Harnesses & Runtime Infra tools for Claude / Anthropic

Lobe Chat - an open-source, modern-design AI chat framework. Supports Multi AI Providers( OpenAI / Claude 3 / Gemini / Ollama / DeepSeek / Qwen), Knowledge Base (file upload / knowledge management / RAG ), Multi-Modals (Plugins/Artifacts) and Thinking. One-click FREE deployment of your private ChatGPT/ Claude / DeepSeek application

(51.2k ⭐) - API service that wraps Antigravity, Codex, Claude Code, and Grok Build behind OpenAI-, Gemini-, Claude-, and Codex-compatible endpoints

Self-hosted multi-model chat. All major providers

(23.7k ⭐) - Routes Claude Code, Codex, Cursor, Cline, and other coding agents to dozens of model providers with automatic fallback

FastGPT is a knowledge-based platform built on the LLMs, offers a comprehensive suite of out-of-the-box capabilities such as data processing, RAG retrieval, and visual AI workflow orchestration, letting you easily develop and deploy complex question-answering systems without the need for extensive setup or configuration

SpaceXAI's open-source terminal coding agent harness: a Rust-based fullscreen TUI with an extensible tool runtime, MCP/skills/hooks support, and headless/embedded modes via ACP. A useful first-party counterpoint to Claude Code and Codex CLI for studying how a new model provider structures the loop, checkpoints workspace state, and exposes the same harness through interactive and scripted surfaces

(1.5k ⭐) - Menu bar and terminal app for choosing model providers for Claude Code, Codex, and other coding agents

(8.5k ⭐) - A local provider proxy and dashboard that routes Codex CLI, the Codex app, SDKs, and Claude Code to multiple LLM providers

(11.6k ⭐) - The power of Claude Code + Gemini / OpenAI / Grok / OpenRouter / Ollama / Custom Model working as one

All-in-one agent sandbox combining browser, shell, filesystem, MCP servers, and VSCode Server in a single Docker container. Native MCP support exposes sandbox capabilities to LLMs via the standard protocol, and files downloaded in the browser are instantly accessible in terminal and VSCode. Optimized startup (4–8s depending on config) with Claude Skills mounting support. The fastest path to a full

Zero-config workflow framework. Chain prompts, branch on results, auto-retry failures

Open-source Go proxy for the Anthropic Messages, OpenAI Chat Completions and Gemini APIs that picks a model per request with an on-box embedder, for Claude Code, Codex, opencode and pi. Self-hostable with BYOK provider keys and OTLP traces. Apache-2.0

(5.2k ⭐) - Self-hosted MCP server that gives ChatGPT, Claude, and other clients access to approved local projects for file edits, code search, and shell commands

(4.0k ⭐) - Turns GitHub Copilot into an OpenAI/Anthropic API compatible server, usable with Claude Code

OpenAI-compatible LLM inference server for Apple Silicon using MLX. 2-4x faster than Ollama with tool calling and prompt caching

Context runtime and MCP server that reduces AI coding agent token costs via session caching, AST-aware compression, and shell output patterns. Website

(3.6k ⭐) - Run Claude Code on OpenAI models

Spec-driven development workflow. Write specs, Codex implements, auto-validates against spec

Qwen's realtime voice runtime that keeps agents talking, working, and present across multiple backend agents (Claude Code, Codex, Qwen Code, Kimi Code). It demonstrates how to wrap terminal-native coding agents in a persistent, interruptible voice shell without losing task continuity — a concrete reference for ambient agent interfaces

Context and token-cost optimizer for AI coding agents. Finds wasted "ghost tokens," survives compaction, and reduces context-quality decay, with native plugins for Claude Code, Codex, OpenCode, and OpenClaw

(2.7k ⭐) - A Claude Code to OpenAI API Proxy

Lightweight local relay that converts Codex Desktop's Responses API to multiple protocol interfaces (OpenAI, Anthropic, Gemini)

(1.7k ⭐) - Use Kimi's latest model (kimi-k2-0711-preview) to drive Claude Code

by dyoshikawa - A Node.js CLI tool that automatically generates configs (rules, ignore files, MCP servers, commands, and subagents) for various AI coding agents. Rulesync can convert configs between Claude Code and other AI agents in both directions

(1.4k ⭐) - An HTTP API for Claude Code, Goose, Aider, and Codex

Lets your agent install self-hosted apps on your own hardware and set them up (admin accounts, API keys, apps connected to each other), asking before each step

Run Claude Code as a one-shot MCP server -- an agent in your agent. Permissions bypassed automatically. By Peter Steinberger. 1,100+ stars

