August 2025 Summaries

In modern applications, the ability to sign out from all devices simultaneously, known as "sign out everywhere," has become a crucial feature for enhancing user security and convenience. This functionality allows users to invalidate all active sessions across various devices, ensuring peace of mind whether they have lost a device, are concerned about security, or simply want to manage their logins efficiently. Implementing this feature using the WorkOS Sessions API is straightforward, as it handles the complexity of session revocation centrally, allowing developers to focus on providing a polished user experience. The tutorial outlines a step-by-step process for implementing this in Node.js, featuring session management, revocation, and the integration of WorkOS webhooks and APIs to handle session events. Additionally, it emphasizes best practices such as revoking sessions upon password changes and providing users with the ability to manage their sessions independently, aligning with the standards set by major platforms like Netflix and Google.

At MCP Night 2.0, Richard Moot from Block showcased an innovative solution for creating an MCP server for a massive API platform, specifically their Square API ecosystem, which includes over 30 APIs and more than 200 endpoints. Block developed a "layered tool pattern" to streamline the platform into three conceptual layers: Discovery, Planning, and Execution. This approach allows AI agents to dynamically discover available services, plan workflows, and execute API calls without needing to understand hundreds of specific endpoints upfront. Moot demonstrated this with a practical example, where an AI agent named Goose successfully created an invoice through natural language commands by leveraging self-directed exploration and the layered tools. This pattern not only reduces errors and context window clutter but also adapts to API changes without requiring tool updates, offering a scalable solution for complex API integrations. The presentation highlighted that higher-level abstractions and self-discovery processes can improve reliability and flexibility, suggesting the pattern's potential application beyond API platforms.

At MCP Night 2.0, held on August 7th at the Regency Ballroom in San Francisco, developers spotlighted cutting-edge tools and frameworks advancing the Model Context Protocol (MCP) ecosystem, with a notable presentation by Valentina Bearzotti from Basement Studio on XMCP. XMCP is a SaaS framework designed to simplify MCP server development, using a declarative, file-system-based approach that eliminates complex setup, enabling developers to focus on tool creation rather than infrastructure. Bearzotti demonstrated XMCP’s capabilities, including its simple tool syntax, middleware support, and flexible server configuration, showcasing how the framework's hot-reload feature and seamless deployment, especially with Vercel integration, significantly lower the barrier to entry for MCP server development. XMCP's integration into existing projects and its focus on enhancing developer experience make it a promising tool for teams aiming to quickly build or experiment with MCP servers. The event encouraged developers to explore XMCP at xmcp.dev and watch Bearzotti's full demo on YouTube, highlighting its potential to broaden MCP adoption.

MCP Night 2.0 highlighted the growing adoption and impact of the Model Context Protocol (MCP) ecosystem, with key insights shared by Eric Zakariasson from Cursor. Eric detailed the journey of MCP's development and adoption, noting significant momentum following the introduction of features like one-click installation and OAuth support, which reduced friction and increased developer engagement. Usage data showed stable growth, indicating genuine utility, particularly for tools that integrate external context into the coding environment. Prominent MCP servers include Context.7 for documentation, Supabase for database management, Playwright for browser automation, and Figma for design integration, each underscoring developers' desire for context-rich tools. A surprising addition was the LinkedIn MCP server, which Eric expanded to allow posting from within Cursor, showcasing the protocol's versatility. Eric emphasized the importance of reducing friction, providing valuable context, and using real data to enhance MCP adoption, offering a roadmap for future development within the ecosystem.

Mux's demonstration at MCP Night 2.0 highlighted the transformative potential of MCP (Managed Content Platform) for AI-driven video infrastructure, showcasing a secure and scalable solution for enterprise adoption. Matthew McClure, Co-Founder at Mux, presented a demo featuring the integration of WorkOS AuthKit to enable secure OAuth authentication for hosted MCP servers, addressing a critical enterprise need for centralized management without local installations. The demo illustrated how authenticated AI agents can utilize Mux's video analysis capabilities, such as accessing storyboards and transcripts, to automate and enhance content workflows. By partnering with WorkOS, Mux rapidly created a robust solution that integrates seamlessly with existing identity systems, paving the way for broader enterprise adoption. This integration not only ensures secure and compliant AI-powered video processing but also enhances developer productivity by making video primitives easily accessible through authenticated connections. Mux's approach exemplifies how established platforms can integrate AI capabilities while maintaining enterprise-grade security, serving as a model for API-first companies seeking to expand their services to AI agents securely.

