A .NET inference engine for local GGUF models offering a CLI, browser chat and Ollama/OpenAI-compatible APIs with CPU, CUDA, Metal and Vulkan backends.zhongkaifu
A self-hosted LLM router and cost optimizer. It classifies prompts, sends simple ones to cheaper or local models, verifies answers, escalates when needed, and exposes OpenAI- and Anthropic-compatible endpoints for coding agents and chat clients.NadirRouter
Termio is a terminal-first agentic development environment that manages parallel CLI coding agents, keeps sessions alive in a daemon, notifies when an agent needs attention, and mirrors sessions to iPhone.termio-sh
Kooky is a minimal macOS terminal built for AI coding, with workspace tabs, split panes, one-click launches for many coding agents, session resume, and live agent activity status.iAmCorey
A universal AI gateway that unifies multiple Ollama, vLLM, llama.cpp and OpenAI-compatible backends behind one managed API with authentication, rate limiting, routing and usage analytics.ParisNeo
Headroom Desktop is a menu-bar app for macOS, Windows, and Linux that routes Claude Code, Codex, OpenCode, and Grok Build through a local proxy, reversibly compressing tool output and logs to roughly halve token costs while showing savings analytics.gglucass
Official docs and code samples for UncensoredLabs, a hosted OpenAI-compatible API and browser chat for uncensored DeepSeek and GLM text models, including pricing, streaming and client setup guides.UncensoredLabs
Pooled runs large open models across browsers by splitting model layers among devices in a room over WebRTC and WebGPU. Users open a link to chat or use Code mode, and the room can expose an OpenAI/Anthropic-compatible API to other tools.Nehanth
A local-first runtime for production AI agents, providing persistent sessions, sandboxed execution, tool governance, credential vaults, memory, audit trails, and a console.sandbaseai
Lynkr is an LLM gateway for AI coding tools. It routes requests across local and cloud providers by task complexity, compresses tool output, caches repeated prompts, and shows spend and routing analytics, all without changing your editor or agent.Fast-Editor
Harness Score is a deterministic CLI that audits the context files, skills, hooks, sensors and CI around an AI coding agent, scoring repository harness maturity from L0 to L4 with concrete fixes.paladini
SkillHub Desktop is a desktop app for managing AI coding skills across tools. Browse a catalog, install or create skills, and sync them between Claude Code, Cursor, OpenCode, and others.skillhub-club
A local-first analytics dashboard that unifies AI coding sessions from 17 editors, tracking costs, models, tools and activity, with an optional relay server for team-wide session sharing.f
A benchmark suite for measuring whether language models can learn new knowledge from context, with expert-annotated tasks, rubrics, inference scripts, and evaluation code.tencent-hunyuan
AgentBox teleports a project into an isolated local or cloud VM and runs multiple coding agents in parallel, with snapshots, pause/resume, a browser, and remote access built in.madarco
A terminal pixel-art office that visualizes AI coding agent sessions as coworkers, showing who is working, waiting for approval, done or burning tokens at a glance.pixtuoid
A vLLM patch and hand-written kernels for serving frontier MoE models on consumer Blackwell GPUs via aggressive quantization and tiered residency.kacper-daftcode
Dubbo-Go-Pixiu is a high-performance AI and API gateway that unifies access to LLMs and MCP servers, bridges HTTP, gRPC and Dubbo protocols, and adds plugin-based governance.apache
A Rust API gateway that translates Anthropic and OpenAI-compatible requests to Kiro/Amazon Q, with credential pooling, proxy rotation, usage accounting, and an admin console.zyphrzero
A Claude Code plugin that draws a live map of the agent's work beside the chat, lighting up tool calls, model steps, hooks, skills and subagents as they happen.alexeyw
A remote control and session manager for coding agents such as Claude Code, Codex, Cursor, and OpenCode. It lets you assign tasks from a phone or web UI, follow progress, and get notified when work finishes.BennyKok
Coralline is a Powerlevel10k-inspired statusline for Claude Code. It renders directory, git, model, context, rate-limit and cost segments, themed subagent rows and optional burn-rate forecasting, installed entirely locally with no API calls.Nanako0129
A transparent reverse proxy for the Tavily API that pools multiple API keys behind one master key, with quota-aware rotation, failover, an MCP endpoint and a web dashboard.xuncv
Authority and security infrastructure for AI agents that checks exact authorization before consequential tool calls and produces verifiable evidence.emiliaprotocol
Heard is a macOS menu-bar app that speaks short summaries of what your coding agents are doing, so you can step away and still know when Claude Code or Codex finishes, fails, or needs approval. The open-source engine runs local Kokoro voices or your own cloud voice keys.heardlabs
ClawLess is a serverless, browser-based runtime for Claw AI agents built on WebContainers. It provides a fully sandboxed Node.js environment with an editor, terminal, YAML policy engine, audit logging and a pluggable SDK, all without a backend.open-gitagent
A macOS Dynamic Island app for monitoring, approving, and jumping into Claude Code and other AI coding agent sessions from the notch, with iPhone sync and remote launch.MioMioOS
A local token and cost tracker for Claude Code, Codex and Kimi Code, adding status lines, GitHub-style heatmaps, a session sidebar and daily or weekly usage reports, all stored on your machine.stormzhang
A local-first desktop and PWA workspace for OpenClaw that runs parallel agent tasks, stores artifacts durably, and coordinates multi-agent teams from one UI.clawwork-ai
A PWA that gives you real terminal access to your machine from a phone or tablet, letting you run and watch AI coding agents such as Claude Code remotely.my-claude-utils
A cross-platform on-device AI runtime and SDK collection for running LLMs, speech recognition and text-to-speech locally in mobile apps, desktop software and games.xybrid-ai
A multi-workspace terminal aggregator with built-in Claude Code and Codex agent panels, file browsing, Git views, snippets, and remote or mobile access.tony1223
Gemini CLI Desktop is a desktop and web GUI for Gemini CLI and Qwen Code, adding project management, visual tool approvals, diff viewing, conversation search, and MCP server management.Piebald-AI
A desktop HUD for Claude Code, Codex and 25+ coding agents that shows quotas, token usage and sessions at a glance, lets you approve requests, and syncs with iPhone and Apple Watch.jazzenchen
Willow Inference Server is a self-hosted, CUDA-optimized inference server exposing Whisper speech recognition, text-to-speech and LLM inference over REST, WebRTC and WebSocket APIs.toverainc
Cordum is an out-of-process governance control plane for autonomous agents, offering policy enforcement, human approval gates, workflow scheduling, and signed audit trails.cordum-io
Termany is an agent-native terminal for running many coding agent sessions in one window, with session management, diffs, ports, skills, and token-cost tracking.thinkany-ai
A local multi-account proxy for Claude Code and Codex that pools subscriptions, API keys, and third-party backends, rotating accounts as quotas run out.KarpelesLab
Agent Cockpit is a local whiteboard and MCP server for AI research agents that traces every claim back to the run and code that produced it, with staleness checks and replay.chutingyean
This DSH plugin connects a ChatGPT or Codex subscription to DeepSeek Harness, adding subscription sign-in, Codex model selection, quota tracking, image generation, web search and optional phone-based remote control without an API key.WSL043
A Slack-native agent gateway that routes team threads to Claude Code, Codex, OpenCode, Docker, or HTTP agents, adding approvals, audit logs, shared memory, and channel-level policy.linxidnju