AI Open Source

Signup

Go to: All AI Tools AI Tools by Tool Type AI Tools by Job Type

AI Open Source

Explore top-tier open-source AI tools, freely accessible and highly customizable. Leverage active communities, transparent development, and powerful algorithms to implement state-of-the-art solutions without licensing constraints.

Signup & Stay Updated on the latest AI Tools

AI Open Source

### AI Tools available

Timmy-TUI

Timmy-TUI makes local AI agent workflows inspectable: capture model choice, tool evidence, filesystem context, workspace launch, logs and sealed MCP→CLI receipts with manifest hashes-open-source, local-first.

AgentOS

AgentOS centralizes AI agent operations in a local-first control surface: organize workspaces, agents, tasks, models, sessions, approvals, onboarding, and runtime visibility for builders and solo founders.

agmsg

agmsg lets CLI AI agents (Claude, Codex, Gemini, Copilot) exchange messages via a shared SQLite file. Vendor-agnostic, persistent, and dependency-light (bash+SQLite). No daemon or network-run unattended agent-to-agent workflows.

Agent Browser Shield

Agent Browser Shield: free, source-available browser extension that strips hidden prompt injections, trackers, cookie banners and noisy page elements so browser-based AI agents see only clean content, reducing attack surface and token waste.

Keen Code

Keen Code is an open-source CLI coding agent in Go. It uses agent-built workflows, distilled turn memory to limit multi-turn context, and skills-driven MCP tools that fetch JSON schemas on demand for efficient, transparent code generation.

Hermes Desktop

Hermes Desktop is an open-source AI agent that grows with you. Native macOS, Windows and Linux app for automating workflows, managing tasks and research pipelines. By Nous Research.

EchoFlow

EchoFlow: an open-source, BYOK AI chat built for Android-native Material UI, local encrypted conversations stored on-device for offline access, no tracking or cloud history.

Second Brain for AI

Second Brain for AI: semantic memory retrieval with deduplication and adaptive summarization. Fetch or compress context smartly, with adjustable similarity thresholds to merge overlapping notes and keep multi-step AI tasks within context limits.

Wallie V2

Wallie V2 is an open-source AI streamer that reacts to your screen and live chat, remembers topics across sessions, animates Live2D with natural lipsync, and runs locally so your API keys stay private.

Wandesk

Wandesk is an AI desktop that builds working apps on your computer from plain descriptions-calorie trackers, invoice generators, reading lists-keeps them with chat, files, tasks and memory, and is 100% local, free, and requires no account.

Parastore

Parastore simulates synthetic shoppers in 3D store layouts: they browse, pick items, impulsively buy near checkout, and generate revenue and behavioral data. Open-source (MIT) framework for layout testing and agent-based retail research.

Crew44

Crew44 is a local-first command center that turns the AI coding agents on your machine into a coordinated team. Create role-based specialists, assign runtimes, share workspace memory and handoffs, no account required, MIT licensed.

Harbor

Harbor runs and auto-wires local AI services with one command. Start frontends, LLM backends, search, voice and more from a 129-service catalog; includes agent launch and an OpenAI-compatible local API.

Phasr

Phasr centralizes engineering workflows: run 100+ persistent workspaces, manage multiple repositories, pin key commands, and coordinate AI agents-reducing context switches so teams stay organized and move faster.

Pi Coding Agent

Pi Coding Agent is a minimal, hackable terminal harness for building and customizing AI coding-agent workflows. Keep the core small and extend or replace context management, sub-agents, permissions and workflows via extensions.

DecisionBox for Databricks

DecisionBox for Databricks connects to your workspace, auto-writes SQL, validates findings against your data, and delivers a ranked backlog of verified, prompt-free insights. Read-only, Unity Catalog scoped; supports Serverless, Pro, Classic SQL w...

MartinLoop

MartinLoop is an open-source control layer for coding agents that enforces hard budget caps, verifier-gated retries, scope/secret safety, rollback evidence, and JSONL run logs-keeping autonomous runs fast and accountable.

CLI Market

CLI Market is a Python connector that maps VTEX storefront APIs into one JSON schema. Query products across retailers (Carrefour BR, Sainsbury's UK, Wong PE) for programmatic and AI-agent access. Open-source MIT; pip install cli-market

AgentRail

AgentRail: an open-source task lifecycle API that turns GitHub Issues and Linear tickets into routable tasks, delivers structured CI summaries, ranks PR comments by severity, and enforces per-agent scoped keys; runs locally, free to self-host.

