September 2025 Summaries

AI agents are increasingly capable of automating workflows, querying knowledge bases, and acting as intermediaries with sensitive APIs, which raises significant security concerns if they operate without proper permission controls. Without safeguards such as Role-Based Access Control (RBAC) and OAuth, these agents risk exposing sensitive data, overstepping intended boundaries, and becoming vectors for malicious activities, paralleling the challenges faced by early web applications before authentication and RBAC became standard. The text outlines several risks associated with unchecked AI agents, such as data leakage, privilege escalation, compliance violations, and trust erosion, emphasizing the necessity for a robust security framework that includes secure authentication, least privilege enforcement, OAuth scopes, user-to-agent delegation, and comprehensive logging. By implementing these security measures, similar to those used for human users, organizations can ensure that AI agents operate safely within defined limits, maintaining user trust and compliance with regulatory requirements.

AuthKit has expanded its capabilities, now supporting localization in 90 languages for a seamless authentication experience, including all user interfaces and emails, with automatic language detection based on user settings. It is now the default authentication provider for Convex, simplifying setup for new projects without the need for API key configuration. AuthKit also introduces multiple roles per organization membership to streamline permission management, and WorkOS has enhanced its offerings with features like custom email provider configurations, Slack notifications for feature flags, and customizable widgets for SSO status and domain verification. Additionally, the platform now supports user authentication via the Vercel Marketplace and provides enhancements like the Standalone SSO API and best practices for secure model-agent interactions.

WorkOS's Fall Launch Week introduced five significant product updates aimed at accelerating the development of Enterprise Ready applications. These updates include the ability for AuthKit to support multiple roles per user, simplifying access models by allowing users to have a combination of roles with a union of permissions. AuthKit also became the default authentication provider for Convex, streamlining project setup without the need for API keys or environment variables. Additionally, AuthKit now supports localization in 90 languages, enhancing user experience with auto-detected language preferences for its UI and emails. WorkOS has integrated custom email provider support, allowing users to utilize their existing email systems like Amazon SES or SendGrid for authentication emails while tracking email events through the WorkOS dashboard. Finally, the introduction of Slack notifications for feature flags ensures team members stay informed about changes in flag creation, updates, and targeting through direct updates in Slack channels.

The newly launched WorkOS Slack app enhances team coordination during feature rollouts by providing real-time notifications of feature flag events directly to Slack, eliminating the need for complex configurations or middleware. This integration, available to all AuthKit customers, ensures that support, engineering, and product teams receive immediate updates on feature changes, including flag lifecycle events and rule updates, which are crucial for managing B2B product launches. Notifications are designed to be clear and actionable, offering direct links to detailed event information, thereby aiding customer support in managing inquiries, helping engineers track changes, and allowing product managers to monitor rollout progress without needing to frequently check dashboards. The app's setup is straightforward, requiring only a few steps to connect to a Slack channel, and while it currently focuses on feature flag notifications, it is designed to accommodate future WorkOS event notifications based on customer feedback.

Email plays a crucial role in authentication processes, but many enterprises face challenges with shared email services, such as poor deliverability and compliance issues. WorkOS has introduced a feature that allows customers to configure custom email providers like Amazon SES, Postmark, Resend, or SendGrid directly from the WorkOS Dashboard. This new functionality offers the control and visibility associated with using a dedicated provider while maintaining the ease of WorkOS-authored authentication flows. It addresses the need for improved deliverability, better observability of email delivery and bounce statuses, and compliance with enterprise requirements to use specific email infrastructures. Customers can configure their chosen provider by entering necessary credentials, with all data securely stored, and verify domains to ensure emails are sent using their verified domains. This setup enhances deliverability and compliance without compromising the developer experience, allowing enterprises to manage their email infrastructure effectively.

