Agent frameworks & infrastructure
LLM gateways & API proxies
4,018 agents · most starred first
LLM gateways & API proxies, page 17 of 41
A standalone, all-on-by-default token-optimization layer for LLM requests: minifies tools/system/messages (code-fence aware), dedups tool results, distills old turns, prunes to a budget, type-compresses output. Async proxy for Anthropic, OpenAI & Ollama + MCP server (stdio) + CLI. MIT; ships orjson/xxhash/tiktoken.
Minimal, provider-agnostic Python client for large language models, built on niquests.
DSH插件-模型余额显示功能
Bring third-party subscription accounts into DeepSeek Harness - 为 DeepSeek Harness 接入第三方订阅账户(当前支持 OpenAI / ChatGPT、xAI (Grok)、GitHub Copilot,全部通过浏览器授权登录)
Local task queue & scheduler that squeezes Claude's 5-hour usage windows dry — and now any Anthropic-compatible subscription (Kimi Code, GLM Coding Plan, MiniMax, MiMo, OpenCode Go, Ollama Cloud) via engine profiles, each with its own cooldown and ledger. Single Go binary, zero-quota orchestration, read-only web board.
High-performance local 3-in-1 gateway unlocking Google Gemini as standard OpenAI, Anthropic (Claude Code), and Gemini API endpoints for coding agents, IDEs, and tools.
Fast Rust proxy that turns GitHub Copilot into OpenAI- and Anthropic-compatible API endpoints. Works with Claude Code, Codex CLI, and any OpenAI/Anthropic SDK.
API gateway for OpenAI Chat Completion using Cloudflare workers
A Ruby wrapper for OpenAI API
Hot-switch custom Codex models, API endpoints, and keys without restarting Codex. Includes a follow-window Windows sidecar.
🐢⚡ The best free API key rotation + automatic model fallback for Hermes Agent. Many keys and models as one pool: every 429 / RESOURCE_EXHAUSTED read, each key rested exactly as long as needed, a late reply raced by your next model. Gemini, OpenAI, OpenRouter, any provider.
Run Claude Code through a local gateway with Opus 5, direct Codex, Kimi K3, Qwen 3.8 Max, and Codex-backed agents.
dsh plugin: DeepSeek Harness 的模型层 = raincode(模型池/缓存/重试) + /skills 浏览
OpenAI-compatible REST API and SSH gateway for CodeBuddy CLI, packaged as a Docker container.
One local gateway for every AI provider. Speaks both the Anthropic and OpenAI wire protocols — streaming, tool calling, vision, automatic failover, circuit breakers, and a live dashboard. Works with Claude Desktop, Claude Code, Cursor, Cline, and any SDK. Self-hosted, zero telemetry.
Lightweight, fast OpenAI-compatible AI router in Rust. Prefix routing, fallback combos, token saver, usage metering, embedded dashboard. One binary, no runtime, no external DB.
Local OpenAI-compatible proxy for your Claude, ChatGPT (Codex), and Grok subscriptions — routed by model name, OAuth login, no API keys.
OpenAI-compatible local proxy for Claude Code CLI, built for OpenClaw and other OpenAI-compatible clients using your Claude Pro/Max login.
🚅 A lightweight, native macOS menu bar proxy for Claude Code — protocol conversion, silent failover, and token stats across 42+ LLM providers. 100% local, privacy-first. 轻量级 macOS 菜单栏 LLM API 代理,支持 42+ 提供商智能路由、自动故障转移与协议转换。
Go OpenAI-compatible LLM gateway route chat/responses/messages across providers with multi-key failover, Anthropic↔OpenAI↔Gemini translation, admin console + telemetry.
Open source, ultra-lightweight, fast,self-hosted, production-ready and enterprise-level llm gateway for multi LLM provider access. OpenAI- and Anthropic-compatible proxy, encrypted master keys, scoped child API keys, and usage/cost analytics.
Factory.ai provider extension for Oh My Pi (omp) with Droid OAuth and Factory LLM gateway routing.
Run GPT-5.6 (on your ChatGPT plan) and Claude side by side in one Claude Code session — full prompt caching, zero injected context, no API keys.
A high-performance LLM API gateway written in Rust. Single binary, zero runtime dependencies, protocol-aware routing across multiple LLM providers.
🪙 40+ Free LLM APIs → One OpenAI-compatible endpoint. CLI, TUI, Proxy, LangChain wrapper. Auto-rotate keys, handle rate limits, track usage.
Claude Code 多模型代理 — 一键切换 DeepSeek、Qwen、GLM、Moonshot 等国产大模型,协议自动转换,零配置即用
Local AI router. One OpenAI-compatible endpoint for every provider you already have, and one key for opencode, claude code, Cursor and any coding agent. Keys stay on your machine.
Apizio — a cheap OpenRouter alternative and OpenAI-compatible AI API gateway. 60 models (GPT, Gemini, Qwen, GLM, DeepSeek, Kimi) on one key and one balance. $5 free credits.
