Barndoor 收购开源工作流运行时 Frags 的创建者 Diaphora
英文标题：🔶 Barndoor acquires Diaphora, creators of open-source workflow runtime Frags
企业 AI 网关公司 Barndoor AI 宣布收购 Diaphora，后者开发了用于构建 AI 工作流的开源引擎 Frags。双方将结合 Diaphora 的工作流技术与 Barndoor 的治理、访问控制能力，让企业将可重复的 AI 工作流以 Blueprints 形式安全分发给获授权员工。Frags 与 Frags Modeling Language（FML）将继续保持开源，并由 Barndoor 团队支持。该交易旨在解决企业 AI 自动化在可靠性、权限访问和规模化分发上的难题，帮助 AI 项目从试点走向生产部署。
5.4 KB
4.3 KB
中文
要点速览
- 将可重复的 AI 工作流构建为在每一步均受治理、并可自动分发给获授权员工的“蓝图”（Blueprints）
- 旨在将 AI 自动化引入仍需要大量人工执行、风险更高的企业工作流
- 底层自动化引擎将继续保持开源，任何人都可以贡献、扩展或验证其工作原理
美国纽约州纽约市，2026 年 9 月 16 日 —— 企业 AI 网关 Barndoor AI 宣布收购 Diaphora。Diaphora 是 Frags 的开发公司，Frags 是一款用于构建 AI 工作流的开源引擎。合并后的产品将使企业能够构建 Blueprints，即通过一系列已定义步骤连接工具与数据、可由企业治理并在获授权团队中分发的可重复 AI 工作流。通过将 Diaphora 的工作流技术与 Barndoor 的治理和访问控制相结合，企业可以一次构建 AI 自动化，并将其安全地扩展至整个组织。
此次收购代表 Diaphora 被“反向并入”（spin-in）Barndoor。Diaphora 最初是由 Simone Pezzano 开发的独立项目，Jay Parisi 随着技术演进参与了协作。Barndoor CEO Oren Michels 随后成为 Diaphora 的顾问，并支持其早期发展。随着技术走向成熟，Barndoor 和 Diaphora 认为两家公司合并后具备更好的发展条件，因此将 Diaphora 的技术与团队纳入 Barndoor。Diaphora 全体团队将作为此次收购的一部分加入 Barndoor。
传统自动化主要聚焦于可预测、基于规则的任务，例如路由记录、同步系统以及处理常规请求。AI 有潜力实现更复杂工作的自动化，但企业仍面临三大主要障碍：可靠性、访问权限和分发。AI 生成的输出可能难以预测；同时，允许 AI 系统跨企业工具和数据采取行动，需要对其可访问及可执行的内容实施严格控制。Diaphora 和 Barndoor 正是为解决这一挑战的两个相对侧面而打造：Diaphora 专注于提高 AI 工作流的可预测性，而 Barndoor 则负责治理在企业中安全部署这些工作流所需的访问权限。
从 AI 试点到企业级采用
企业迅速引入了 AI 工具，但将实验转化为可重复工作流仍是一项挑战。在 Diaphora 中构建的自动化可以成为 Barndoor 内受治理的 MCP 工具。Barndoor 基于角色的访问控制决定了谁能够发现和运行每项自动化，无需单独的配置流程。员工随后可以访问获其角色批准的自动化，使企业能够在维持集中式治理的同时，安全地将 AI 工作流分发至各个团队。
为让 AI 工作流更可靠、更可预测而构建
Diaphora 构建于 Frags 之上。Frags 是一款开源 AI 工作流运行时，使用 Frags Modeling Language（FML）定义 AI 如何执行工作流，并限制 LLM 可以采取行动的位置。
“多数 AI 系统根据一份计划运行，也就是你希望发生什么。蓝图定义的则是实际会发生什么，”Barndoor 联合创始人兼 CEO Oren Michels 表示。“通过将 Diaphora 的工作流技术与 Barndoor 的治理能力结合，企业能够让 AI 自动化更具可预测性、更易部署，并让员工无需成为引导 AI 的专家也能使用它。”
蓝图通过一系列定义明确的步骤连接工具和数据，以创建可重复的工作流。AI 可以用于需要判断的环节，其余步骤则按照蓝图稳定执行。例如，一名销售人员在结束客户通话后，可能需要记录通话、更新客户资料、安排下一步并通知客户团队。蓝图定义了每次应如何执行这些步骤。如果它无法完成某个必要步骤，它将停止并指出问题，而不是用看似合理的内容填补空白。
治理使采用成为可能
Barndoor 治理 AI 在企业内可访问的工具、数据和模型，而 Diaphora Blueprints 将继承这些控制措施。员工只能看到和运行其角色获授权的自动化，同时企业会保留每项自动化访问了什么、修改了什么以及产生多少成本的记录。这使工作流能够一次构建并安全分发给获授权用户，而无需单独的配置流程。
“人们通常将 Barndoor 视为一款治理与安全产品，”Michels 表示。“这正是治理开始推动采用、而不是拖慢采用的时刻。没有人会使用自己不信任的 AI，也没有人会使用自己不知道如何使用的 AI。这一次同时解决了这两个问题。”
“Diaphora 的存在，是为了让 AI 工作流变得可预测，而不是让其执行听凭偶然，”Diaphora 联合创始人、Frags 创建者 Simone Pezzano 表示。“加入 Barndoor 后，我们可以将这种可靠性与企业级治理结合，并为团队提供一种实用方式，让他们无需为每项自动化重复创建访问控制，就能构建和部署 AI 工作流。”
这将开启的工作类型
能够行动的 AI，而不只是起草内容。如今许多在工作场所使用的 AI 工具会生成一份输出，之后仍需由人工审核和提交。蓝图可以被设计为自行完成操作，例如识别账单错误并发布已更正的发票，而不是仅提出一项修正建议供某人批准。
跨多个系统存在的答案。一些工作流需要来自不同系统的信息，而这些系统具有不同的所有者和访问控制。例如，季度客户健康度评审可能需要从一个系统提取合同、从另一个系统提取使用数据、从第三个系统提取支持历史、再从第四个系统提取账单信息。蓝图可为每一个步骤携带恰当的权限，使获授权员工无需直接访问每一个底层系统，也能运行该工作流。