Standboy

Standboy is a Game Boy in your editor sidebar that wakes when your AI agent runs and hides when it stops. Play GB/GBC/GBA ROMs with saves stored locally. Free, MIT-licensed, no telemetry.

Gradient Bang

Gradient Bang: an open-source, massively multiplayer LLM-driven space trading game. Talk to and manage ship AIs, design sub-agents in a Vercel sandbox, handle long LLM contexts, and compete on live leaderboards.

Mi

Mi is a 30-line agentic loop with two tools and an LLM plug-in-compact, optimized for local LLMs, token-efficient, and featuring sandbox mode to run in a container for safer, isolated execution.

ShioriCode

ShioriCode: a desktop UI for coding-agent CLIs that keeps each run as a project-aware thread tied to branch/workspace, streams activity into a timeline, and previews generated diffs without leaving the app. Supports any authed CLI or Shiori host.

Linchpin

Linchpin is an open-source runtime for hosting managed agents on a single VM: per-session Docker sandboxes, persistent event logs, HTTP/SSE API, built-in tools and encrypted credential vaults for quick self-hosting.

Agentmemory

AgentMemory gives AI coding agents persistent cross-session memory: it captures agent activity, compresses it into structured memories, indexes via hybrid search, and injects relevant context. 100% open source.

Memori

Memori captures AI agent traces and tool results into structured, persistent memory-enabling scoped, async recall, automated daily briefs, and reduced prompt bloat across projects.

Pipali

Pipali is a desktop AI coworker that automates work: research files and the web, draft briefs, spreadsheets, emails, reports and personal apps, run recurring tasks and monitor events with sandboxed permissions.

OpenHuman

OpenHuman: a privacy-first AI agent with two-minute setup and one-click Gmail/Slack/Telegram/Notion/GitHub connections. Runs locally with encrypted, persistent memory. Open-source and free to start.

Radar

Radar is an open-source, local-first Kubernetes UI that consolidates cluster workflows: ownership-aware topology, live watches, Helm + GitOps, traffic flows, cost insights and audits-no signup required.

Gas City 1.0

Gas City 1.0 is a software factory that runs coding agents to deliver enterprise-grade features, quality and schedule-accelerating development for large-scale projects while preserving production standards.

Gas City 1.0

Gas City 1.0 is a software factory that runs coding agents to deliver enterprise-grade features, quality and schedule-accelerating development for large-scale projects while preserving production standards.

Claude Code & Codex Usage Trading Cards by Rudel

Rudel turns Claude Code & Codex sessions into trading cards that summarize archetype, token usage, model mix, repo breadth, cost, outputs and errors. Free, open-source, and self-hostable.

Miaw AI secretary

Miaw AI Secretary is a lightweight desktop assistant with a floating chat that helps you brainstorm, draft, summarize and handle quick tasks without leaving your workflow. Open-source and minimal.

PandaProbe

PandaProbe captures full AI-agent sessions (LLMs, tools, APIs) for tracing, evaluation, monitoring and debugging, so teams can measure production quality and reliability. Open-source + cloud.

Redesign by Nodewave

Redesign by Nodewave: a local app that uses your React components and Claude Code prompts to create branded social posts. Edit in a visual editor and export PNGs for LinkedIn/Instagram. Open-source (MIT). No cloud, no accounts.

WUPHF by Nex.ai

WUPHF by Nex.ai: a self-hosted, open-source Slack-style hub for AI agents, to coordinate multiple models, run playbooks, and auto-build a git-backed team wiki so agents onboard themselves.

Euphony

Euphony is an open-source tool that renders Harmony JSON/JSONL and Codex logs into interactive timelines. Load files or paste, filter with JMESPath, inspect metadata, edit in-browser, embed as a Web Component; client-side by default.

MiMo-V2.5 Voice

MiMo-V2.5 Voice - an 8B MIT-licensed ASR for bilingual Chinese-English transcription. High accuracy on dialects, noisy audio, code-switching and lyrics, with prosody-driven punctuation for production-ready transcripts. Available on HuggingFace.

