Perforce Software is making the Model Context Protocol (MCP) that artificial intelligence (AI) tools and agents invoke to access data available across its entire portfolio of DevOps platforms.

Jake Hookom, executive vice president of product for Perforce, said the code management, application testing and infrastructure management tools and platforms the company provides will now all be accessible to any AI tool or agent.

The overall goal is to make it simpler for DevOps teams to safely incorporate AI into their existing DevOps workflows, he added. That approach makes it possible to extend existing controls, traceability, and security into workflows that are being augmented by AI in a way that minimizes any additional risks that might be created, said Hookom.

Originally developed by Anthropic, MCP is now being advanced under the auspices of the Agentic AI Foundation, a newly created arm of the Linux Foundation. It is rapidly becoming a de facto open source standard for connecting AI applications to external systems to enable an AI agent to, for example, perform actions using current data rather than only data that the large language model (LLM) they depend on was trained on.

Additionally, MCP also makes it simpler for organizations to swap out AI agents and large language models (LLMs) as they best see fit, noted Hookom.

It’s not clear how pervasively AI tools and agents are being embedded into DevOps workflows, but a recent Futurum Group survey finds 60% of respondents said their organization is now actively using AI to build and deploy software. Overall, top areas of investment are AI Copilot/AI code tools (38%), AI agent development (37%), AI-assisted testing (37%) followed closely by DevOps (37%), automated deployment (34%), software security testing (31%).

Each DevOps team is, of course, adopting AI tools as they best see fit depending on the level of trust they have in the output being generated. In fact, the latest generation of AI tools are already taking advantage of more advanced reasoning capabilities to significantly improve the quality of the code being generated, noted Hookom. As the cost of using those tools inexorably declines, the next challenge will then be determining to what degree to enable an AI agent to autonomously perform an action either on behalf of an individual developer, engineer or the entire DevOps team, added Hookom.

The most important thing is to develop an overall strategy now versus randomly adding AI coding tools that lack the context enabled by, for example, providing access to the most relevant data via an MCP server, he added.

While adoption of AI currently remains uneven, the one thing that is certain is that as the number of silos that make up the software development lifecycle (SDLC) is reduced or outright eliminated using AI the more collaboration there will inevitably be, said Hookom.

Ultimately, it won’t be too long before every DevOps tool and platform becomes accessible via a, hopefully, secure implementation of an MCP server or client. The challenge and the opportunity might then become determining how best to organize a DevOps team when many of the silos that led to the creation of one bottleneck or another are finally eliminated.