OAuth 2.1 is an updated version of the OAuth 2.0 standard, designed to enhance security by simplifying implementation and mandating best practices, while eliminating insecure options like the implicit and password grants. Although still in draft form as of April 2025, it is being adopted in production by leading organizations, such as Anthropic, which uses it in their Model Context Protocol for secure AI agent interactions. This iteration consolidates scattered guidance into a unified specification, making it easier for developers to implement OAuth securely in web apps, APIs, and beyond. Key changes include making Proof Key for Code Exchange (PKCE) mandatory for all authorization code flows, prohibiting the use of bearer tokens in URLs, requiring refresh token rotation or sender constraints, and enforcing exact redirect URI matching to prevent vulnerabilities. These enhancements reflect a decade of lessons learned, offering a more secure foundation for both human-driven applications and autonomous systems, with tools like WorkOS providing an out-of-the-box solution for seamless compliance.

MCP Night 2.0, held on August 7, 2025, in San Francisco, showcased the maturation of the Model Context Protocol ecosystem, gathering 700 engineers, founders, and researchers to explore how AI systems integrate with real-world tools. The event highlighted MCP-UI, presented by Liad Yosef and Ido Solomon from Monday.com, which enhances AI interactions by introducing interactive user interfaces within AI conversations, addressing the limitations of text-based interactions. The framework enables the construction of rich UI components as MCP Resources that can be rendered by clients, maintaining visual identity and user experience. Notably, Shopify has integrated MCP-UI to improve commerce interactions, illustrating the framework's real-world applications by handling complex shopping scenarios through an intent-based messaging system. MCP-UI represents a shift toward more dynamic AI interfaces, potentially extending beyond visual to voice interactions, promoting a standardized approach that transcends individual AI platforms. As major players like Shopify adopt this technology, the focus shifts toward broader adoption, signaling a future where AI agents deliver interactive experiences directly, moving beyond mere descriptive interactions.

The Model Context Protocol (MCP) is a significant advancement in AI infrastructure, allowing large language models to interact with external tools and APIs, transforming them into active agents capable of executing tasks. However, this capability introduces unique security risks, including prompt injection, tool poisoning, command execution, and token theft. These vulnerabilities arise from the potential manipulation of model-generated inputs, the broad permissions often granted, and unverified endpoints. Several documented cases, such as SQL injection flaws and token leakage, highlight the potential for exploitation. To mitigate these risks, best practices are outlined for both MCP servers and clients, emphasizing strong authentication, input sanitization, token security, and endpoint verification. The need for secure discovery, trust anchoring, and robust audit mechanisms is critical, as is a shared responsibility between servers and clients to maintain the integrity and safety of MCP interactions. Tools like WorkOS can help implement these security measures, providing infrastructure that supports authentication, granular permissions, login monitoring, and session control, ensuring a resilient and secure deployment of MCP. Overall, the security of MCP is an ongoing process, requiring vigilance and adaptation to protect against both current and future threats.

The Model Context Protocol (MCP) faces significant challenges in securely authenticating and authorizing access to its servers, which are increasingly handling sensitive data. Developers often struggle between building custom authentication systems, which are time-consuming and complex, or relying on basic methods that are not adequate for production environments. WorkOS offers a solution by acting as an OAuth bridge, simplifying this process and allowing developers to focus on the core functionality of their MCP servers. WorkOS provides two main paths: AuthKit for those building new applications or migrating to modern authentication solutions, and Connect for those needing to integrate OAuth with existing systems. This approach eliminates the need for developers to delve into complex OAuth concepts, such as dynamic client registration, and has been successfully implemented by companies like Mux, which utilized WorkOS to launch their MCP server without having to develop in-depth OAuth support internally. By offloading authentication complexity to WorkOS, developers can concentrate on enhancing the unique features of their MCP servers, ensuring both security and functionality.

Many product teams face challenges with low-quality sign-ups, such as disposable emails, bots, and repeat free trial sign-ups, which can inflate growth metrics but ultimately damage business accuracy and trust. These fake accounts distort analytics, waste resources, create compliance risks, and erode trust with legitimate users. WorkOS Radar offers a comprehensive solution through a layered defense system that blocks disposable emails, detects and stops bots, prevents repeat sign-ups, and enforces compliance with sanctioned countries. By implementing these defenses, companies can maintain accurate analytics, reduce fraud-related costs, and ensure a smoother user experience, ultimately focusing on genuine customer growth and satisfaction.

