Tag
一篇关于在生产环境中部署 agentgateway 的实战指南，介绍这个 Apache-2.0 的 Rust 代理如何为 MCP 服务器、代理和模型提供统一入口，并逐步演示 JWT 鉴权、策略配置与协议版本演进对部署的影响。
作者主张开发者应自建并控制 LLM 代理（proxy），以获得真正的 token 用量、请求日志、缓存命中等可观测性数据，而不是依赖各大模型厂商提供的无用仪表盘。
用户发帖询问 Magpie 的使用体验，Magpie 是一款本地网关工具，可让 Claude Code、Codex 等任何 AI 代理跨厂商调用模型，并在多个订阅与 API 账户间智能路由和分配额度。
一个在 GitHub 上获得 11,502 stars 的免费前沿模型插件被发现其实是客户端而非模型提供方，其免费通道将用户提示词路由到 opencode 的 Zen 网关和 Kilo AI 的公开网关，且可能被上游记录日志用于改进服务。作者指出应用界面不会显示上游提供商名称，提醒用户在安装前了解提示词的实际去向。
开源工具 workbuddy2api-hub 将腾讯 WorkBuddy/CodeBuddy 的原生模型服务反向代理为标准 OpenAI 兼容接口（含 Anthropic Messages），支持多账号调度、设备指纹隔离与 Web 看板，可直接接入 ClaudeCode 等终端 Agent 工具。
开源项目 dot2api 通过 MCP 连接器将 OpenAI Dot 封装为兼容 OpenAI 与 Claude 的 API 端点，无需逆向网页接口或使用账号凭证，可用于自动化工作流与 Agent 委派任务。
Liquid Inference 是一款 LLM 路由器，让多家模型提供商针对每次 prompt 动态竞价，从而为用户提供最低推理成本，兼容 Claude Code、Cursor 等主流 agentic 编码工具，支持数百个开源与闭源模型。
Uber engineers detail the design of their MCP Gateway, a centralized microservice platform that unifies how hundreds of teams expose existing HTTP, gRPC, and TChannel services as MCP tools for AI agents — now hosting 800+ MCP servers and 5000+ tools internally.
Vercel AI Gateway的CEO @rauchg 表示，大多数token聚合器充斥着噪音，包括那些要求你把数据用于训练的token促销，以及声称支持ZDR（零数据保留）但信誉存疑的提供商；而Vercel AI Gateway凭借真实客户用量、40万+付费客户规模和零加价，成为理解全球AI token流向的最可信数据源。
FastRouter.ai launches on Product Hunt as a unified AI gateway and control plane that routes LLM requests across 200+ models via a single OpenAI-compatible API, optimizing for cost, latency, quality, and reliability with failover, observability, and governance.
A developer built a small gateway to simplify multi-model agent workflows by handling streaming, tool calls, and provider-specific integration challenges across models like Claude and GPT.
LiteLLM enables using over 100 AI models through a single connection, supporting major providers like OpenAI and Anthropic, with open-source and self-hosted options.
GoModel is an open-source AI gateway written in Go, providing a single OpenAI-compatible API for multiple providers with features like budgeting, caching, guardrails, and load balancing, serving as a self-hosted alternative to OpenRouter and LiteLLM.
FreeLLMAPI is a GitHub tool that aggregates free quotas from 34 AI service providers into a single OpenAI-compatible API, featuring intelligent routing, automatic failover, and quota management for developers.
The article provides a tutorial on obtaining free access to Claude Opus 5 by signing up on a third-party website, generating an API key, and configuring Claude's developer mode with that key and base URL.
The article discusses the need for tool gateways to secure AI agents' API access by limiting unpredictable behavior and seeks recommendations for products or libraries that provide such functionality.
Google Cloud API Gateway now offers model routing in Public Preview, providing a serverless ingress layer that accepts OpenAI-compatible requests and dynamically routes them to Gemini, Claude, or OpenAI models.
Google Cloud API Gateway announces public preview of model routing, letting developers access Gemini, Claude, and OSS models via a single endpoint using OpenAPI specs.
TokenShield, an open-source FastAPI gateway proxy, now uses per-turn sliding window hashes to prevent agent retry loops in CrewAI, reducing token waste and cost.
The article describes how building an intelligent caching gateway (Hawiyat Composer) saved significant AI API costs by eliminating repeated token waste through exact-match caching, semantic caching, model routing, and local routing.