July 2025 Summaries

The Parallel Search MCP Server is now available in Devin's MCP Marketplace, providing AI software engineers with advanced web research capabilities that enhance development workflows by allowing real-time access to ranked URLs with excerpts optimized for AI reasoning. This infrastructure eliminates the need for complex scraping and parsing processes, enabling developers like Devin to query technical forums such as Stack Overflow and GitHub for real-time debugging and troubleshooting solutions. The integration facilitates learning from open-source projects and accessing current API documentation, framework guides, and technical specifications, ensuring compatibility and up-to-date information. By offering insights into the latest library versions and dependency management, Parallel Search MCP transforms how developers manage packages and dependencies, ultimately streamlining the development process with a web-aware approach.

The Parallel Task API has introduced Tool Calling via MCP Servers in beta, allowing developers to integrate their own tools, data sources, and execution logic into web research workflows with a single API call. This feature enables tools hosted on MCP-compatible servers to be accessed through the Task API, facilitating connections to private databases, code execution sandboxes, or proprietary APIs without requiring custom orchestrators or standalone MCP clients. Tool Calling supports the use of existing infrastructure while ensuring that the execution logic remains private and secure, as developers control which tools are accessed and how they are authenticated. The API provides complete traceability by including details of all tool calls, arguments, responses, and errors in the response. Users can begin integrating their systems by connecting their MCP server and consulting the provided documentation for guidance.

Source Policy is a new feature available for the Parallel Task API and Search API, designed to give users precise control over the domains accessed by AI agents, enhancing the reliability and accuracy of research results. By specifying include_domains and exclude_domains, users can ensure that queries are restricted to trusted sources and irrelevant or unreliable domains are blocked. This feature is particularly beneficial for production AI applications that require access to reliable and verifiable sources, such as conducting due diligence, sales research, or competitive analysis. Source Policy allows users to constrain the web research pipeline from the outset, reducing noise and ensuring that outputs are derived from trustworthy sources, without the need for post-processing filtering. This tool is now accessible across all Task API and Search API processors, and developers can begin utilizing it through the Developer Platform or by consulting the documentation.

Parallel Tasks are optimized for handling large-scale workloads by facilitating the execution of numerous independent web research tasks through the Task Group API, which consolidates these operations into a single batch for streamlined management. Users can create a task group, initiate runs within that group, and monitor live events through real-time result streaming, reducing the complexity of managing multiple tasks simultaneously. This API is particularly beneficial for large-scale data operations such as enriching CRM records, performing bulk due diligence, or executing competitive intelligence workflows by offering unified monitoring of task statuses and dynamic expansion capabilities that allow additional tasks to be added mid-execution. Available in public beta, the Task Group API enables developers to efficiently build and scale their applications using the Parallel Developer Platform.