In 2025, launching an iOS app on the Apple App Store requires adherence to Apple's authentication rules, particularly if the app supports third-party logins like Google or Facebook. A key requirement is implementing "Sign in with Apple" as a privacy-focused alternative, ensuring users' data privacy and preventing tracking across apps. WorkOS AuthKit offers a solution to integrate "Sign in with Apple" alongside other social logins, using a secure OAuth 2.0 PKCE flow. The process involves configuring redirect URIs, setting secrets, and managing authentication flows through a hosted login, ensuring compliance with Apple's guidelines. Additionally, securing tokens, implementing logout processes, and adhering to security best practices are crucial for maintaining app integrity and user trust.

In the rapidly evolving field of AI, the comparison between the Model Context Protocol (MCP) and the Agent-to-Agent Protocol (A2A) has gained significant attention, as each serves different purposes in the development of AI systems. MCP, developed by Anthropic, is an open standard designed to provide AI models with a consistent way to access external tools and data, acting as a universal adapter for seamless integration with various systems. It is especially useful for single-agent applications that require reliable tool access. Conversely, A2A, spearheaded by Google, facilitates the interaction and collaboration between multiple AI agents, enabling them to discover each other's capabilities and coordinate tasks in a decentralized, peer-to-peer manner, making it ideal for multi-agent workflows. While MCP focuses on agent-to-tool communication, A2A is about agent-to-agent interaction, and both rely on JSON-RPC to standardize agent interactions, reducing the need for custom integrations. Despite their distinct functions, combining MCP and A2A can provide a comprehensive foundation for scalable and secure AI systems, with MCP ensuring effective tool execution and A2A managing orchestration and collaboration among agents.

Secure and seamless authentication is crucial for SaaS platforms, with Single Sign-On (SSO) being a key component of this process. Building SSO in-house is often impractical for startups due to its complexity and the necessity of maintaining security and compliance standards, making specialized SSO providers a preferable option. These providers offer advantages such as handling complex integrations, providing robust security features, and supporting a wide range of Identity Providers (IdPs), which can lead to a lower total cost of ownership and higher reliability. The article evaluates several SSO solutions available in 2025, recommending WorkOS for its ease of integration and scalability, while also discussing other options like Auth0, Okta, Microsoft Entra ID, and Keycloak, each with its own strengths and ideal use cases. Ultimately, the choice of an SSO provider should align with the specific needs, infrastructure, and resources of the SaaS business, with WorkOS highlighted as a suitable choice for quickly implementing enterprise SSO.

In modern enterprise applications, Single Sign-On (SSO) and SCIM provisioning are crucial for seamless authentication and automated user lifecycle management. WorkOS facilitates the implementation of these features, allowing developers to integrate Okta SAML SSO and SCIM provisioning efficiently with Node.js, thereby providing secure authentication and streamlined user management for enterprise customers. The guide outlines the steps required to set up these integrations using the WorkOS SDK, including configuring connections on both the WorkOS and Okta dashboards, managing authentication processes, and ensuring real-time synchronization of user data. It also discusses methods for testing integrations, syncing users and groups, handling events using APIs or webhooks, and optionally streaming events to Datadog for enhanced monitoring and reporting. Additionally, the guide offers resources for further customization and integration options, such as example applications, SDKs, and detailed migration and configuration instructions.

Supporting multiple identity providers (IdPs) for single sign-on (SSO) in enterprise environments is more complex than simply implementing a standard like SAML, as different IdPs such as Azure AD, Okta, and Ping have unique implementations and quirks that require intricate handling. Initial SSO integrations often lead to unforeseen challenges because assumptions made for one IdP may not hold true for others, leading to breakdowns in authentication flows. The ongoing maintenance of these integrations is a significant burden due to varying certificate rotations, metadata handling, and error diagnostics, turning support teams into de facto identity management teams. WorkOS addresses these challenges by providing a unified integration that abstracts the complexities of different IdPs, offering a consistent interface for authentication, normalizing attributes, and centralizing error handling. This abstraction not only simplifies the integration process but also ensures scalability and reliability for enterprises engaging with numerous IdPs, positioning identity management as a critical platform-level concern rather than a mere feature.

The WorkOS Sessions API provides developers with the tools to manage user sessions across multiple devices, enhancing security and user control. By allowing the listing and revocation of active sessions, the API supports features such as "sign out everywhere," stricter password reset protocols, and account lockdowns in case of security threats. This capability is crucial in environments where users access applications from multiple devices, as it helps mitigate risks associated with long-lived session tokens that remain active even after potential breaches. Implementing these features can empower users to manage their sessions independently while enabling support teams to revoke sessions when necessary, such as when an employee leaves an organization. The Sessions API is designed to work seamlessly with platforms like Node.js, providing detailed session metadata, and supports integration with user interfaces to allow users to view and manage their active sessions.

