Financial analyst working across market data terminals

Why Leave a Successful Company?为什么离开一家成功的公司？

The AI image platform was working.AI 图像平台运转良好。

Liblib had built one of the most popular AI image generation platforms in China. The product was growing, the team was strong, and the market was expanding. So why leave?Liblib 已发展成为中国最受欢迎的 AI 图像生成平台之一。产品在增长，团队实力雄厚，市场也在扩大。那么，为什么要离开？

The infrastructure gap was obvious.基础设施的缺口十分明显。

Three observations drove the decision and exposed the need for a shared capability-routing layer.三个观察促成了这一决定，也揭示了共享能力路由层的必要性。

Because the next decade of AI isn't about generating images. It's about agents that act. And agents that act need infrastructure to find, route, and execute real-world capabilities at scale.因为 AI 的下一个十年不只是生成图像，而是能够采取行动的 Agent。而采取行动的 Agent 需要基础设施，才能大规模发现、路由并执行真实世界的能力。

Observation 1观察一

Agents were multiplying. Every major model lab was shipping agent features. Developers were building agent products. The ecosystem was forming fast.Agent 数量迅速增加。各大模型实验室都在推出 Agent 功能，开发者也在构建 Agent 产品，生态系统正在快速形成。

Observation 2观察二

Tools were fragmented. 10,000+ APIs existed for agents. But they were scattered, unindexed, and incompatible. No unified discovery or routing.工具高度碎片化。面向 Agent 的 API 超过 10,000 个，却分散各处、缺乏索引且互不兼容，没有统一的发现与路由机制。

Observation 3观察三

The window was open. Infrastructure layers get built once, early, and become defaults. The window to build the capability routing network for agents was open — but not forever.机会窗口已经打开。基础设施层往往在早期建立，随后成为默认选择。为 Agent 构建能力路由网络的窗口已经打开，但不会永远存在。

The Vision: Action Infrastructure for the Agent Era愿景：Agent 时代的行动基础设施

Kubernetes + Zapier + Homebrew — for AI tools面向 AI 工具的 Kubernetes + Zapier + Homebrew

🖥

Like Kubernetes像 Kubernetes 一样

Standardizes how agents discover and route to capabilities — the same way Kubernetes standardized how services discover and route to each other. One control plane for all tool access.标准化 Agent 发现能力并路由至能力的方式，就像 Kubernetes 标准化服务之间的发现和路由一样。通过一个控制平面管理所有工具访问。

🔌

Like Zapier像 Zapier 一样

Connects agents to thousands of real-world tools and services — without requiring custom integrations for each one. One connection, all capabilities.将 Agent 连接到数千种真实世界的工具和服务，无需逐一编写自定义集成。一次连接，访问全部能力。

📦

Like Homebrew像 Homebrew 一样

A trusted, community-verified registry of capabilities — so agents and developers can find and install exactly what they need. Verified, versioned, ready to call.提供可信且经社区验证的能力注册表，让 Agent 和开发者找到并安装所需能力。经过验证、支持版本管理，随时可调用。

But purpose-built for AI agents. Not adapted from something else.但它是专为 AI Agent 构建的，而非由其他系统改造而来。

What We're Building我们正在构建什么

The shared infrastructure layer for the agent ecosystem.面向 Agent 生态系统的共享基础设施层。

Capability Discovery能力发现

A semantic search engine for tools. Agents describe what they need in natural language. QVeris returns ranked matches with quality signals. No API names to memorize. No documentation to hunt down.面向工具的语义搜索引擎。Agent 用自然语言描述需求，QVeris 返回附带质量信号的排序结果。无需记忆 API 名称，也无需四处查找文档。

Intelligent Routing智能路由

When multiple providers offer the same capability, QVeris routes to the best match — based on task requirements, provider availability, cost, latency, and success rate. Not random. Not hardcoded. Optimized.当多个供应商提供同一种能力时，QVeris 根据任务要求、供应商可用性、成本、延迟和成功率，将请求路由到最匹配的选项。不是随机选择，也不是写死配置，而是经过优化。

Sandboxed Execution沙箱执行

Every tool call runs in an isolated sandbox with parameter validation, authentication handling, and consistent JSON output. One protocol for all capabilities. No per-provider parsing. No broken pipelines.每次工具调用都在隔离沙箱中运行，包含参数校验、身份验证处理和一致的 JSON 输出。所有能力使用同一种协议，无需逐家解析供应商结果，避免流程中断。

Full Audit Trail完整审计记录

Unique execution IDs, session-level tracing, and full call records for debugging, cost tracking, and compliance — built in from day one. Every call logged. Every result traceable.从第一天起就内置唯一执行 ID、会话级追踪和完整调用记录，用于调试、成本跟踪与合规。每次调用都有记录，每项结果都可追溯。

Why Now为什么是现在

2023

LLMs go mainstream大语言模型走向主流

GPT-4 launches. Developers start building agent prototypes.GPT-4 发布，开发者开始构建 Agent 原型。

2024

Agent frameworks emergeAgent 框架出现

LangChain, AutoGPT. Orchestration solved. Tool access fragmented.LangChain、AutoGPT 等框架出现。编排问题得到解决，工具访问仍然碎片化。