BaseAI — The Web AI Framework. The easiest way to build serverless autonomous AI agents with memory. Start building local-first, agentic …

Coding agent session manager supporting Claude Code, Gemini CLI, Codex, Cursor, Copilot, Cline, OpenCode, Kimi CLI. Smart auto-approval via Haiku, devcontainer support. 940+ stars

Unified TUI and CLI to index and search local coding agent session history across 11+ providers including Codex

(992 ⭐) - A load balancer proxy for multiple Claude OAuth accounts with automatic failover, request tracking, and web dashboard

Persistent context and a code knowledge graph for coding agents, with a Codex CLI launcher

Intelligent API gateway for Claude Code and Codex CLI. Rotate endpoints, monitor usage, seamlessly integrate OpenAI, Gemini, and others

by es617 - An outstanding, creative library that converts Claude Code session transcripts into self-contained, embeddable HTML replays - interactive playback with speed control, a local editor, collapsible tool-call and thinking blocks, redaction for private information - the output is fantastic. Also supports other providers

Open-source, OpenAI-compatible LLM gateway you run yourself. One endpoint for 40+ providers, with virtual keys, budgets, and usage tracking

DeepSeek-V4 terminal coding agent harness that treats prefix-cache economics as a first-class design constraint: byte-stable system prompts, cache-reusing forks for reflection and memory, and a self-verifying cross-session memory layer keep real SWE-bench-style tasks at ~95.8% cache hit and ~30× cheaper than Claude Opus. A concrete demonstration that harness-level cache engineering can substitute

Scaffold factory that turns any repo into a branded agent harness with its own npx CLI, MCP server, scoped memory, governance policy, and Darwin Mode self-evolution. The clearest open-source embodiment of "the model is replaceable, the harness is the product" — it generates the owned scaffolding rather than prescribing a fixed framework

(614 ⭐) - An agentic proxy that enhances any AI agent workflow with instant history compaction and context optimization tools

by disler - Fuses 2–5 frontier models instead of racing them — one architect, one primary builder, and up to three secondary builders — giving N-way opinions, debate, fusion, coordinated implementation, direct one-agent routing, and gate-first validation. Ships as a composable Pi extension with pre-built model stacks and install/prime commands, and a video walkthrough of the whole approach

(480 ⭐) - Proxy that captures and visualizes in-flight Claude Code requests and conversations

Linter for AI agent configurations. Validates GEMINI.md, SKILL.md, hooks, MCP, and more with 156 rules, auto-fix, and LSP server for real-time editor diagnostics

One endpoint for every LLM vendor. Any SDK, any model, any tool. One Rust binary

(415 ⭐) - A proxy server that converts Anthropic API requests to OpenAI format and sends them to OpenRouter, used to use Claude Code with OpenRouter

(406 ⭐) - Intelligent integration between Claude Code and Google Gemini for large-scale code analysis

(390 ⭐) - A basic proxy to use Kimi K2 on Claude Code through Groq

(384 ⭐) - A Simple Proxy enabling Claude Code to work with OpenRouter

(348 ⭐) - A Python program allowing the use of Claude Code with Google's Gemini models

by akin-ozer - Immensely detailed set of skills for DevOps Engineers (or anyone who has to deploy code, really). Works with validations, generators, shell scripts and CLI tools to create high quality IaC code for about any platform you've ever struggled painfully to work with. Worth downloading even just as a source of documentation

LLM plugin for interacting with the Claude 3 family of models

An innovative open-source Code Interpreter with (GPT,Gemini,Claude,LLaMa) models

One API for 25+ LLMs, OpenAI, Anthropic, Bedrock, Azure. Caching, guardrails & cost controls. Go-native LiteLLM & Kong AI Gateway alternative

by Prassanna Ravishankar - A live mesh network that connects Claude Code and other providers so they can talk to each other in real time. Built for emergent coordination. Agents discover each other and share context across project boundaries. Uses tmux sessions as the coordination layer with logical "circles" for peer isolation. Includes a web dashboard and Telegram bot

(253 ⭐) - Seamlessly forward Claude Code requests to any OpenAI-compatible API service with smart model mapping, streaming support, deployed on Cloudflare Worker

(245 ⭐) - Connect Claude Code with Google's Gemini AI for powerful AI collaboration

(232 ⭐) - A lightweight integration that brings Google's Gemini AI capabilities to Claude Code through MCP (Model Context Protocol)