In the B2B SaaS industry, managing feature access for different customers can be complex, with enterprise accounts, strategic partners, and internal teams often receiving varying levels of service and capabilities. Effective feature management is crucial for scaling a SaaS product, as it can enhance sales, streamline rollouts, and boost customer satisfaction, while poor management may result in code inefficiencies and deployment issues. Feature flags are an efficient tool for managing customer-specific access and gradual feature rollouts, serving as both entitlement and rollout flags to control access in multi-tenant applications. They offer flexibility and ease of use, allowing toggling features without redeployment, and can be managed via hardcoded checks, configuration files, custom entitlement systems, or third-party tools like WorkOS Feature Flags. These flags can be integrated into user authentication processes, enabling organization-level feature control with minimal engineering input, thus aligning product behavior with customer agreements and facilitating targeted beta programs. WorkOS Feature Flags provide a streamlined solution with instant changes, org-based access control, and a unified interface for both engineering and non-engineering teams, available at no extra cost for AuthKit users.

MCP Night 2.0 highlighted the rapid evolution and growing adoption of the Model Context Protocol (MCP) in connecting large language models to real-world systems, showcasing its potential to transform AI interactions into action-execution capabilities such as creating tickets and managing files. With over 700 attendees, including developers and product managers, the event emphasized the need for robust security and observability as critical factors for enterprise adoption. Presentations and discussions underscored challenges like security vulnerabilities, context window limitations, and the importance of seamless developer experiences. The event featured a panel discussion with representatives from Anthropic and OpenAI, focusing on making MCP enterprise-ready and addressing security challenges, particularly as AI systems gain access to organizational tools and data. Additionally, the evening revealed significant themes, such as the necessity of integrating AI systems with existing workflows and the development of new security frameworks to manage AI-driven interactions. The event concluded with a lively afterparty, further cementing MCP's role as a standard for safely integrating AI with enterprise operations.

WorkOS's technical screening process aims to assess candidates' problem-solving abilities, collaboration skills, and understanding of complex systems rather than their ability to produce perfect syntax or quickly memorize information. The interview is structured as a 60-minute session led by one or two engineers, designed to reflect real-world problem-solving scenarios at WorkOS, rather than testing obscure algorithm knowledge. Candidates are encouraged to use their preferred tools, but the emphasis is on navigating ambiguity, testing ideas, and communicating effectively with the interviewer, who acts as a collaborator. The evaluation focuses on three core competencies: technical ability, debugging skills, and communication, with an understanding that finishing the exercise is not the main objective. The goal is to simulate the collaborative, thoughtful, and curious work environment at WorkOS and to find engineers who enjoy tackling complex challenges with supportive teammates.

The Model Context Protocol (MCP) is a standardized system that enables large language models (LLMs) to securely interact with external tools, APIs, and services by providing a universal interface layer that bridges the gap between LLMs and real-world software. As AI systems become more integrated with external systems, MCP provides a framework for authentication and authorization to ensure secure operations, adopting OAuth 2.1 as its standard to handle these processes. MCP consists of components such as the MCP host, client, and server, each playing a role in maintaining secure interactions and managing permissions. Despite the benefits, the adoption of OAuth 2.1 poses challenges, such as the operational and security burdens placed on MCP servers, including metadata management and token handling. The community is working on evolving the protocol to balance compliance, developer usability, and security needs, aiming for a more flexible integration model that supports a range of deployment scenarios while ensuring the safe interaction of AI models with external environments.

AI is revolutionizing business innovation but simultaneously enabling increasingly sophisticated fraud tactics that challenge traditional defenses. As AI systems become central to modern applications, they become vulnerable to various types of attacks, including adversarial attacks, model extraction, data poisoning, and prompt injection, which exploit the decision-making processes of these systems. Emerging trends such as deepfake scams, synthetic identity fraud, and AI-powered crypto scams illustrate the growing complexity and scale of fraud, with incidents involving deepfakes and identity fraud on the rise. To counter these threats, businesses must adopt dynamic and advanced security measures that evolve at the pace of AI-driven attacks. These measures should include real-time anomaly detection, automated blocking, and zero-trust security approaches. Solutions like WorkOS Radar offer modern fraud detection and prevention capabilities by integrating behavioral signal collection, device fingerprinting, and customizable rules, ensuring that authentication processes remain robust against evolving threats.