Free LLM API

Free LLM API aggregates free tiers from major LLM providers into one open-source API. Plug in your keys and get routing, rate limiting, and automatic fallback for prototyping without billing or credit cards.

kimiflare

kimiflare: open-source Claude/Kimi code clone running Kimi K2.6 on Cloudflare Workers AI. Self-hosted, low-latency code assistant for private, fast code generation; MVP was built recursively using kimiflare itself.

Cai

Cai is a privacy-first, local Mac assistant you own that eliminates constant ⌘Tab switching by handling quick copy-paste, prompts and follow-ups on your machine. Free, MIT-licensed.

YourMemory

YourMemory filters agent context with biological decay, letting noise fade while reinforcing frequently used facts. Local-first DuckDB storage and a graph engine surface architectural neighbors for higher recall and fewer stale hallucinations.

Stet

Stet is a minimalist macOS voice-input app that cleans up your dictation while preserving your voice. Open-source and privacy-minded: bring your own API for free or use hosted transcriptions with zero-data-retention for $6.99/month.

SureThing.io

Turn any GitHub repo into a one-click, production-ready AI employee with memory, a live dashboard and your business context-so founders and teams deploy open-source AI fast, no setup.

Reka Edge

Reka Edge is a 7B vision-language model for Physical AI. Its ConvNeXt V2 encoder uses 3× fewer image tokens for sub-second video analysis, low-latency object detection and agentic tool use on edge devices.

CC-BEEPER

CC-BEEPER: a floating retro pager for Claude Code that surfaces agent status, permission prompts and results, with voice dictation, spoken recaps, hotkeys and auto-accept modes. Runs locally in Swift-no accounts, telemetry, or API keys.

Caveman

Caveman: a Claude prompt that removes verbosity for coding tasks-≈75% fewer tokens, ≈3x faster responses, plus terse commits and one-line PR comments. Free, MIT, one-line install.

Open Agents

Open Agents is an open-source reference platform to build and deploy cloud coding agents with agent brains, durable workflows, secure sandboxes and a multi-model AI gateway-optimized for large monorepos, integrations and institutional knowledge.

ContextPool

ContextPool is a local persistent memory layer for AI coding agents: it extracts actionable engineering insights from past sessions and auto-loads relevant context while keeping raw transcripts local.

Assemble

Assemble: a config-first AI work system that routes tasks by difficulty, keeps cross-session memory, and converts complex jobs into spec-driven workflows with review and test boards-zero dependencies, multi-platform.

Clicky

Clicky is an on-Mac AI companion beside your cursor that sees your screen, replies by voice, and points to UI elements. It guides you through tasks in real time inside your workflow, with private streaming and open-source code.

Waypoint-1.5

Waypoint-1.5 generates playable first-person action environments with fluid shooting and responsive frame rates on regular GPUs, delivering mechanics-first, locally-run FPS experiences that feel like a polished game.

Rosentic

Rosentic checks every PR against all active branches to detect cross-branch conflicts before merge. Deterministic scans, one YAML install, no signup - catch compatibility issues before they hit main.

Marmot

Marmot is a lightweight, open-source data catalog (single binary + Postgres) that indexes 25+ sources, provides full lineage and an MCP server so LLMs query a single source of truth - deployable in minutes.

OpenBrowser-AI

OpenBrowser-AI gives LLMs direct Chrome control via raw CDP, running Python in a persistent namespace to batch actions and compress page state-using ~2.6x fewer tokens for faster, efficient browser automation.

/dev for Claude Code

/def for Claude Code turns Claude into a Tech Lead: PRD alignment, architecture and task decomposition, isolated worker worktrees, QA checks, security audits, and structured code review. Open-source (MIT).

Open Claude in Chrome

Open Claude in Chrome: open-source (MIT) clean-room clone of the Claude extension. Adds browser automation with 18 tools: click/type/screenshots/JS/eval/accessibility/console+network. Chrome & other Chromium browsers; no domain blocklist.

Google Gemma 4

Google Gemma 4 - an open (Apache 2.0) model family for advanced reasoning, multimodal inputs, 256K-context and agentic workflows. Offers frontier-level intelligence-per-parameter that runs efficiently on phones, laptops and local servers.