如果你开通了腾讯 WorkBuddy 或 CodeBuddy的会员/套餐积分,拥有大量的免费或高配模型额度(GLM-5.2/5.3、DeepSeek-V4、Kimi、混元等),但在日常编码中,你更习惯使用 Codex CLI、Claude Code、Claude Desktop 桌面端,或者 Cherry Studio / Chatbox 等外部工具
Free OpenAI-compatible API — drop-in ChatGPT API replacement. Use OpenAI SDK with Groq, Gemini, Mistral, Cohere, OpenRouter. No OpenAI bill, no credit card. GPT-4 / GPT-3 alternatives, chat completions endpoint.
A lightweight FastAPI proxy that translates Anthropic API requests to NVIDIA NIM's OpenAI-compatible endpoint.
Local gateway that routes Claude Code, Codex & any OpenAI-compatible coding tool across 26 free-tier AI providers — with auto-failover, quota-aware routing, and request auto-fit.
Automatic model and effort routing for Claude Code: Haiku, Sonnet, Opus per prompt, zero token classification
Self-hosted gateway for LLMs and decision models. Claude Code, Codex, OpenCode and any OpenAI/Anthropic SDK app reach any provider (OpenRouter, Groq, Ollama…) with context pruning; Jev and open-rlcd System One decisions get audit and calibration. Live dashboard, single Go binary.
A circuit breaker for your LLM API bill — hard budget caps + per-feature cost attribution for OpenAI & Anthropic.
DeepSeek Harness plugin: Sign in with Claude Pro/Max (no API key). Caps Anthropic images at 2000px and slims /compact for 1M context.
Claude Max account usage and auto-switching for bb — proactive on utilization, reactive on real rate-limit failures, with thread auto-continue.
Claude Desktop onboarding for third-party inference gateways in one command: model discovery, profiles, backups, macOS / Windows
OAuth token manager, subscription quota dashboard and native wire-protocol gateway. The LiteLLM plugin part now lives in litellm-mysubs.
模型路由器:在 Codex Desktop 和 Claude Desktop 使用 Claude、第三方 API 与自定义模型。Local LLM proxy for Codex & Claude Desktop: OpenAI Responses / Chat Completions & Anthropic Messages, Claude Code subscription routing. macOS & Windows.
Local load-balancing reverse proxy for multiple Anthropic Max OAuth accounts
A small self-hosted API in front of the Claude CLI. It answers at /v1/messages in the shape of the Anthropic API, so an SDK client works with only a base URL change.
⚡ OpenAI-compatible proxy with smart routing across 146 free AI models. Zero failures, auto-fallback, admin dashboard, Docker ready. Powered by Puter.js.
⚡ Route requests to the most cost-effective model with ClawRouter—one wallet, over 30 models, no API keys needed.
反代本地:把 Converge/enter.pro 反代成本地 OpenAI / Anthropic 兼容服务,多账号池自动切号 · prompt cache · 工具调用 · 单页控制台。 | Local OpenAI/Anthropic-compatible proxy for Converge/enter.pro with multi-account pooling, prompt caching, tool calling and a web console.
Use a Claude Code session as a local Anthropic-compatible API
Claudy lets you switch between Anthropic, Z.AI, OpenRouter, Ollama, and custom endpoints with a single command — keeping credentials, config modes, and Claude frameworks cleanly isolated per profile. Multi-provider · Config isolation · Channel bridge · Local agent bridge · Usage analytics
Free agentic AI coding assistant for VS Code. Multiplexes many free-tier LLM providers into one self-healing surface — auto-routes, auto-fails-over, learns your codebase. Zero setup with keyless providers; bring your own keys anytime.
把 Notion AI 网页登录态反代成本地 OpenAI / Anthropic / Responses 兼容接口:30+ 模型、Claude Code / Codex / cc-switch 接入、单文件 exe、自动改代码 agent。
Local reverse proxy that fixes Cline Desktop 403s and stops free-tier models from being billed. OpenAI + Anthropic compatible.
Merge multiple Claude API gateways into one local endpoint. Automatic failover, sticky prompt-cache routing, model aliasing. Zero dependencies.
use glm/minimax/openai/claude api in your deepseek harness
Per-step reasoning effort for coding agents, picked by a classifier, without breaking the prompt cache. OpenCode plugin, Pi extension, and a local proxy for Codex CLI.
Serverless AI gateway on Cloudflare Workers for resilient access through rate limits and failures
Self-hosted gateway: one OpenAI- and Anthropic-compatible endpoint in front of many upstream accounts, with round-robin failover, a console, a CLI and an admin API.
One local endpoint for all your AI subscriptions: agents send a pool name, the gateway holds the credentials and rotates accounts as quotas run out.
A universal gateway for AI models.
Remote-first, multi-user AI gateway with BYOK failover, shared memory, and a local MCP agent that runs tools on your own machine.