AuthKit has introduced a robust localization feature, now supporting over 90 languages to enhance global accessibility and user experience without any additional setup or cost. By automatically detecting the user's preferred language through the browser's accept-language header, AuthKit renders all user-facing elements, such as sign-up boxes and emails, in the appropriate language, broadening the potential user base and increasing conversion rates. This feature is crucial for businesses aiming for international expansion, as it eliminates the need for engineering efforts or translation costs and supports global teams. The system relies on pre-computed translations, ensuring no extra runtime overhead, and provides a fallback to US English when a locale is unrecognized. Users can preview multilingual interfaces and customize text to maintain brand consistency, with translations automatically generated. Additionally, user locale data is accessible via API for enhanced marketing and analytics insights, reinforcing that global accessibility should be the standard for all AuthKit customers.

AuthKit has become the default authentication option for Convex projects, simplifying the process of building real-time applications with production-grade authentication. The integration allows developers to set up a fully authenticated Convex application in under 60 seconds, eliminating the need for manual configuration and multiple dashboard interactions. Convex, a reactive backend platform, is popular for its ability to seamlessly sync data across clients in real-time applications such as collaborative apps and multiplayer games. The integration automates the setup of WorkOS resources, including team creation, API key management, JWT template configuration, and CORS origins, thereby enhancing the developer experience. This new setup provides a seamless, scalable solution with multiple environments and team collaboration capabilities, all at no extra cost, enabling developers to focus on feature development rather than infrastructure management.

AuthKit has introduced a significant upgrade to its Role-Based Access Control (RBAC) system by allowing multiple roles per organization membership, which aligns more closely with real-world organizational structures. Previously, AuthKit restricted users to a single role per organization, which often led to complex workarounds like creating hybrid roles for users with cross-departmental responsibilities. The new multi-role feature enables teams to assign users multiple roles, automatically calculating the union of their permissions, thus streamlining role management and reducing unnecessary complexity. This update offers several advantages, including clearer role design, reduced role proliferation, predictable access control, and scalability as organizations grow. It also simplifies modeling and maintenance of RBAC setups, even as organizations become more complex. Although multi-role support disables IdP role assignments for now, future updates will integrate this feature and enhance permission syncing across environments. Multi-role capability is currently available and can be activated through the Roles & Permissions page on the AuthKit dashboard.

Model-Client Platform (MCP) introduces novel security challenges as it allows models to generate natural-language instructions executed by agents with significant privileges, leading to a unique attack surface. Unlike traditional client-API designs, model-agent interactions are unpredictable and context-sensitive, posing risks such as prompt injection, over-privileged agents, and data exfiltration. Security measures include utilizing strict schemas, authentication, request signing, and role-based access controls to limit privileges and contain potential breaches. Man-in-the-Middle (MitM) attacks can be mitigated with mutual TLS and certificate pinning, while replay and lateral movement threats require nonce usage and identity-based routing. Data exfiltration can be controlled through data loss prevention checks and privacy-aware training, and supply-chain risks demand maintaining a software bill of materials and sandboxing third-party dependencies. By implementing comprehensive security practices, such as those offered by platforms like WorkOS, organizations can establish a robust, secure model-agent ecosystem, transforming the risks of MCP into manageable challenges and ensuring a resilient AI infrastructure.

In the context of SAML (Security Assertion Markup Language) integrations, self-signed certificates are typically sufficient and secure, as SAML does not rely on the traditional Certificate Authority (CA) trust model used in HTTPS/TLS. Unlike web communications that require CA validation to establish trust, SAML certificates are used for cryptographic signing and verification directly between an Identity Provider (IdP) and a Service Provider (SP). This approach ensures that assertions are genuinely from the IdP and have not been altered, with trust established directly between the parties through the exchange of public certificates. While CA-signed certificates can be useful in certain situations, such as when organizational policies require them or for simplifying certificate management, they are not necessary for most SAML use cases. Self-signed certificates are widely used and accepted for SAML, with CA-signed certificates primarily mandated by internal security policies or regulatory requirements rather than by the SAML specification itself.

Air-gapped environments, often used in high-security, regulated, or enterprise settings, isolate systems from external networks to meet compliance, security, or operational requirements. This isolation can be physical or logical and is common in industries like defense, healthcare, and finance, where regulatory frameworks demand strict data protection measures. WorkOS provides solutions to address authentication challenges in such environments, where traditional network-based authentication methods may fail due to isolation. By offering environment isolation and controlled network interactions, WorkOS enables secure authentication in air-gapped, on-prem, and hybrid deployments without compromising security. It allows organizations to maintain modern authentication processes while adhering to stringent security and compliance needs, supporting both cloud-based and isolated systems through a dual-implementation approach.

