# 融效通 EachRun

> Every AI run, accounted for. 让每一次协同，都融会贯通。

EachRun is an enterprise AI execution and governance platform for managing project workspaces, approved models and agents, access, quotas, execution records and handoffs.

中文简介：融效通 EachRun 是企业 AI 执行与治理平台，用于统一管理项目工作区、模型与 Agent、访问权限、AI 额度、执行记录和项目交接。

## Why the name

“融效通” names what happens when work from many people and many agents converges: 融 for convergence, 效 for the speed that follows it, 通 for delivery that is not blocked. “EachRun” applies the same requirement to every model call, agent run and business task.

中文品牌说明：“融效通”指多人、多 Agent 发起的执行汇到一处（融），汇聚之后更快出结果（效），全程不被阻塞地交付（通）；EachRun 表示每次模型调用、Agent 运行和业务任务都按这个要求管理。

## Who uses it

Three kinds of user run on the same platform. **Companies** use it to put AI into specific jobs: workflows and role-split AI teams give staff a path to follow, proven methods become reusable templates, and usage and cost land on a person, a project and a model. **Government and public institutions** use it to settle the boundary first: access control decides who may use which models, data and tools; requests, approvals, login logs and execution records leave a trail; private or hybrid deployment covers data-residency requirements. **Individuals** use it as one place that holds a project, its files and its history, where a workspace is a real machine with its own vCPU, memory and disk rather than a chat box.

中文：三类用户用的是同一套平台。**企业**用它把 AI 落到具体岗位上——工作流和按角色分工的 AI 员工团队让员工照着走就能开工，做通一次的方法沉淀成模板，用量和成本落到人、项目和模型上。**政府与事业单位**用它先管住边界——访问控制决定谁能用哪些模型、数据和工具，申请审核、登录日志和执行记录都留痕，数据不出域的要求用私有化或混合部署满足。**个人**用它作为一个装得下项目、文件和历史的入口——工作区是一台有自己 vCPU、内存和磁盘的真机器，不只是一个对话框。

## What EachRun solves

The core proposition is that a company can open AI up to its staff and still run the place, which takes three things at once: one entry point (employees do not build their own setup), one set of permissions (project roles decide which models, agents and data are available) and one ledger (every run is costed and traced). Workflows, role-split AI teams, shared templates, real workspace machines and private deployment sit on top of that core, not in place of it.

核心主张：把 AI 放开给员工用，公司还管得住。要三件事同时成立——一个入口（员工不用自己搭环境）、一套权限（项目角色决定可用的模型、Agent 和数据）、一笔账（每次执行都算得清、查得到）。工作流、按角色分工的 AI 团队、共享模板、真机工作区和私有化部署属于核心之上的增值层。

Personal AI tools help an individual complete a task. We add shared project workspaces, approved capabilities, access rules, quotas and execution records for team use.

In the current product, AI activity can be associated with a user, project, workspace, runtime and model. Administrators can review status and token usage; supported agents can ask for approval before selected tool or command operations; shared project resources can be assigned viewer, editor or administrator roles with expiry and revocation.

我们管理 AI 进入团队项目后的执行过程。现有产品可以把用户、项目、工作区、运行环境、模型、状态和 Token 用量放到同一条记录中；受支持的 Agent 可在部分工具或命令执行前申请确认；共享资源可设置角色、有效期并撤销。

豆包、Kimi、WorkBuddy 等产品适合员工直接完成聊天、研究、文档或办公任务。人数少、任务不接触内部系统时，直接选顺手的工具即可。多个团队共用模型与 Agent，或 AI 开始访问公司数据和业务系统时，我们管理经过接入与授权的能力、工作区、权限、额度和执行记录。具体产品能否接入取决于其接口、许可和部署条件。

## Product structure

Employees work through seven entry points:

- Workbench: projects, resource requests, open tasks and a guided first-run flow.
- Workspace: a cloud environment per project holding files, context, tools and execution records, which colleagues can join.
- Runtime: the vCPU, memory and disk allocated to a workspace, with live usage, port-level publishing of services inside it, and the option to attach existing servers.
- Discover: the models, agents and tools the company has connected and this person may use.
- Share center: prompts, task templates and processes published for reuse by other projects.
- AI quota: quota plans and user limits, with usage aggregated by user, workspace, project, runtime and model.
- Admin console: workspace resources, organization directory, access control, model providers, approval requests, scheduled jobs and login logs.

