Inc.
Status not recorded
The autonomous agent platform: GUI, toolkits, schedules — AutoGPT's more productized cousin.
BabyAGI
Yohei Nakajima
Status not recorded
The task-loop minimal: create, prioritize, execute, repeat — AutoGPT's more minimal-native cousin.
Caveman
Caveman
Status not recorded
Context compression, token measurement and guidance for cutting AI agent costs, installed into an existing coding agent.
CowAgent
CowAgent
Status not recorded
An open-source assistant and agent harness that plans tasks, runs tools and skills, and keeps long-term memory. Multi-model and multi-channel.
DeepSeek Harness
DeepSeek
Status not recorded
DeepSeek's open-source agent harness in developer preview, where model, tools, skills, sessions, sandboxes and scheduling are all plugins.
GPT Engineer
gpt-engineer community
Status not recorded
The repo that became Lovable: specify a project in one file, watch the agent build it.
Guardrails AI
Guardrails AI
Status not recorded
Validation and control layer for LLM outputs, checking structure, safety and policy compliance on the way in and out.
Hermes Agent
Nous Research
Status not recorded
A self-hosted open-source agent with persistent memory and self-created skills, reachable through messaging platforms.
OmniRoute
Cheaper Inference
Status not recorded
An open-source router that puts hundreds of model providers behind one OpenAI-compatible endpoint with automatic fallback.
OpenAI Swarm
OpenAI
Status not recorded
OpenAI's educational agent sketch: routines and handoffs in a few hundred lines — the blueprint the Agents SDK shipped.
YuE2
Multimodal Art Projection
Status not recorded
An open music generation model built on symbolic planning, with zero-shot covers and agentic editing of existing tracks.
ECC
ECC
Status not recorded
An operating layer across coding agents: shared skills, hooks, planning canvas, memory and security policy.
Orca
Orca
Status not recorded
An agent development environment for running several coding agents in parallel across isolated worktrees.
AgentGPT
Reworkd AI
Status not recorded
Name a goal in a browser tab and watch an agent chain tasks — AutoGPT's friendlier web cousin.
CC Switch
CC Switch
Status not recorded
Manages provider configuration, routing, MCP servers, skills, sessions and usage across several AI coding CLIs.
MemPalace
MemPalace
Status not recorded
Local-first memory for AI workflows, organised as rooms and wings, with published benchmark results and MCP tooling.
Sweep
Sweep AI
Status not recorded
Sweep pivoted from an AI code reviewer into a coding assistant aimed at JetBrains IDEs, which is what its own site now sells. The open-source repository that carried the original reviewer is still maintained, but it is no longer the product.
OmO
OmO
Status not recorded
An agent invoked by a keyword in a prompt, aimed at research and deliverable-producing work rather than coding alone.
OpenMontage
OpenMontage
Status not recorded
Open-source agentic video production: give it a brief and materials, and agents research, script, storyboard, narrate, score and render an editable video.
HyperFrames
HeyGen
Status not recorded
An open-source framework from HeyGen that renders HTML, CSS and seekable animations into deterministic MP4 video, driven from a CLI or by coding agents through installable skills.
OpenClaw
OpenClaw Foundation
Status not recorded
An open-source assistant that runs on your own machine and answers in the chat apps you already use, with a local gateway for sessions, tools and channels.
Paperclip
Paperclip Labs
Status not recorded
An open-source control plane that runs a team of AI agents — org charts, task tickets, budgets and approvals — on top of agents you bring yourself.