The WorkOS team hosted a successful 2-day Enterprise MCP hackathon, attracting around 100 participants who collaborated on innovative projects using WorkOS technology. The event was marked by a vibrant and positive atmosphere, with attendees expressing interest in using the office space for future events. Despite logistical challenges such as managing resources and maintaining the event flow, the team effectively coordinated speakers, judges, and supplies. Noteworthy projects included WebMCP, which dynamically creates MCP server tools, a documentation server that innovatively navigates through Markdown files, and a security scanner addressing vulnerabilities in MCP servers. The hackathon fostered a strong sense of community, with participants from various companies supporting each other and sharing entrepreneurial insights. The enthusiasm generated has already set plans in motion for future events, with an emphasis on better resource stockpiling.

As the demand for robust identity and access management (IAM) solutions grows, many companies that initially opted for Auth0 are exploring alternatives due to rising costs and limitations in flexibility. By 2025, several strong contenders have emerged in the customer identity and access management (CIAM) market, including WorkOS, Microsoft Entra ID, Amazon Cognito, Firebase Authentication, and Keycloak. These alternatives offer varying strengths such as ease of integration, customization, ecosystem alignment, and scalability, catering to different organizational needs, whether for enterprise, consumer, or hybrid models. WorkOS, for instance, is ideal for SaaS products requiring enterprise features without the need to build from scratch, while Keycloak provides a self-hosted, highly customizable solution with significant control over identity infrastructure. The choice of IAM provider hinges on specific factors like user base, ecosystem compatibility, customization needs, operational complexity, and cost scalability. For those transitioning from Auth0, detailed steps and documentation are available for migrating to platforms like WorkOS, ensuring a smooth and strategic shift in identity management solutions.

Authentication is a crucial feature in application development, and WorkOS simplifies the integration of modern authentication methods like Single Sign-On (SSO), Multi-Factor Authentication (MFA), and Passwordless login for .NET applications. This tutorial guides developers through implementing these authentication strategies using WorkOS SDKs and APIs, allowing them to choose the method that best suits their product needs, such as SSO for enterprise clients or Passwordless for lightweight apps. It outlines the process of setting up SSO by creating authorization endpoints and handling callbacks, adding MFA through TOTP or SMS codes for enhanced security, and enabling Passwordless login using one-time codes sent via email. The tutorial also covers the prerequisites and steps for installing the WorkOS .NET SDK and securing API keys. By incorporating these authentication methods, developers can build applications with scalable, enterprise-ready authentication flows. Additionally, WorkOS offers advanced capabilities such as Directory Sync, Admin Portal, Roles and Permissions, and Audit Logs to further enhance security and facilitate enterprise adoption.

OAuth 2.0 introduced a flexible framework for delegated authorization, evolving over time to address emerging security concerns, leading to the introduction of Proof Key for Code Exchange (PKCE) as an enhancement to the Authorization Code Flow, particularly for public clients like single-page applications and mobile apps. With OAuth 2.1, PKCE has become mandatory for all applications, including confidential clients, due to its ability to protect against code interception attacks and ensure a consistent and universal security model. PKCE works by adding parameters such as code verifier and code challenge, offering additional safeguards beyond client secrets, and is recommended for both public and server-side apps. Despite OAuth 2.1 being a draft, adopting its guidelines ensures compatibility with current servers, enhances security, and aligns with future standards. WorkOS supports OAuth 2.1 adoption by providing a platform that incorporates modern security practices, enabling developers to focus on user experience while ensuring secure authentication processes.

