We Built Our Own MCP Server for Engineers & Release Managers

Speed or safety? 

We hear from Platform and DevEx teams that they feel like they’re at a crossroads right now: embrace AI-led development and move faster—or maintain rigorous security and governance standards but risk falling behind.

Luckily, there’s a solution that turns this crossroads into an opportunity, rather than an ultimatum. 

Feature flags come out of the gates swinging in the showdown for safe and fast software development, allowing teams to unlock both. For our customers operating in highly-regulated environments, this balance is especially important. 

Introducing the Flagsmith MCP Server 

With that, we’re excited to announce the Flagsmith MCP Server—built to make feature flag management easier at scale and bring feature flagging directly into AI-driven development flows. Feature flags allow for the safe testing, rolling out, and rollback of AI-generated code. The MCP server reduces the admin work involved in responsible feature flagging and allows for automated checks and balances. 

Our aim with this extension of our platform is to provide teams with the flexibility to decide which workflows they run through our UI and which they run through the MCP Server. AI Agents can use the MCP server alongside humans and it easily integrates with your CI/CD pipeline. Our mission, always, is to provide teams with the flexibility to meet their own unique and changing needs, safely.

Why we built the Flagsmith MCP Server

As the ratio of developers using AI assistance when coding continues to grow along with AI-powered IDEs and “vibe coding” platforms, it’s obvious where things are headed. If AI is increasingly writing the code, it only makes sense that it should also handle wrapping that code in feature flags, choosing environments, and managing rollouts from the start.

AI has made development faster than ever before, but this isn’t without its drawbacks. The need for safety checks and guardrails has become increasingly apparent. 

“Feature flagging has always been powerful but never frictionless,” said Asaph Kotzin, Principal PM at Flagsmith. “With this MCP server, flags become part of the coding flow itself. Cursor, Windsurf, Copilot, or IDEs - wherever you're coding, Flagsmith is there with you, providing all the tools engineers and release managers need without breaking their stride.”

The idea is deceptively simple. Instead of exposing all 500+ API endpoints to an AI agent, we’ve built a server that only surfaces the tools a given role needs. For developers, that means creating and configuring flags directly from inside the IDE, with AI updating the code automatically. For release managers, it means running canary releases, scheduling staged deployments, and managing change requests, all by simply chatting with their AI assistant.

Enabling deep work

As an engineer, the Flagsmith MCP Server gives you everything you need—at your fingertips via natural language—without having to learn a new UI. You can set and forget key practices that once took up valuable time. 

Engineers are constantly being asked to context-switch between tools and platforms, despite the negative impact on deep work. The MCP server takes over tasks that were once manual, acting as a virtual assistant running in the background so that engineers can focus on their most important tasks. Additionally, the Flagsmith MCP Server allows you to set naming conventions and best practices at an organisational level, reducing cognitive burden.

Creating flags from the CLI

Imagine a developer midway through building a new checkout flow. They type: ^{Wrap this in a flag called checkout_v2, default off in staging.}

‍Instead of pasting snippets or wiring SDK calls manually, the Flagsmith MCP Server creates the flag in Flagsmith, initialises its values, and updates the code automatically. The flag exists, the code is safe, and the developer never leaves their editor.

“The code just comes back already wrapped and fully functional. Developers don’t even need to think about the plumbing. It’s just there,” Kotzin told me.

Release managers, too, finally get a smoother path. Historically, staged rollouts required clicking through dashboards, double-checking segments, and creating tickets for approvals. Now, they can ask:

^{What’s the status of search_v3 in staging? Start a 10% canary and ramp it to 50% tomorrow at 9am.}

The MCP Server checks the status, configures the rollout, schedules the increase, and files the change request automatically.

“Releases used to mean late-night dashboards and endless checklists, especially for larger enterprise organisations,” Kotzin said. “Now you just say what you want, and the rollout configures itself.”

Fighting tech debt 

Keeping your feature flags healthy—monitoring and removing stale flags from your codebase—is an important part of feature flag management, but it can be hard to stay on top of this. With the Flagsmith MCP Server, you can identify and remove stale flags, compare the status of different flags across multiple environments, and create change requests with multiple flags, all from the CLI. No more manual clean-up—Flag Hygiene becomes automatic. 

MCP Servers are the future

For engineering leaders, this launch is more than convenience—it’s strategy. Teams everywhere are under pressure to increase velocity by leaning on AI, but leadership worries about the risks of moving too fast and using fully AI-generated code. By embedding Flagsmith directly into the AI coding loop, those leaders get confidence that every new feature ships with guardrails from the outset.

AI has made development faster than ever before, but this isn’t without its drawbacks. The need for safety checks and guardrails has become increasingly apparent. And what good is speed if PRs simply languish in endless queues and QA? 

“With our MCP server, leaders get the best of both worlds,” Kotzin explained. “Your teams move faster with AI in the loop, but feature flags ensure no risky deployments slip through. It’s higher velocity without gambling with production.”

That vision scales far beyond a single release. If AI agents are going to become first-class members of development teams, they need reliable runtime controls—and Flagsmith is positioning itself to be exactly that. We see a future where vibe-coding platforms like Bolt and Lovable, which already let users “build by chatting,” automatically wrap every new feature in a flag, ready for safe rollout.

“As AI takes on more of the coding, it just makes sense that every new feature comes wrapped in a flag from the start,” Kotzin said. “From creation, to environment setup, to rollout, agents can handle the whole lifecycle. We’re building a future where feature flagging is so deeply woven into the workflow that it becomes invisible.”

With this release, Flagsmith isn’t just removing friction. It’s laying claim to being the default runtime control layer for AI-native software development, where faster no longer has to mean riskier.

As we continue to invest in this new MCP Server as well as our powerful and user-friendly UI, teams will get to decide where execution, policy, and governance live—dictated only by their needs and ambitions.

The MCP server is available now. It can be installed locally with a single click, or deployed remotely for SaaS, private cloud, or self-hosted environments.

Try it out: https://app.getgram.ai/mcp/flagsmith-mcp/install