Cut up to ~89% of cloud LLM tokens. Self-evolving local-first smart router (R0-R3, ~100% routing accuracy after evolution) - Stanford & Together AI baseline: 60-80% cost/energy savings, 80%-accurate router already captures ~80% of oracle gains. OpenAI/Anthropic gateway + MCP. Rust/Python/Node/C ABI.
occ gateway proxy for Claude Code provider routing
Intelligent model routing for OpenClaw with quota prediction, task classification, and automatic optimization
Lightweight implementation of the OpenAI open API on top of local models
🦙 Llama - Welcome to the Llama-API repository! This repository is dedicated to testing the llamaapi Python library to access LlamaAPI. Whether you're a developer, researcher, or just a tech enthusiast, you'll find valuable resources and examples here.
Local-first LLM gateway for Ollama clusters and agentic clients, with OpenAI-compatible streaming, model routing, Prometheus metrics, and an embedded admin UI.
Access AI services without any API keys!
Stop overpaying to run your agents. Kalibr routes every request to lower-cost model and tool paths without degrading performance. TypeScript SDK.
🔑 批量检测 AI API Key 有效性的 Web 工具,支持 OpenAI、Claude、Gemini 等 12+ 平台
Open-source desktop tool for managing LLM API keys and syncing them into local AI agents and clients.
API key rotation and model fallback proxy — battle-tested on OpenRouter, supports Anthropic, OpenAI, and Gemini
Stop waiting for Opus on every grep — routes 80% of tasks to faster models. Zero dependencies.
Lightweight AI gateway for multi-provider routing, fallback, budget control and usage logging. Zero runtime dependencies.
A lightweight CLI tool to seamlessly switch between AI providers for Claude Code and Codex
OpenAI-compatible inference proxy backed by Claude Code
SwiftAI, Swift, AI, Grok, xAI, LLM, Anthropic, Claude, OpenAI, ChatGPT
Plug-and-play LLM connector via YAML config — 21 providers, async, tool calling, structured outputs, embeddings, middleware. Zero bloat.
Run Claude Code using DeepSeek's API to cut costs by 50x while keeping the same workflow and token experience.
The official Python library for the Edgee AI Gateway
Switch Claude Code between LLM providers — no proxy, no config editing
Kairo: Securely orchestrate multiple AI providers via Anthropic. Hardened with X25519 (age) encryption, integrated audit trails, and a terminal-native interface.
The official Node.js / Typescript library for the Edgee AI Gateway
Budgeted LLM runs with hard caps + typed outcomes + async job mode. Python/CLI/HTTP. OpenAI-compatible.
🤖 Enhance analysis with a customizable AI integration tool that unifies multiple AI services through a standard API for streamlined outputs.
Runnable, framework-focused recipes for the Ferro Labs AI Gateway: build LangChain & LangGraph agents that route across OpenAI, Anthropic & Gemini through one endpoint — docker compose up and go.
Cut your Claude bill: local proxy that auto-routes every request to Haiku/Sonnet/Opus by task difficulty. Zero code changes, works with Claude Code.
MCP server that auto-classifies prompt complexity and routes to the right Claude thinking mode
A transparent, 100%-local semantic cache for LLM APIs — drop-in proxy, one line to integrate, written in Rust
Transparent Rust proxy that lets upstream models without native tool calling participate in OpenAI/Anthropic tool protocols.
Rotate across multiple Anthropic Claude Code (Max/Pro) subscriptions — picks the account with most 5h/weekly quota headroom, writes a full-scope ~/.claude/.credentials.json, and execs the real claude. OAuth-native, alias-drop-in, no shadow dirs; Remote Control and session-scope features work out of the box.
Sample Amazon Bedrock proxy demonstrating how developers can use any AI framework (LangChain, LangGraph, Strands Agents, CrewAI) through a centrally governed API Gateway with Cognito JWT authentication, request tracking, and response streaming — all transparent to the framework via boto3's event system.
Blazing Fast Semantic Cache & API Optimizer for LLMs (OpenAI, Anthropic, Gemini)
gradio region:us
docker region:us
docker region:us
Use your Claude Max subscription as a local OpenAI-compatible API. Multi-account load balancing, streaming, rotating logs, zero per-token cost.
让 Claude Code 在科大讯飞 Anthropic 端点上正常工作的轻量透明代理(thinking 思维链 + WebSearch 网络搜索适配)
Provider-neutral LLM router and model orchestration engine with auto-learning and budget zones
SuperAI — Intelligent multi-model AI orchestration platform with self-learning and automated backups
OpenAI-compatible provider portability: convert chat conversations between OpenAI, Anthropic and Gemini. Tool calls, multimodal parts, zero deps.
A highly compatible proxy bridge with translation for API requests to multiple backend providers (OpenAI-Oauth, RunPod, Ollama,,etc.). Configure endpoints through a web admin UI and map virtual model names to actual backend models. Create API keys to share and monitor usage and costs by tools / teams. Opencode/Claude to queue servers too.
Yet another proxy for LLMs