User provisioning is a fundamental component of enterprise-ready applications, and the System for Cross-domain Identity Management (SCIM) protocol facilitates this by standardizing the exchange of user and group data between Identity Providers (IdPs) and applications. While most enterprise IdPs like Okta and Azure AD come with built-in SCIM clients, custom IdPs require the development of a SCIM client to integrate with service providers like WorkOS. This process involves understanding authentication, supported endpoints, and crafting requests for user and group management operations such as creation, update, deactivation, and deletion. SCIM operates over HTTP using predictable REST endpoints and JSON schemas, as outlined in RFC 7644 and RFC 7643, and requires careful handling of authentication, idempotency, and error responses to ensure reliable synchronization. The guide emphasizes the importance of maintaining consistency, managing payload sizes, and handling deactivated users to prevent security vulnerabilities. A checklist for going live is recommended to ensure compliance, resilience, and interoperability with enterprise IdPs, thereby streamlining user management and enhancing enterprise adoption of applications.

The MCP Registry aims to streamline the discovery of servers within the MCP ecosystem, offering a centralized catalog that alleviates the previous fragmentation seen in server discovery. While this represents progress, traditional API key authentication methods pose significant challenges, creating friction due to their cumbersome distribution and management processes. The text advocates for OAuth as a superior solution, transforming the connection experience by enabling seamless, single-step authentication flows that enhance security and reduce operational overhead. OAuth's token-based system provides robust security features, surpassing the static nature of API keys, and allows for integration with trusted identity providers, thus leveraging existing authentication infrastructures. By standardizing on OAuth, the MCP ecosystem can foster increased experimentation and application development, as server developers are encouraged to implement OAuth to ensure both backward compatibility and a future-ready authentication strategy. This transition promises to unlock the full potential of the MCP Registry, making it both secure and accessible while encouraging innovation within the community.

Multi-factor authentication (MFA) is a critical security measure that significantly reduces the risk of account takeovers by requiring users to provide multiple forms of verification, such as passwords, device tokens, or biometrics. Effective MFA strategies balance security with user convenience, differing for consumer apps (B2C) and enterprise platforms (B2B). Best practices include offering multiple authentication options, using adaptive risk-based authentication, enforcing MFA across all organizational access points, and guarding against MFA fatigue. It's crucial to have fallback options for device failures and to integrate MFA with single sign-on (SSO) and access controls to create a resilient security framework. B2C applications prioritize ease of use and recovery, while B2B applications demand robust security and policy enforcement. Tools like WorkOS can streamline MFA implementation, allowing for enterprise-grade security without extensive development time, and ensuring both security and seamless user experience.

At WorkOS, product engineering is defined by a comprehensive project lifecycle that integrates customer understanding, market analysis, and rapid product development. This process, inspired by 37signals’ Shape Up methodology, involves several stages: project brief, shaping, Hilltop, build, ship, and post-ship. Each stage is designed to ensure clarity, encourage collaboration, and empower engineers to act as product owners. The lifecycle begins with a project brief outlining the problem and potential solutions, followed by shaping, where engineers deeply engage with the problem, collaborate with designers, and draft a solution. The Hilltop stage involves presenting this solution for feedback, which informs the subsequent build phase, where the product is developed with a focus on user experience and developer joy. Shipping involves enabling teams and selecting an appropriate rollout strategy, while post-shipping ensures continued alignment with customer needs and addresses any issues. This methodology emphasizes engineers' autonomy and expertise, fostering an environment where they become domain experts and integral to successful product delivery.

The Model Context Protocol (MCP) has made significant advancements with the introduction of MCP-UI, an extension designed to integrate interactive web components into AI agent conversations, transforming human-AI interfaces by moving beyond text-only interactions. Highlighted at MCP Night 2.0, MCP-UI aims to bridge the gap between text responses and actionable UI elements, particularly for domains like commerce and data visualization. It extends MCP's existing framework with a new UIResource interface, offering three rendering approaches: inline HTML, external URL resources, and Remote DOM integration, each tailored for specific applications and security needs. The system's architecture supports both server-side SDKs for resource creation and client-side rendering with React components. MCP-UI's event system ensures agents maintain control over application logic while enabling rich user interactions through structured events. Real-world implementations, such as Shopify's commerce applications and Block's Goose agent, demonstrate its capability for handling complex interactions seamlessly. Despite its benefits, MCP-UI introduces performance and security challenges, and its reliance on certain frontend frameworks may limit broader adoption. Future directions for MCP-UI include support for declarative UI patterns, cross-platform expansion, and integration with generative AI to create adaptive interfaces, signaling a shift towards more dynamic and interactive agent interfaces.