2025

Agents enter productionAgent 进入生产环境

79% of enterprises deploy agents. Fragmentation becomes a bottleneck.79% 的企业部署 Agent，碎片化成为瓶颈。

2026

Infrastructure layer builds基础设施层开始建立

The window to become the default routing network is open. QVeris is building it.成为默认路由网络的机会窗口已经打开，QVeris 正在构建这一网络。

Future未来

Every agent on shared infra所有 Agent 使用共享基础设施

Like DNS for the web. Every agent routes through a shared capability network.就像互联网中的 DNS，每个 Agent 都通过共享能力网络进行路由。

What We've Shipped我们已推出的产品

QVeris NetworkLIVE已上线

10,000+ verified capabilities across 15+ categories. The core routing infrastructure.覆盖 15 个以上类别、超过 10,000 项已验证能力，构成核心路由基础设施。

View Providers →查看供应商 →

QVeris CLILIVE已上线

Universal API gateway from the terminal. Discover, inspect, and call any tool in natural language.终端中的统一 API 网关。使用自然语言发现、检查并调用任意工具。

Install CLI →安装 CLI →

MCP ServerLIVE已上线

Tool gateway for IDE agents — Cursor, Claude Code, and all MCP-compatible environments.面向 IDE Agent 的工具网关，支持 Cursor、Claude Code 及所有兼容 MCP 的环境。

Set Up MCP →配置 MCP →

QVerisLabLIVE已上线

Production AI assistant with native access to 500+ data providers and 10,000+ APIs.面向生产环境的 AI 助手，原生接入 500 多家数据供应商及超过 10,000 个 API。

Try QVerisLab →试用 QVerisLab →

Frequently Asked Questions常见问题

Who founded QVeris AI?谁创立了 QVeris AI？

QVeris was founded by the former CTO of Liblib AI, one of China's leading AI image generation platforms. After observing the rapid growth of AI agents and the lack of shared tool infrastructure, the founder left a successful company to build the capability routing network that every AI agent will need.QVeris 由 Liblib AI 前 CTO 创立。Liblib AI 是中国领先的 AI 图像生成平台之一。在观察到 AI Agent 的快速增长以及共享工具基础设施的缺失后，创始人离开了一家成功的公司，转而构建每个 AI Agent 都将需要的能力路由网络。

What problem is QVeris solving?QVeris 正在解决什么问题？

AI agents can only use tools they were pre-configured with at build time. QVeris solves the tool discovery and routing problem — letting agents find and call any of 10,000+ verified capabilities at runtime, using natural language, the same way a search engine lets humans find web pages they've never visited.AI Agent 通常只能使用构建时预先配置的工具。QVeris 解决的是工具发现与路由问题：让 Agent 在运行时使用自然语言，从超过 10,000 项已验证能力中查找并调用所需能力，就像搜索引擎帮助人们找到从未访问过的网页一样。

Why is 2026 the right time to build AI agent infrastructure?为什么 2026 年是构建 AI Agent 基础设施的合适时机？

Enterprise AI agent deployment has crossed 79% adoption. Agents are moving from prototypes to production workflows. The tool fragmentation problem — scattered APIs, no unified discovery layer — is now the primary bottleneck at scale. Infrastructure layers get built early and become defaults. The window is open now.企业 AI Agent 部署采用率已超过 79%。Agent 正从原型走向生产工作流。API 分散、缺少统一发现层等工具碎片化问题，已成为规模化应用的主要瓶颈。基础设施层通常在早期建立，随后成为默认选择。机会窗口就在当下。

How is QVeris different from LangChain or other agent frameworks?QVeris 与 LangChain 等 Agent 框架有何不同？

LangChain and similar frameworks handle agent orchestration — how agents reason and chain steps together. QVeris operates at a different layer: capability discovery and routing. It works alongside any orchestration framework, not instead of it. Orchestration manages the reasoning flow. Routing manages the tool access.LangChain 及类似框架负责Agent 编排，即 Agent 如何推理并串联步骤。QVeris 工作在另一个层面：能力发现与路由。它与各种编排框架协同工作，而非取代它们。编排管理推理流程，路由管理工具访问。

What does "action infrastructure for the agent era" mean?“Agent 时代的行动基础设施”是什么意思？

It means the shared layer that lets AI agents discover, inspect, route to, and execute real-world capabilities — tools, APIs, live data, and external services — at production scale, with quality signals, audit trails, and consistent execution guarantees. The same way DNS, HTTP, and cloud infrastructure became the shared foundation of the web, capability routing becomes the shared foundation of the agent ecosystem.它指一个共享层，让 AI Agent 能够以生产级规模发现、检查、路由并执行真实世界的能力，包括工具、API、实时数据及外部服务，同时提供质量信号、审计记录和一致的执行保障。正如 DNS、HTTP 和云基础设施成为互联网的共享基础，能力路由也将成为 Agent 生态系统的共享基础。

Former Liblib CTO Builds Infrastructure for the Agent Era前 Liblib CTO 构建 Agent 时代基础设施