AI agents

unreallabsai/unreal-agent

A Go harness built around async tool calls and resumable sessions — Unreal Agent is a Go library for building LLM agents where tool calls become serializable operations executed asynchronously, not blocking the event loop. It includes a coordinator, forkable append-only session store, and both a cloud runner and TUI, aimed at developers needing crash recovery and idempotent redelivery.

Async-first agent harness

BuilderIO/skills

Composable slash-command skills that plug into coding agents — A CLI-installable set of modular skills for AI coding agents, covering PR review, delivery pipelines, visual planning/recaps, and watchdog-style auditing. Useful for teams automating agent workflows who want ready-made, swappable commands instead of building each capability from scratch.

Skills for agents

warpdotdev/warp

Warp turns the terminal into an agentic coding workspace — Warp reimagines the terminal as a development environment with a built-in coding agent, or lets you plug in Claude Code, Codex, Gemini CLI and others. Its new Warp Factories let teams automate issue triage, spec writing, and PR review, with progress visible on a public contributions dashboard.

Warp is an agentic development environment, born out of the terminal.

XiaomiMiMo/MiMo-Code

Xiaomi's terminal-native coding agent with persistent project memory — MiMoCode is a terminal AI coding assistant that reads/writes code, runs commands, and manages Git while keeping a persistent memory of your project across sessions. It also powers Xiaomi MiMo Desktop, which adds smart multi-model orchestration, cached long-task cost control, and browser automation for broader office and design workflows.

MiMo Code: Where Models and Agents Co-Evolve

superdesigndev/superdesign

The original IDE extension behind Superdesign, now archived — This was the first open-source AI design agent living inside IDEs like Cursor, Windsurf and VS Code, turning prompts into UI mockups, components and wireframes. It's no longer maintained: Superdesign is now a web app, with IDE users pointed to a separate design-skill repo for Claude Code and Cursor.

AI Product Design Agent - Open Source

spotify/portal-ai-plugins

Spotify Portal joins Claude Code, Codex, Cursor and pi — Plugins that bring Spotify Portal's internal developer catalog into AI coding agents via the Portal CLI, letting agents search services, generate briefings, run diagnostics, and invoke actions. Includes the shunt plugin, which offloads bulk file reads and boilerplate to cheaper worker models. Useful for teams already using Portal who want catalog-aware coding agents.

Fosowl/agenticSeek

A fully local, voice-enabled agent that browses and codes for you — AgenticSeek positions itself as a private, local alternative to Manus AI: it browses the web, writes and debugs code, and plans multi-step tasks using local reasoning models like DeepSeek-R1, with no cloud or subscription fees. Voice interaction is in progress. Useful for privacy-conscious developers wanting autonomous agents without API costs.

Fully Local Manus AI. No APIs, No $200 monthly bills. Enjoy an autonomous agent that thinks, browses the web, and code for the sole cost of electricity.

google/agentmesh

A private, zero-trust P2P network for AI agents — Agent Mesh lets AI agents publish and discover tools and models across a secure peer-to-peer network, with nothing reachable by default and identity tied to your OIDC provider. It works across laptops, containers, clusters and NAT'd phones, routing model requests via an OpenAI-compatible endpoint. Useful for teams building distributed agent systems needing verifiable, policy-controlled service access.

Agent Mesh

agno-agi/agno

Own your full agent stack, from build to runtime to UI — Agno is a Python framework paired with the AgentOS runtime and UI, letting teams build, run, and manage agent platforms themselves. It adds JWT-based RBAC, Postgres-backed traces, and starter templates for Railway, Docker, AWS or GCP deploys, giving developers control over data, memory, and security instead of relying on a hosted service.

The self-improving agent platform

elvisun/newsjack

Skill pack that turns coding agents into a newsjacking PR team — Newsjack.sh installs slash-command skills like /newsjack-detector, /fact-check and /coverage-tracker so agents such as Claude Code or Codex can spot timely story angles, verify sources, and deliver reports to Slack. Aimed at marketers and PR teams who already run AI agents and want ready-made monitoring workflows instead of building prompts from scratch.

The open-source skills that turn your agent into a full PR team.

Companion-Inc/feynman

A CLI agent that automates literature reviews and paper audits — Feynman is an open-source CLI research agent with built-in workflows like /lit, /deepresearch, /audit and /replicate for literature review, paper critique, code-vs-claims auditing and replication planning. It targets researchers and ML practitioners who want structured, citation-checked investigation instead of ad-hoc chat prompts.