At MCP Night 2.0, representatives from Anthropic and OpenAI discussed their collaborative embrace of the Model Context Protocol (MCP) as an open standard for AI tool integration, underscoring its potential to revolutionize AI workflows by standardizing communication between models and integrations. Theo from Anthropic shared how MCP emerged from the inefficiencies of manually managing context across AI models, leading to a standardized approach that simplifies building integrations. OpenAI's Dimitri highlighted their adoption of MCP, driven by community enthusiasm and its promise for standardization over proprietary methods. Both companies have seen significant internal adoption of MCP, using it to create agile, microservices-style architectures that prioritize intelligence and user experience over integration. Security concerns remain a significant barrier to wider adoption, especially as organizations seek to automate sensitive workflows. The panelists emphasized the importance of open development, with MCP's governance model fostering transparency and collaboration via GitHub and Discord. They also noted opportunities in improving server infrastructure, authentication, and developer ergonomics to facilitate broader adoption. MCP's open and collaborative development approach suggests a promising future for AI tool integration, potentially establishing MCP as essential infrastructure in the agentic future of AI.

At MCP Night 2.0, Marissa Felix and Kenneth Sinder from Notion showcased the capabilities of Notion's Model Context Protocol (MCP) server, illustrating its potential to streamline development workflows by connecting documentation directly with code implementation. They demonstrated how a Product Requirements Document (PRD) in Notion could be transformed into a functional prototype using a single prompt, highlighting the system's bidirectional integration capabilities that allow updates to both documents and code. Notion's MCP implementation has evolved from a self-hosted package to a sophisticated, remote server, featuring simplified installation, enhanced Markdown support, and user-based permissions, which facilitates easier integration with development environments like ChatGPT, VS Code, and others. This integration enhances documentation-driven development, promotes living documentation, and supports project coordination directly within the development environment. Despite challenges with client compatibility and complex schemas, Notion is actively working on improving the MCP ecosystem's security, functionality, and server discovery protocols. The implementation signifies a shift towards AI-enabled development workflows, where documentation and code remain synchronized, reducing friction between product requirements and actual development.

At MCP Night 2.0, Harald Kirschner from VSCode addressed five significant pain points of the Model Context Protocol (MCP) ecosystem, transforming them into opportunities for improved user experience. He introduced solutions such as simplifying the installation process of MCP servers by integrating an official registry, enhancing security by managing API keys more securely, and improving tool discovery through a built-in picker and "context engineering building blocks" for reusable chat modes. Additionally, VSCode's latest release supports a vast number of tools simultaneously, allowing developers to maintain flexibility and efficiency without overwhelming their systems. To combat context bloat, a new sampling feature condenses search results, and resource links aid in managing information. Furthermore, VSCode's developer mode enhances the debugging experience by offering automatic server restarts, comprehensive logs, and full debugging support, addressing the real frustrations developers face. Kirschner's systematic approach to resolving these issues demonstrates a commitment to refining the MCP user experience and sets a precedent for the broader ecosystem to follow, emphasizing that technical success relies on enhancing developer satisfaction.

Automated user provisioning through SCIM (System for Cross-domain Identity Management) is crucial for SaaS platforms targeting enterprise customers, as it ensures seamless user lifecycle management by synchronizing user information between identity providers and apps. While building SCIM in-house can divert resources from core product development, many startups opt for specialized SCIM providers to navigate the complexities of varying implementations, compliance needs, and scaling challenges. SCIM providers like WorkOS, Auth0, and Stytch offer differing approaches to integration, pricing, and scalability, each catering to specific needs and growth stages. WorkOS is highlighted for its straightforward integration and predictable costs, making it ideal for fast-growing SaaS companies. Auth0 provides a comprehensive identity platform with extensive customization options, albeit at a higher complexity and price. Stytch offers a simple API with a generous free tier, suitable for those experimenting with modern authentication features, though its SCIM capabilities are still maturing. Choosing the right SCIM provider depends on the specific requirements and strategic goals of the SaaS company.