“我们都构建过在笔记本电脑上运行出色的 AI 自动化，但要在整个企业内可靠地运行它们，则完全是另一回事。以 Agent 经济的速度构建自动化，需要的是快速、分布式且受治理的工作流，而不是任何人都无法扩展或监管的一次性脚本，”Diaphora 联合创始人兼 CEO Jay Parisi 表示。“这就是为什么加入 Barndoor 是合理的选择。我们正在共同将自动化创建与企业治理带到同一个平台。”
这一组合已经引起希望将 AI 工作流从实验推进至更广泛部署的企业共鸣。薪酬公平与决策智能领域公司、Barndoor 客户 Syndio 认为，可预测性和可见性对于将 AI 扩展到各团队至关重要。
“可靠性是将一个 AI 项目转变为我们能够交付给整个团队使用的东西的关键。有了 Barndoor，如果我们知道一项工作流每次都会遵循相同的既定流程，并且能够准确看到哪些内容与其发生了交互，我们就能比目前运行的任何工具更广泛地部署它，”Syndio CTO Nimrod Vered 表示。“Barndoor 通过收购 Diaphora 获得的新工作流能力开启了许多可能性。我们如今在 AI 中面临非确定性问题，这限制了我们能够将其扩展至各团队的程度。这可能是解决企业内部对 AI 自动化犹豫情绪的关键一环。”
开始使用
企业现在即可开始构建受治理、可用于生产环境的 AI 自动化。访问 diaphora.ai 注册，并让你的第一项自动化投入运行。
关于 Diaphora 和 Frags
Diaphora 构建于 Frags 之上。Frags 是一款 LLM 工作流引擎，可执行以 Frags Modeling Language（FML）编写的工作流。Frags 将数据检索和业务流程等复杂工作流视为受控操作，AI 未经授权无法修改这些操作。它确保 LLM 仅在必要的环节参与工作，从而节省成本并降低风险。Frags 和 FML 将继续保持开源，并由 Barndoor 团队提供支持，使每个人都能获得构建可信、可靠 AI 工作流所需的工具。
关于 Barndoor
Barndoor 是面向企业的 AI Gateway。Barndoor 为 IT、安全和 AI 团队提供治理 Agent、自动化和模型所需的可见性及细粒度访问控制。它保护每一项 AI 操作，治理 Agent 和自动化可通过 MCP 访问哪些工具和数据，在阻止 PII 和敏感数据泄露至模型或工具之前，保持 LLM 支出的可预测性。访问 barndoor.ai 了解更多信息。
English / 原文
TL;DR
- Build repeatable AI workflows as “blueprints” that are governed at every step and automatically distributed to authorized employees
- Designed to bring AI automation to higher-stakes enterprise workflows that still require significant manual execution
- The underlying automation engine will remain open source, allowing anyone to contribute, extend or verify how it works
NEW YORK, NY – September 16, 2026 – Barndoor AI, the AI gateway for enterprises, has acquired Diaphora, the startup behind Frags, the open-source engine for building AI workflows. The combined product will allow enterprises to build Blueprints, repeatable AI workflows that connect tools and data through a defined series of steps that can be governed and distributed across authorized teams. By combining Diaphora’s workflow technology with Barndoor’s governance and access controls, enterprises can build AI automations once and securely scale them across the organization.
The acquisition represents a “spin-in” of Diaphora, which began as an independent project developed by Simone Pezzano, with Jay Parisi collaborating on the technology as it evolved. Barndoor CEO Oren Michels later became an advisor to Diaphora and supported its early development. As the technology matured, Barndoor and Diaphora determined the companies were better positioned together, bringing Diaphora’s technology and team into Barndoor. The full Diaphora team will join Barndoor as part of the acquisition.