The open source AI research agent.

bealesh/neckbeard

Cloud infrastructure plugin that won't let agents guess costs — neckbeard plugs into coding agents like Claude Code and maps an app's needs onto a versioned, CI-tested OpenTofu module catalog instead of letting the model invent HCL. It outputs a hash-verified blueprint, supports AWS, GCP, and Azure, and flags unsupported needs (like Redis) rather than faking them.

Agent plugin that scaffolds tested, cost-honest cloud infrastructure from a versioned module catalog

egma-ai/jev-code-reviewer

A local-only PR reviewer that sorts diffs by human attention priority — Jev classifies PR changes into P0/P1/P2 priority tiers so reviewers see only what matters by default, with natural-language explanations of diffs and original code one toggle away. It runs locally via CLI, agent skill, and a Chrome extension, posting nothing to GitHub—built for devs reviewing their own coding agent's PRs.

Review behavior, not just diffs. Jev prioritizes human attention; OpenAI explains the changes. Local CLI + agent skill + GitHub extension.

stacklok/toolhive

Containerized, policy-enforced MCP servers for production use — ToolHive runs MCP servers in isolated containers, applying per-request identity and access policy and cutting token usage up to 85% via semantic tool search. A Kubernetes operator adds IdP integration, OTel tracing, and audit logs, serving developers, platform engineers, and self-hosting enterprises.

ToolHive is an enterprise-grade platform for running and managing Model Context Protocol (MCP) servers.

atomicstrata/llm-wiki-compiler

A CLI compiler that turns raw sources into an interlinked wiki — llmwiki compiles raw documents into an interlinked, Obsidian-friendly wiki using LLMs. Version 1.5 decouples wiki behavior from storage, adds bearer-authenticated Converse gateways, removes the fixed concept page limit, and gives the live viewer auto-refresh after compiles. Useful for anyone building and maintaining a growing knowledge base from scattered source material.

The knowledge compiler. Raw sources in, interlinked wiki out. Inspired by Karpathy's LLM Wiki pattern.

openclaw/openclaw

A self-hosted AI assistant that lives in your chat apps — OpenClaw runs locally and meets you across Discord, iMessage, Slack, Teams, Telegram, WhatsApp and native apps, keeping state, memory and credentials on your own hardware. Model backends like Claude, Codex or local models plug in interchangeably, and one Gateway scales from a personal laptop setup to shared team deployments.

The AI that really does things. Any OS. Any Platform. The lobster way. 🦞

browserbase/stagehand

An SDK that gives Playwright scripts AI-driven browser agents — Stagehand combines Playwright automation with AI actions, letting developers mix deterministic code and natural-language commands to extract data or interact with any website. It integrates with tools like Claude Code, Codex, Eve, and Mastra, targeting developers building browser agents in TypeScript or Python.

The SDK to extract data and interact with any site on the web. Get started with Claude Code, Codex, Eve, Mastra, and more.

arc53/DocsGPT

Self-hosted AI agents that answer strictly from your own documents — DocsGPT is a private, self-hosted platform for building AI agents and assistants grounded in your own docs, with a visual Agent Builder, deep research, document analysis, multi-model support and API connectors. A one-line install script makes it quick to run locally, aimed at teams wanting enterprise search without sending data to third parties.

Private AI platform for agents, assistants and enterprise search. Built-in Agent Builder, Deep research, Document analysis, Multi-model support, and API connectivity for agents.

atsushi-green/ds-ai-coding-skills

Pre-built AGENTS.md skills to guide AI coding tools for data scientists — A collection of AGENTS.md router files, Claude/Codex skills, and prompts designed for data science workflows, covering Python, SQL, Polars, visualization, and statistics. It includes paired 'diagnostics' skills that force AI assistants to output standardized charts alongside analysis results, with a gallery of generated examples.

A collection of AGENTS.md configurations, skills, and prompts tailored for Data Scientists to maximize productivity with AI coding assistants.

nwiizo/rust-learning-lab

A shared Rust-learning skill for both Codex and Claude Code — A plugin bringing a single learn-rust skill to both Codex CLI and Claude Code, explaining Rust syntax, ownership, and control flow while reviewing small code changes. Supports English and Japanese, and can turn detailed reviews into revisitable Markdown documents. Useful for developers learning Rust alongside AI coding assistants.

