July 2025 Summaries

The text outlines Merge's introduction of two new models, Pay Group and Bank Info, to its Unified HRIS API, which aim to enhance payroll management and employee benefits services. The Pay Group model allows companies to categorize employees based on payroll criteria, simplifying the creation of payroll runs, while the Bank Info model facilitates managing employees' bank account details for transactions and payments. These features address customer requests for improved payroll support and enable organizations to perform tasks like purchasing benefits on behalf of employees, linking bank accounts, and regulating pay rates efficiently. Merge encourages users to explore these models through a free account or demo, highlighting its role as an integration platform that manages customer integrations beyond being just a Unified API product.

The Model Context Protocol (MCP) is increasingly used for connecting AI agents with third-party applications, but its implementation presents several challenges. Many MCP servers were expedited to market due to external pressures, leading to vague tool descriptions and improper tool calls by AI agents, potentially exposing sensitive information and degrading performance. Additionally, poor maintenance as companies prioritize marketing over functionality can lead to ineffective and risky servers, with issues such as incorrect schema definitions causing operational errors. Security risks are significant, including the potential for malicious actors to exploit AI agents into exposing sensitive data and the emergence of fraudulent MCP servers designed to steal credentials. Extensive testing is crucial to ensure MCP servers meet performance requirements, though it is complex and resource-intensive. Merge is developing a platform to facilitate the integration and management of AI agents with MCP servers, aiming to streamline these processes and mitigate associated risks.

Model Context Protocol (MCP) servers are often poorly built, which can lead to issues with AI agents underperforming due to unclear tool descriptions, ambiguous error messages, and varying authentication mechanisms. To prevent these issues, it is essential to comprehensively test an MCP server before using it in production, using sandbox data for every test to avoid risking sensitive information. Setting up a wide range of test scenarios that define expected behaviors and evaluating the tools' hit rates and success rates can help identify strengths and weaknesses of the MCP server's tools. Additionally, tracking unnecessary tool calls and using a third-party solution to run tests can help manage the complex and time-intensive process of testing an MCP server, ultimately ensuring that AI agents perform optimally and reducing potential errors and costs. By following best practices for testing MCP servers, developers can build more effective and efficient AI systems that integrate seamlessly with various tools and services.

Merge, a unified API platform, prioritizes data security and has invested heavily in its infrastructure, encryption, and compliance frameworks, achieving GDPR compliance before onboarding its first customer. The company has introduced Destinations, a new offering that allows customers to store data directly in their own environment instead of Merge's server, providing enterprise-grade security and compliance, control over data hosting and access, and faster access to integration data. Destinations is designed for companies with strict data and compliance requirements and is currently available in closed beta for those on the professional or enterprise plan, supporting file storage category and Postgres as a data storage option, with plans to expand support to additional platforms in the future. By providing this capability, Merge aims to support the strictest data security requirements of enterprise companies, making it the first and only unified API platform to do so.

The emergence of Anthropic's Model Context Protocol (MCP) has led to confusion among developers and product managers, with some viewing it as a replacement for APIs, while others see it as inferior. However, the reality is that MCP and APIs are complementary, with many well-designed AI systems needing both. MCP provides a standardized way for large language models to interact with external data sources through existing APIs, managing context between them, while APIs handle core data transmission, authentication flows, and connections to different applications. The combination of MCP's context management capabilities with robust API infrastructure is necessary to address security challenges, such as access control and data loss prevention. Ultimately, understanding the relationship between MCP and APIs will give companies a competitive advantage in building sophisticated AI assistants, and recognizing when to leverage each will depend on the specific scenario and the complexity of the system being built.

Merge's Unified API is designed to provide engineers with flexibility and control over data access and transformation, offering a range of features that allow for customization and extensibility. The platform's Common Models can be extended through Field Mapping, which enables the definition of custom fields and mapping to third-party APIs, while Remote Fields allow for programmatic retrieval and writing to fields not normalized by Merge. Additionally, Remote Data provides access to raw, untouched payloads from source APIs, and Passthrough Requests enable fully authenticated, direct API calls to supported integrations. These features give engineers fine-grained control over data without compromising the simplicity of a unified API, making Merge a deeply customizable platform that combines the flexibility of direct API access with the ergonomics of a normalized integration system. By using these features, engineers can tailor Merge's integration infrastructure to their specific needs, and the company offers demos for those interested in exploring their own use cases.