Simplify interactions with Large Language Models

Comprehensive usage analytics platform that tracks and optimizes your Gemini CLI workflows, helping teams understand and improve their AI-assisted development patterns

by Mykola Maksymenko - Local, multi-provider agent communication tool. Adds hooks and a skill to exchange questions and reviews, and saves decisions, verification results and handoffs for later sessions. Each session keeps its own client, model, conversation and permissions

Delegates coding grunt-work to other harnesses and models from the terminal while your main session stays the orchestrator

(178 ⭐) - An MCP server for Claude Code conversation history

by Erik Nielsen - A native-Swift macOS menu-bar app (10 MB, no Electron) that covers and input-locks your screen with one hotkey while your agents keep running, then makes the locked screen glow when Claude Code pauses for permission or finishes. Polished and well-engineered — CI with 50 tests, signed/notarized builds, an honest "visual privacy tool, not a security boundary" disclosure, no analyti

A full stack app for interruptible, low-latency and near-human quality AI phone calls built from stitching LLMs, speech understanding too…

Drop-in governance layer for AI coding tools. Backlog-first delivery, cross-session memory, decision tracking, QA gates, and autonomous delivery daemon. Works with Claude Code, Cursor, Codex CLI, Gemini CLI, Windsurf. Markdown + YAML + bash, zero dependencies

Agent harness with Slack, GitHub, and Linear integrations. Useful reference for how real-world tool wiring works inside a harness

Saves Gemini CLI logins and API keys as named profiles and switches the live ~/.gemini state in one command, alongside Claude Code, Codex CLI and Antigravity CLI. Docs: aiswitcher.dev

A simple NPM interface for seamlessly interacting with 36 Large Language Model (LLM) providers, including OpenAI, Anthropic, Google Gemin…

by S.F. Kislev - A tiny bridge that lets Claude Code drive desktop software — Photoshop, Premiere, Blender, Unity, InDesign, Office, 13 apps total — by writing one-time scripts against each app's own automation runtime (COM / AppleScript / CEP), instead of MCP servers or fragile screenshot-based computer use. A genuinely novel approach that unlocks the apps' full scripting surfaces (InDesign alone

(90 ⭐) - Use Claude Agent SDK and Claude Code with other providers/models

Windows app that reads, searches and backs up your Gemini CLI chats across accounts, and continues any chat in Gemini CLI through a handoff file that carries the whole conversation

Lightweight continual harness optimizer (April 2026) built on the Claude Agent SDK. Runs an outer loop that reads task traces, rewrites harness configs, and re-evaluates — achieving 67% → 87% on tau-bench with no labeled training data. Demonstrates that even small, focused meta-harness loops can yield large reliability gains when harness configs are treated as learnable parameters

AI Agent Orchestration Kanban Board. Route tasks to Claude Code, Codex CLI, and Gemini CLI with role-based auto-assignment

by openweb-org - An agent-native skill that accesses 90+ websites by calling their underlying APIs directly (typed JSON in, JSON out) instead of screenshotting and parsing the DOM, with auth auto-resolved locally from your existing browser session. Security is first-class — every operation is tagged read/write/delete/transact behind permission tiers, SSRF protection runs on each request, and there

(68 ⭐) - Claude Code with any LLM provider (OpenRouter, Gemini, Kimi K2)

Optimize agentic capabilities with agents. Reflective optimizer for AI agent prompts, tool code, MCP tool surfaces, and skill packages; learning from evaluation traces, val-only significance gate with a sealed test split scored once. Adapters for τ²-bench, SWE-bench, SkillsBench. Zero runtime deps, Apache-2.0

Sync Skills, MCP servers, Agents & Commands across Claude Code, Cursor, OpenCode, Codex, and Gemini CLI with automatic format conversion (JSON ↔ TOML ↔ JSONC)

March 2026 Claude Code plugin that autonomously evolves LLM agent harnesses using multi-agent proposers in isolated git worktrees, LangSmith-backed evaluation, and regression guards. Iterates on prompts, routing, retrieval, and orchestration code based on full-trace counterfactual diagnosis. The most practical published implementation of the Meta-Harness outer-loop optimization paradigm

Self-hosted web app that turns one set of AI coding rules into the file each tool reads: Gemini CLI's GEMINI.md, AGENTS.md, CLAUDE.md, Cursor rules and Copilot instructions, 30 rule-file formats in all. Blueprints carry variables so a team shares one set across projects, and the lynxp CLI writes the files into each repository. AGPL-3.0

by congmnguyen - A focused collection of documented scripts that fix the most painful Claude Code papercuts on WSL2 + Windows Terminal — clipboard screenshot paste via a Go daemon, Windows notifications on Stop/PermissionRequest hooks, LSP wiring, a usage-aware status line, and voice-mode audio routing. Each fix is written up with root cause and exact config (including non-obvious traps like Sessi