Traditional automation has largely focused on predictable, rules-based tasks such as routing records, syncing systems and handling routine requests. AI has the potential to automate more complex work, but enterprises still face three major barriers: reliability, access and distribution. AI-generated outputs can be unpredictable, while allowing AI systems to take action across enterprise tools and data requires strict controls over what they can access and do. Diaphora and Barndoor were built to address opposite sides of this challenge: Diaphora focuses on making AI workflows more predictable, while Barndoor governs the access required to deploy them securely across an enterprise.
From AI pilots to enterprise adoption
Enterprises have rapidly introduced AI tools, but turning experimentation into repeatable workflows remains a challenge. Automations built in Diaphora can become governed MCP tools within Barndoor. Barndoor’s role-based access controls determine who can discover and run each automation without requiring a separate provisioning process. Employees can then access the automations approved for their roles, allowing enterprises to securely distribute AI workflows across teams while maintaining centralized governance.
Built to make AI workflows more reliable and predictable
Diaphora is built on Frags, an open-source AI workflow runtime that uses the Frags Modeling Language (FML) to define how AI executes a workflow and constrain where an LLM can act.
“Most AI systems operate from a plan, what you hope is going to happen. A blueprint defines what actually happens,” said Oren Michels, co-founder and CEO of Barndoor. “By combining Diaphora’s workflow technology with Barndoor’s governance, enterprises can make AI automation more predictable, easier to deploy and accessible to employees without requiring them to become experts at directing AI.”
A blueprint connects tools and data through a defined series of steps to create a repeatable workflow. AI can be used where judgment is required, while the remaining steps execute consistently according to the blueprint. A salesperson finishing a customer call, for example, may need to log the call, update the customer record, schedule the next step and notify the account team. The blueprint defines how those steps are executed each time. If it cannot complete a required step, it stops and identifies the issue rather than filling the gap with something plausible.
Governance makes adoption possible
Barndoor governs which tools, data and models AI can access within an enterprise, and Diaphora Blueprints will inherit those controls. Employees only see and run automations authorized for their roles, while enterprises maintain a record of what each automation accessed, changed and cost. This allows a workflow to be built once and securely distributed to authorized users without a separate provisioning process.
“Barndoor is thought of as a governance and security product,” said Michels. “This is where governance starts driving adoption instead of slowing it down. Nobody uses AI they don’t trust, and nobody uses AI they don’t know how to use. This answers both at once.”
“Diaphora exists to make AI workflows predictable rather than leaving their execution to chance,” said Simone Pezzano, co-founder of Diaphora and creator of Frags. “By joining Barndoor, we can pair that reliability with enterprise-grade governance and give teams a practical way to build and deploy AI workflows without having to recreate access controls for every automation.”
The kind of work this opens up
AI that acts, not just drafts. Many AI tools used in the workplace today produce an output that a person must then review and submit. A blueprint can be designed to complete the action itself, for example, identifying a billing error and posting a corrected invoice rather than simply proposing a correction for someone to approve.
Answers that live across multiple systems. Some workflows require information spread across systems with different owners and access controls. A quarterly customer health review, for example, might require pulling a contract from one system, usage data from another, support history from a third and billing information from a fourth. A blueprint can carry the appropriate permissions for each step, allowing an authorized employee to run the workflow without requiring direct access to every underlying system.
“We’ve all built AI automations that work great on a laptop, but running them reliably across an enterprise is another story. Building automations at the speed of the agent economy demands fast, distributed and governed workflows, not one-off scripts nobody can scale or oversee,” said Jay Parisi, co-founder and CEO of Diaphora. “That’s why joining Barndoor made sense. Together, we’re bringing automation creation and enterprise governance onto one platform.”
The combination is already resonating with enterprises looking to move AI workflows from experimentation into broader deployment. Syndio, a pay equity and decision intelligence leader and Barndoor customer, sees predictability and visibility as critical to scaling AI across teams.
“Reliability is what turns an AI project into something we can put in front of the whole team. With Barndoor, if we know a workflow will follow the same defined process every time and can see exactly what touched it, we can deploy it far more broadly than anything we run today,” said Nimrod Vered, CTO of Syndio. “The new workflow capabilities Barndoor gains through the Diaphora acquisition open many possibilities. We have issues with nondeterminism in AI today, which limits how far we can scale it across teams. This could be a critical piece of solving the internal hesitation around AI automations.”
Get Started
Enterprises can start building governed, production-ready AI automations today. Visit diaphora.ai to sign up and put your first automation to work.
About Diaphora and Frags
Diaphora is built on Frags, an LLM workflow engine that executes workflows written in FML, the Frags Modeling Language. Frags treats complex workflows, like data retrieval and business processes as controlled actions that the AI cannot alter without authorization. It ensures that the LLM is only working where it’s needed, saving money and lowering risk. Frags and FML will remain open source and supported by the Barndoor team to empower everyone with the tools they need to build trusted, reliable AI workflows.
About Barndoor
Barndoor is the AI Gateway for enterprises. Barndoor gives IT, security, and AI teams the visibility and fine-grained access control they need to govern agents, automations, and models. It secures every AI action, governing which tools and data agents and automations can access via MCP, and keeping LLM spend predictable, while blocking PII and sensitive data leaks before they reach a model or tool. Learn more at barndoor.ai.