Abnormal AI Innovation: How Abnormal Accelerates Developer Velocity with MCP

At Abnormal AI, we utilize the Model Context Protocol (MCP) to enhance our AI development workflows, combining cutting-edge technology with security. While AI tools such as Cursor accelerate research and development, MCP takes this a step further by securely integrating our developer tools with specific contexts and development environments like Jira, GitHub, and service logs. This secure integration enables our developers to seamlessly interact with third-party applications, accelerating development velocity.

As part of our Abnormal AI Innovation series, we’re showcasing how our engineering team is harnessing AI to build truly AI-native products—MCP being a prime example. Let's take a closer look.

MCP bridges Abnormal's development environments with our internal tools and infrastructure bidirectionally. The in-house MCP server operates locally on developer machines, while a secure proxy manages interactions with internal applications and data on GitHub, Jira, or our hosted metrics. This architecture ensures that sensitive data stays protected within Abnormal’s internal environment, upholding our strict security standards.

Abnormal’s MCP implementation is designed to operate entirely on local development machines to maintain control over sensitive code and data. This approach mitigates risks related to data exposure, external vulnerabilities, and compliance issues associated with third-party MCP solutions.

Access is tightly controlled through precise role-based authentication. When our developers use agentic integrated development environments (IDE) like Cursor, they are connected to the MCP server. The server then interacts with the authentication proxy and leverages existing VPN permissions and authentication systems to create a secure channel. This setup allows us to integrate AI assistance into our workflow while upholding the highest security standards.

At Abnormal AI, we build custom tools designed specifically for our workflows—because general-purpose solutions don’t meet the high bar we set for security, performance, and developer experience. Every tool we create is guided by three principles: security-first design, accelerated development velocity, and a unified, integrated environment.

By operating entirely within our MCP server and permission group system, these tools ensure that sensitive data remains protected and access is tightly controlled. We combine custom-built interfaces with vetted external APIs—accessible only through the MCP server—to maintain strict authentication and authorization standards.

This tailored approach enables us to streamline common engineering tasks, reduce context switching, and keep our teams focused on building the core products that protect our customers. Key components of our toolset include:

1. Seamless Git and GitHub Integration: Manages branches, commits, and pull requests entirely within the MCP environment—no tab switching required.

2. Effortless Jira Access: Surface issue tracking data directly in the IDE to keep engineers focused and informed.

3. Instant System Knowledge: Pulls real-time context on services, deployments, and incidents—right when it’s needed.

4. Automated Smart Testing: Detects the correct test framework based on file type and run tests in secure, pre-defined environments—no manual setup or rules required.

5. Batch Refactoring at Scale: Makes codebase-wide changes efficiently by applying transformation patterns across multiple files, all within secure boundaries.

This is how we accelerate development without compromising trust: purpose-built tools that reflect our AI-native mindset and security-first culture.

Abnormal's MCP implementation has significantly improved engineering workflows across our organization. Common tasks like branch management, automated testing, and batch refactoring are now automatically executed by MCP, saving time and reducing the effort spent on routine operations. This streamlined workflow minimizes context switches and enhances code quality, allowing developers to focus on their core mission of developing products that protect our customers.

By maintaining scalability and high code quality standards, our MCP implementation helps Abnormal AI stay ahead of emerging security threats. Integrating AI assistance into our development process has created a more efficient environment, enabling teams to work smarter and faster while upholding stringent security measures.

Ultimately, Abnormal's use of MCP translates into higher research and development velocity, allowing us to better protect our customers. In the ongoing battle against sophisticated threats, combining AI technologies creates a virtuous cycle that keeps us ahead of malicious actors. We leverage good AI to combat bad AI in a fast-evolving threat landscape.

This is just the beginning of how Abnormal is using AI to build AI securely and at scale to combat AI-powered threats. We'll continue to share our advancements in future posts in our AI Innovators series.

If challenges such as architecting secure AI systems, building internal platforms for developer velocity, or designing the future of software development interest you—consider joining our team. We're always looking for AI-native engineers passionate about security.

See Abnormal’s AI capabilities in action by scheduling a demo today!