Model Context Protocol (MCP) is emerging as a critical framework for enabling large language models (LLMs) to interact with external tools and data, offering features such as Tools, Resources, Prompts, Sampling, Roots, and Elicitation. MCP servers expose capabilities that allow clients to perform actions, access data, and guide models through reusable instructions, while tools require explicit user approval for execution, ensuring transparency and control. Resources provide read-only data for browsing without triggering operations, and prompts offer standardized templates for consistent task execution. On the client side, Sampling allows AI models to perform tasks on behalf of servers, Roots define secure filesystem boundaries, and Elicitation enables models to request additional user input during sessions, ensuring accurate responses. Together, these server and client features enable seamless, coordinated workflows that maintain privacy and user control, positioning MCP as a foundation for scalable and trustworthy AI applications.

In the realm of enterprise application security, Multi-Factor Authentication (MFA) and Single Sign-On (SSO) are complementary technologies that enhance user experience and security. SSO acts as a digital master key, allowing users to access multiple applications with a single login, reducing password fatigue and simplifying user management. However, it poses a risk if a malicious actor obtains the login credentials. To mitigate this, MFA introduces additional security layers by requiring users to verify their identity through various methods, such as biometrics or one-time codes, making unauthorized access much harder. The combination of SSO and MFA offers a balance between user convenience and robust security, particularly crucial for enterprises handling sensitive data or requiring compliance with security standards. Real-world applications, such as in finance, healthcare, and software development, demonstrate how SSO provides seamless access while MFA kicks in for high-risk actions. Solutions like WorkOS enable organizations to integrate SSO and MFA easily without extensive custom development, facilitating scalable and efficient security implementations.

In the realm of identity and access management, SAML (Security Assertion Markup Language) and OpenID Connect (OIDC) are key protocols, each serving distinct needs. SAML has long dominated enterprise authentication due to its rich, standardized semantics that cater to complex, hierarchical identity information, which is vital for legacy applications and compliance-heavy environments. Its architecture supports a hub-and-spoke federation model, enabling seamless cross-domain identity federation, especially useful in large-scale federations like universities and government. Conversely, OIDC is modern, lightweight, and more suited for consumer-facing and API-based applications, though it lacks some of the depth and interoperability provided by SAML. While OIDC adoption is increasing for new applications, SAML remains entrenched due to the massive investment in existing systems and the network effect of its widespread enterprise use. As organizations navigate hybrid environments, platforms like WorkOS are emerging to bridge the gap between these protocols, offering a unified solution that abstracts the complexities, allowing seamless integration with both SAML and OIDC, thus future-proofing applications as the identity landscape evolves.

In 2024, the advancement of large language models (LLMs) shifted towards reasoning models such as OpenAI’s o1, Claude 3.7 Sonnet, and DeepSeek R1, which focus on structured, multi-step reasoning rather than just providing quick answers. These models generate extensive internal reasoning traces, improving performance on tasks requiring logic, planning, and tool use, though they also increase latency and cost. The reasoning approach involves chain-of-thought (CoT) processes, allowing models to decompose problems, correct errors, and explore multiple solutions, significantly enhancing capabilities in mathematics, coding, and scientific reasoning. However, these models face challenges such as high computational demands, limited generalization, and the risk of misleading outputs, as they primarily rely on pattern matching rather than true logical reasoning. While they excel at complex tasks, they are inefficient for simpler ones, prompting selective deployment by developers to balance accuracy with computational costs. As the industry works on hardware optimization and hybrid approaches, reasoning models are expected to become more integrated and cost-effective, driving innovation in AI deployment strategies.

The tutorial outlines the process of integrating automated user provisioning between Microsoft Entra ID and an application using SCIM, Node.js, and WorkOS, emphasizing the importance of automated provisioning alongside SSO for enterprise applications. It guides users through syncing, updating, and deprovisioning users automatically without building a SCIM server from scratch. Key steps include installing the WorkOS Ruby SDK, configuring secrets, and setting up connections with Google Workspace. The tutorial also covers real-time synchronization of directory events using the WorkOS Events API or webhooks, recommending the events API for better control and reliability. Additionally, it provides instructions for retrieving specific users and groups, and offers optional steps for streaming events to Datadog for enhanced monitoring and analysis. The document highlights the dynamic nature of software tools, advising users to consult updated documentation when encountering issues.