Trinity-Large-Thinking by Arcee

Trinity-Large-Thinking by Arcee: an open-source SOTA model for developers who want models they can inspect, post-train, host, distill and own. Ranked #2 on PinchBench (KiloClaw); $0.90 per 1M output tokens (~96% cheaper).

Skills Janitor

Skills Janitor cleans your skills folder: finds unused, duplicate or broken skills, previews proposed removals, and safely prunes them. Single-file bash+python3 (no deps), MIT-licensed, free, no telemetry-reclaim context space and tidy your setup.

OpenGyver

OpenGyver is a free, open-source CLI and Claude Code skill that converts formats and units-currency, color codes, timestamps, images, JSON, UUIDs-extensible via modules for custom conversion commands.

Arlopass

Arlopass is a browser extension that lets web apps use your AI providers without exposing API keys-approve requests, pick models, and route calls locally. No server proxies; keys remain in your OS keychain.

CraftBot

CraftBot: an independent proactive agent that saves tokens via dynamic tool loading and task-mode switching, plans and initiates daily/weekly/monthly tasks with your approval, consolidates nightly memory, and runs from one command.

Domscribe

Domscribe links live DOM to source code so AI agents find exact components quickly: fetch DOM snapshots, props and state from a file+line, or click elements to resolve and edit the correct source on first try.

Glance

Glance gives Claude Code a real Chromium browser via MCP-tools for browsing, screenshots, form-filling and a plain-English test runner that executes steps and returns pixel-accurate inline screenshots. Open-source; one-command install.

Lexaclaw

Lexaclaw runs locally to find your corporate documents, extract compliance obligations, and centralize deadlines into a step-by-step system. Open-source, private, and free for startups - soon with autonomous filing via your AI agent.

CrabTalk

CrabTalk: an 8MB Rust daemon that composes your CLI tools via PATH. Components connect as services, crash or swap independently. Open-source, one-curl install-build apps on top, not beneath.

Jentic Mini

Jentic Mini is a self-hosted API execution layer that lets agents discover and call real APIs without leaking credentials. It brokers requests, injects Fernet-encrypted credentials at runtime, and issues per-agent revocable toolkit keys with audit...

AlphaClaw Apex

AlphaClaw Apex centralizes AI analytics, automates workflows, and delivers concise insights to speed decisions and reduce operational costs.

AI Skills Manager

AI Skills Manager unifies skills across AI coding agents, lets you copy skills between agents and install from GitHub repos. Free for Windows - streamlines managing varied formats and folders.

Scheduled

Scheduled is an open-source AI agent inside Gmail: reads threads, checks your calendar, and drafts meeting replies in your voice for quick review (draft-only) or autonomous sending-no extra app or links.

Lore

Lore is a local, free, open-source memory assistant that answers questions from your notes and journals with one keystroke. 100% private, no cloud or API keys; fast local LLM search for researchers, devs, and thoughtful users.

Parallax

Parallax is a local AI coding agent that turns Linear or GitHub issues into task plans, waits for your approval, implements changes in isolated branches, and opens a reviewable PR so you keep control of planning and review.

Masko Code

Masko Code overlays animated mascots on macOS 14+ to supervise AI agent sessions: speech-bubble notifications, one-key approvals (⌘1), session jumps (⌘M) and quick shortcuts. 100% local, open-source (MIT), customizable mascots.

Google Workspace CLI

AI tool for Google Workspace CLI

DenchClaw

DenchClaw is an open-source OpenClaw framework and local Mac app: a DuckDB-backed CRM, coding agent and file-tree interface to query, edit and automate workflows. Integrates with Gmail, Apollo and Notion - private, extensible, and local.

OpenMolt

OpenMolt is an open-source Node.js framework for programmatic AI agents. Define agents in code with instructions and tools to gather data, call APIs, generate outputs and update systems.

Phi-4-reasoning-vision

Phi-4-reasoning-vision: a compact open-weight multimodal model using mid-fusion to combine rapid visual perception with deep chain-of-thought, enabling efficient computer-use agents and precise complex-math problem solving.

LTX Desktop

LTX Desktop is a free, open-source local video editor built on the LTX-2.3 engine, optimized for NVIDIA GPUs. Edit and render on-device with no cloud dependency or per-generation fees-keep your data local and workflows private.