中文产品结构：员工侧有工作台、工作区、运行环境、能力市场、共享中心和 AI 额度，管理侧提供组织与角色、工作区、模型渠道、额度、共享资源和 AI 对话记录等管理能力。

## How employees get work done

Employees do not have to know how to prompt. A workflow breaks a job into steps, each step carrying the skills it needs; starting one produces a visual workflow whose output is filed back to the project. The capability market also holds AI teams packaged by function — academic research, design, engineering, finance, healthcare, marketing — where a team is split by role and carries anywhere from a few to dozens of skills. Teams and skills are maintained and distributed by the company, and installed into a workspace.

中文：员工不需要先学会怎么跟 AI 说话。工作流把一件事拆成几步，每一步挂着这一步要用的技能，按流程开工会生成可视化工作流，产出物自动挂回项目。能力市场里还有按职能打包的 AI 员工团队（学术研究、设计、工程研发、财务、医疗健康、市场营销），每支团队按角色分工，装进工作区即可使用；团队和技能由公司统一维护和分发。

## Current scope

Available product functions include organization members and roles, login audit, workspace application and approval, resource and lifecycle management, model providers and channels, connection testing, quota plans and user quotas, project sharing, and AI conversation records with status and token usage.

Some controls depend on the selected runtime and deployment configuration. Tool or command approval is available only for supported agents. Connections to CRM, ERP, OA, ticketing, internal APIs, repositories and knowledge bases are confirmed during a pilot; they are not presented as universal, ready-made integrations.

当前产品包含组织成员与角色、登录审计、工作区申请审批与资源管理、模型供应商和渠道、连接测试、额度计划和用户额度、项目共享，以及带运行状态和 Token 用量的 AI 对话记录。工具或命令确认取决于所选 Agent；业务系统、内部 API、代码仓库和知识库的接入范围需要在试点中逐项确认。

The homepage includes product screens showing AI conversation records, Agent access settings, sharing roles, governance statistics, usage attribution and run outcomes. The screens retain the product structure and states, while people, projects, workspaces, Agents, models and operating figures are replaced with demo data.

官网展示 AI 对话记录、Agent 授权、共享权限、治理统计、用量归属和运行结果等产品界面。页面结构和功能状态来自产品，人员、项目、工作区、Agent、模型及经营数字均替换为演示数据。

## Deployment

Available as SaaS service, enterprise team edition, enterprise governance edition, private deployment and hybrid deployment. Business system integration scope is confirmed during the pilot. Where data may not leave the organization, private or hybrid deployment defines the boundary.

部署方式包括 SaaS 云服务、企业团队版、企业治理版、私有化部署和混合部署，业务系统接入范围在试点阶段确认。

## Commercial pilot

A pilot normally starts with one engineering or IT team and one or two recurring, reviewable tasks for four to six weeks. The current time, rework, cost and results are recorded before the pilot and compared afterwards. The comparison supports a clear decision to stop, adjust or expand.

商业试点通常从一个研发或信息化团队的一到两个高频、可验收任务开始，运行四到六周。开始前记录耗时、返工、成本和结果，结束后对照基线决定停止、调整或扩大。

Pricing depends on team size, deployment, model and execution resources, governance requirements and integration scope. Improvement is measured in the pilot rather than guaranteed in advance.

## Important URLs

- English homepage: https://eachrun.com/
- Chinese homepage: https://eachrun.com/zh/
- Chinese insights hub: https://eachrun.com/zh/insights/
- Enterprise AI tool choice: https://eachrun.com/zh/insights/doubao-kimi-workbuddy-enterprise-ai-choice/
- Enterprise AI agent management guide: https://eachrun.com/zh/insights/enterprise-ai-agent-management/
- AI agent access and audit guide: https://eachrun.com/zh/insights/ai-agent-access-audit/
- AI deployment decision guide: https://eachrun.com/zh/insights/ai-private-deployment-guide/
- AI usage and cost management guide: https://eachrun.com/zh/insights/ai-usage-cost-management/
- Enterprise AI pilot checklist: https://eachrun.com/zh/insights/enterprise-ai-pilot-checklist/
- Sitemap: https://eachrun.com/sitemap.xml
- Full AI context: https://eachrun.com/llms-full.txt

## Contact

- Email: sales@eachrun.com
- Phone: 400-030-8696
- Company: Chengdu March Orange Intelligent Information Technology Co., Ltd / 成都三月橙智能信息技术有限公司