A framework to connect a flow of ML models by applying graph theory

Tool for managing Claude Code + Codex API + Gemini CLI + OpenCode service provider configurations

Rust rewrite of CLIProxyAPI with a built-in dashboard. Serves the Gemini API on a local endpoint (GOOGLE_GEMINI_BASE_URL), so Gemini CLI can run on Gemini API keys or on signed-in Claude and ChatGPT accounts

Automatically saves coding agents' conversations in Git Notes. Works with Codex, Claude Code, Gemini CLI, OpenCode

Local dashboard that starts Gemini CLI sessions and follows them through the telemetry file Gemini writes, on the same screens as Claude Code, Codex and OpenCode: live activity, token cost per repository, searchable history

AI agent skill that analyzes git history to reveal codebase hotspots, bug magnets, bus factor risks, and development momentum before reading any code. Works with Gemini CLI, Claude Code, Cursor, and 20+ other coding agents

by Amin Dhouib (Devino Solutions) - A zero-dependency, cross-platform notifier that fires a desktop toast and a free phone push (via ntfy) the moment Claude Code (or Codex/Cursor/Gemini) finishes a task or needs input, wired up by a one-command setup. The standout among notifiers is its testing rigor — CI drives the *real* agent CLIs end to end and asserts real ntfy round-trips and real OS toast d

Free design CLI and skill pack for Gemini CLI, Claude Code and Codex. Describe a UI in plain words and get web, Figma and SwiftUI output checked by deterministic taste gates; installs with npx @easestart/design-os init. Source-available (PolyForm Shield)

Local guardrail with a BeforeTool hook for Gemini CLI (laya-guard install gemini-cli) that judges each shell command before it runs, using regex rules, protected paths, and a small on-device model. Blocked commands are denied; ones marked for review pass with a note. Also hooks into Claude Code, Codex, Cursor, and other agents. MIT

Rust inference server for Apple Silicon: MLX models behind OpenAI and Anthropic APIs, routing to remote providers, desktop dashboard

Tmux manager for running, monitoring, saving, and restoring Codex and other agent CLI sessions

(32 ⭐) - Use Claude Code as an LLM provider with your subscription flat fee instead of pay-per-token API keys

Human-in-the-loop workflow. Codex pauses at decision points and asks for your input

Routes LLM, image, speech-to-text and embedding requests across a fleet of machines you own. Auto-discovers nodes over mDNS and scores each on memory fit, thermal state, queue depth and already-loaded models. OpenAI-, Ollama- and Anthropic-compatible

OpenAI-compatible /chat/completions, /embeddings & /models endpoints

Self-hosted mesh router for local LLM backends — OpenAI /v1 + Anthropic Messages on :11400, mDNS LAN peers. Complements cloud gateways with local LAN mesh for Cursor, Claude Code, Codex, and Honcho

by Endeavor Yen - A zero-dependency Claude Code skill (68 commands) that connects to your *real* Chrome — logged-in tabs, cookies, live page state — to give the agent a perception layer: layout, visible styles, per-action "what changed" evidence, CSS-cascade-to-source tracing, and session replay/Playwright export. Notably rigorous about its own claims, with a dogfood benchmark gate that blocks per

Local AI gateway for routing clients across providers, regions, accounts, and local models

Dashboard and macOS menu bar for tracking AI subscription quotas in one place (Claude Code, Codex, Gemini CLI, GitHub Copilot, and Cursor). Supports multi-accounts, notifications, and session-timer start optimization

TypeScript CLI that syncs agents, MCP configs, skills, commands, and workflow assets across Gemini CLI, Claude Code, Codex, OpenCode/OpenClaw, and GitHub Copilot workflows

Evidence-driven development pipeline that uses Gemini CLI as one of three independent reviewers in a multi-model code audit panel (alongside Claude and Codex)

Local-only TUI + web dashboard that tails Gemini CLI sessions alongside Claude Code, Codex, Cursor, Hermes, and OpenClaw on one unified timeline. Parses tokens, tools, and per-turn cost from each Gemini CLI session (gemini-2.5-pro / flash rates), plus context compaction visualizer, MAD z-score anomaly detection, MCP server mode, and OpenTelemetry exporter. No cloud, no telemetry. macOS + Linux. MI