Learn and review Rust through syntax, rationale, practice, and debugging. A bilingual plugin for Codex and Claude Code.

microsoft/apm-action

A GitHub Action that wires APM agent packages into CI — This action installs Microsoft's APM (Agent Package Manager) CLI and deploys agent primitives like instructions, prompts, skills and agents directly in CI, with no config needed beyond a one-line step. It adds isolated mode for inline dependencies without an apm.yml, a setup-only mode mirroring actions/setup-node, and an update mode to refresh branch/tag-pinned dependencies against the lockfile.

GitHub Action for Agent Package Manager

ai-mode/ai-mode

A modular AI assistant layer built into Emacs itself — AI mode brings code completion, interactive chat, and code manipulation into Emacs, with buffer-bound chat sessions and context-aware assistance. Its modular design supports multiple AI backends and lets users extend it with custom commands, aimed at Emacs users wanting integrated AI tooling rather than separate copilot plugins.

AI mode for Emacs

wileai/pi-codex-connectors

Bridge Codex connectors and web search into any Pi model — A Pi extension that exposes your already-connected Codex (ChatGPT) apps and tools to other Pi models, using three generic tools instead of hundreds of schemas. It also adds direct web search via Codex's alpha search endpoint, without triggering a model turn. Useful for Pi users who want ChatGPT connector access outside Codex itself.

Use every connected Codex (ChatGPT) connector from any pi model

anupray95/AI-Agent-on-a-NOKIA

Turning a Nokia 110 feature phone into an AI chat agent — A hobbyist reverse-engineered the Nokia 110 4G firmware, hijacking the Calculator app to build a native chat interface wired to DeepSeek's API. Typed messages trigger real phone functions like calls, alarms, torch and battery checks via tool calls. It's a RAM-loaded prototype, interesting for firmware hackers and retro-phone tinkerers.

Reverse-engineered an old Nokia's firmware and added a custom AI Agent. Native functions like calls, alarms, or alarms can now be easily handled just by chatting.

rociiu/talorys

A personal AI agent you own, hosted on your own Cloudflare account — Talorys is a self-hosted AI assistant deployed with one npx command into your own Cloudflare account, no external server or vendor database involved. It handles chat, long-term memory with optional semantic recall, tasks, notes and scheduled automations, aimed at single users wanting a private, free-tier-friendly, telemetry-free assistant.

Your personal AI agent in your own Cloudflare account. Chat, memory, tasks, notes and scheduled reminders, deployed with one command: npx create-talorys@latest. Free-tier friendly, single-user, no telemetry.

anthropics/financial-services

Claude agents built for investment banking and wealth management desks — A collection of reference Claude agents, skills, and connectors for IB, equity research, private equity, and wealth management workflows. Deploy as Cowork plugins or via the Managed Agents API from one source; outputs like memos and models are staged for human review, not automated decisions.

zeroclaw-labs/zeroclaw

A self-hosted Rust runtime for owning your AI agent end to end — ZeroClaw is a single Rust binary that runs a personal AI agent on your own machine, connecting to ~20 LLM providers and 30+ channels like Discord, Telegram, Matrix and email. It targets people who want full control over their agent, data and host rather than relying on a hosted service.

Fast, small, and fully autonomous AI personal assistant infrastructure, any OS, any platform — deploy anywhere, swap anything 🦀

esengine/DeepSeek-Reasonix

A terminal coding agent built around DeepSeek R1 reasoning — Reasonix is a TUI/CLI coding agent that pairs DeepSeek R1 with tool-use and prompt caching for complex software engineering tasks. It ships an agent framework, ACP protocol, plugin system (dsh), and extension docs, making it useful for developers who want a scriptable, extensible terminal coding assistant rather than a chat-only tool.

A reliable coding agent for complex software engineering tasks.

appwrite/appwrite

Appwrite now bills itself as an MCP and agent-first backend — Appwrite adds a hosted MCP server so coding agents like Cursor, Claude Code, and Codex can operate directly on a live project, alongside its existing Auth, Databases, Storage, Functions, Messaging, and Hosting modules. It's for developers and AI agents building apps that need self-hosted or managed backend infrastructure.

The open-source cloud for agents & devs. Including Auth, Databases, Storage, Functions, Messaging, Hosting, Realtime, WAF and more

composo-ai/DishwasherBench

Benchmarking LLMs by making them load a dishwasher — DishwasherBench turns dishwasher loading into an LLM benchmark, using rigid-body physics and simulated wash chemistry on an exact model of a Bosch SMV4HAX40G. Humans play the same game, called Spotless, in a browser, while models act through MCP, with every episode deterministic and replayable for verification.

A dishwasher-loading game on a physical simulation of one real machine, built to benchmark language models and playable in a browser. The game inside is called Spotless.