Vet

Vet is a fast, local, open-source code-review tool from Imbue that verifies coding agents using conversation history, catching silent failures, unrun tests, logic errors, and missing edge-case handling in PRs.

Step 3.5 Flash

Step 3.5 Flash: 196B sparse MoE (11B active/token) delivering high-throughput coding (up to 350 tok/s), 74.4% SWE-bench, clean long-context handling and seamless OpenClaw/OpenRouter integration for reliable agentic workflows.

Context Gateway

Context Gateway is a context-compression proxy that reduces token costs, lowers latency, and improves model accuracy for agents (Claude Code, OpenClaw, Codex). Includes instant compaction, spend caps, Slack alerts; core code open source.

Personal AI Memory

Personal AI Memory silently captures and indexes your chats locally - no servers or API keys - so you never lose context across LLMs or browsers. Import or export ChatGPT and Google Takeout data with one click.

DialogLab

DialogLab is an open-source framework to author, simulate, and test multi-party conversations with humans and AI. Use a visual scene builder, phase-based flow control, human-in-the-loop testing and verification dashboard.

OpenObserve

OpenObserve: a cloud-native observability platform (Rust) that unifies logs, metrics and traces in one high-performance, stateless engine. Store petabytes on S3/MinIO/GCS and query everything with SQL while keeping infrastructure minimal.

Springfield Oracle

Springfield Oracle is a free, open-source database that verifies viral Simpsons clips and claims. Every entry is sourced to original episodes, checked against real events, and open to community submissions.

Openstatus MCP Health Checker

Openstatus MCP Health Checker performs protocol-level AI checks by running full JSON-RPC handshakes (initialize, ping, tools/list), showing exact payloads and negotiated versions, and parsing RFC 9728 headers to reveal token requirements.

PromptURLs

Type a prompt once to generate pre-filled, shareable links for ChatGPT, Claude, Gemini, Perplexity, Grok and more. Compare outputs instantly, save to Notion, build a one-click prompt library. Free, browser-only, no sign-up, open source.

IronClaw

IronClaw isolates and protects API keys while proxying AI requests, preventing key leaks and providing secure, drop-in integration with OpenAI-compatible services.

Claudebin

Easily share Claude conversations. Claudebin captures prompts, responses, and edits for seamless collaboration and review of your AI sessions. Open source.

Mengram

Mengram - open-source memory API capturing semantic, episodic and procedural memories. When an AI completes a task, Mengram saves the steps as reusable procedures so agents recall optimal workflows. Free; 60-sec setup.

Kollect Voice Agent

Kollect Voice Agent converts static forms into real-time AI conversations: speak answers, the AI listens and guides the survey, and you can generate forms by describing them. Open-source and feedback welcome.

Baseline Core

Baseline Core: free open-source skills system that wires your business context into any AI tool. Run a command, answer a few questions, and AI agents produce product work-research, PRDs, sprints, user flows, design, prototypes.

ClawMetry for OpenClaw

ClawMetry for OpenClaw: real-time observability-inspect sub-agent activity, files, commands, tool calls, thoughts, session timelines and per-session cost. Open-source; pip install clawmetry.

claude-devtools

claude-devtools reconstructs Claude Code sessions from local logs: shows file reads with syntax highlighting, inline diffs, tool calls and per-turn context breakdown. Open-source, local-only, zero network calls.

MiniMax-M2.5

MiniMax-M2.5: an open coding model with SOTA SWE-Bench 80.2%-the first open model to beat Sonnet; only Anthropic's Opus and OpenAI's GPT-5.2 Codex score higher. Try it free for a week on Kilo Code (offer ends Feb 19).

OpenBug

OpenBug is an open-source CLI+GUI AI that diagnoses bugs by reading logs and code, generates code diffs to fix issues, saves fixes as runbooks in Git, and runs in your terminal with no SDK or config.

Atomic Bot

Atomic Bot gets you from download to a running OpenClaw agent on your Mac in ~60 seconds-local execution keeps your workspace private and setup is simple so you can focus on tasks, not configuration.

TexTab

TexTab: Select text, hit a shortcut, get AI responses without opening a browser. Open source, BYO API keys (OpenAI, Anthropic, Gemini, OpenRouter), no subscriptions - you own your data and control costs.

.

As seen on