A unified API platform provides a single, standardized interface to access multiple integrations within a software category, abstracting the complexity of individual integrations by normalizing data models, authentication flows, rate limits, and pagination. The market for unified API platforms is becoming increasingly saturated, with over 23 vendors now listed on the G2 Grid for Unified APIs, each offering different types of integrations, tooling, and support measures. Some unified API platforms specialize in specific software categories or industries, while others support several categories and are constantly expanding to new ones. When evaluating unified API solutions, it's essential to consider factors such as integration observability, sync flexibility, and white-glove support, as well as the platform's ability to support enterprise-grade integrations and provide scalable pricing models. Top unified API platforms, including Merge, Apideck, Kombo, Nango, StackOne, and Unified.to, offer varying degrees of integration coverage, customization options, and support for different categories and industries, making it crucial to assess each platform's pros and cons to determine the best fit for a particular business.

Merge's website utilizes cookies to enhance user experience, and by using the site, users agree to the collection and use of data as outlined in their Privacy Policy. The company also provides information on AI agents, which are software-based systems that perform tasks autonomously, and the Model Context Protocol (MCP), a standardized protocol that enables AI agents to interact with third-party applications. AI agents can manage repetitive tasks, answer questions, and provide insights, while MCP allows them to access tools from various applications, enhancing their capabilities. The relationship between AI agents and MCP is one of enablement, where MCP facilitates AI agents' interactions with external data, but AI agents do not directly impact MCP servers. Merge offers a Unified API that allows for hundreds of integrations, and the company provides resources on testing MCP servers, securing AI agents, and managing customer integrations.

Securing AI agents is crucial to prevent sensitive information from leaking, mitigate the risks of AI agent sprawl, and gain a competitive advantage. A single security incident can erode customer trust and put companies at risk of violating compliance regulations. To secure AI agents, companies can establish rules on the data they can receive and share, build alerts to notify teams of rule violations, use logs to track and troubleshoot individual tool calls, require authentication before users can invoke AI agents, and leverage an audit trail to detect and address internal risks. By implementing these measures, companies can ensure the security and integrity of their AI agents and protect sensitive data from unauthorized access. Additionally, using a 3rd-party solution or implementing these measures internally can help companies achieve this goal and demonstrate their commitment to security to prospects and customers.

AI agent observability refers to the measures put in place for monitoring and managing AI agents, including fully-searchable logs, rules to block or redact sensitive data, audit trails, custom alerts, and access controls. Observing AI agents is crucial as it prevents them from accessing and using sensitive data, enables teams to debug issues quickly, builds trust and credibility with prospects, provides insights into areas for improvement, and helps companies comply with key data privacy and protection frameworks. To perform AI agent observability, companies can set rules to keep sensitive information secure, leverage logs to track every action their AI agents take, implement alerts to detect security and performance issues in real-time, and use an audit trail to help admins address potential issues on time. By using AI agent observability tools, businesses can overcome challenges associated with observing and managing AI agents, such as scaling and resolving issues quickly, and can ensure that their AI agents are secure, compliant, and performing optimally.

The process of building an API integration is unique due to differences in how API providers approach pagination, authentication, and rate limits, as well as varying integration requirements for each product or internal workflow. To scope an API integration project, one must research the API provider's available endpoints, define build requirements, outline responsibilities among team members, assign timelines for each phase, and present the plan to stakeholders. This process involves several steps, including investigating API documentation, defining efforts based on endpoint details, planning for integration monitoring workflows, and clarifying roles and responsibilities within the team. By following these steps, teams can successfully scope and implement API integrations, and tools like Merge's Unified API can simplify this process by allowing developers to add hundreds of integrations through a single API, saving time and improving product reliability.