Developer Tools & Devops · Launch darkly + Claude
LaunchDarkly MCP with Claude: Safer Feature Releases Through Plain-Language Control
The short answer
Claude works with LaunchDarkly through the Model Context Protocol (MCP), an open standard that lets Claude read and act inside the tool under permissions you set. Once connected, Claude can help create feature flags, adjust rollouts, manage environments and review flag usage from a plain-language request, while your team approves consequential changes.
What is Launch darkly?
LaunchDarkly is a feature management platform used by software teams to switch features on and off without redeploying code. Developers, product managers and release engineers use it to roll out changes to a small group of users first, test in different environments and turn off a problem feature in seconds. It is widely used by teams that ship software often and need to release with less risk.
How Claude connects to Launch darkly through MCP
The Model Context Protocol is an open standard that works like a universal adapter between Claude and other software. It lets Claude use a tool securely, reading information and taking actions, with permissions the organisation controls. For LaunchDarkly, the connection can run through an MCP server or a direct API connection, depending on what suits your setup.
Once connected, Claude can typically create a new feature flag, check which environments it exists in, adjust a percentage rollout or review how a flag has been used. A team member can simply ask in everyday language, for example to show which flags are still active in staging. Claude then carries out the request within the access granted to it.
Access is kept tight. We set up least-privilege permissions so Claude can only do what each role needs. Sign-in uses an API key or OAuth, as the tool supports. Consequential changes, such as switching a flag on in production, can require a named person to approve first, so people stay in charge of every release decision.
What you can automate
Creating flags for new features
A product manager describes a new checkout feature in plain words. Claude can set up a matching feature flag with a clear name and description, so developers can start building behind the flag. This keeps unfinished work hidden from customers until it is ready, and reduces the back-and-forth between product and engineering teams.
Staged rollouts with approval
A release engineer asks Claude to move a feature from 5 percent of users to 25 percent after checking the error rate. Claude can prepare the change and show the proposed update for sign-off before it takes effect. This gives teams a steady, controlled rollout and a clear record of who approved each step.
Fast switch-off during incidents
When a new feature causes trouble, an on-call engineer can ask Claude to turn that flag off in production. Because the action is simple and traceable, the team can recover quickly without waiting for a new deployment. Typical responses are quicker than a full rollback, which helps limit customer impact.
Environment housekeeping
Over time, old flags pile up across development, staging and production. Claude can list flags that have not changed for a long period or appear inconsistent between environments, and prepare a clean-up list for the engineering lead to review. This keeps the platform tidy and reduces confusion for new team members.
Audit questions for compliance and leadership
A compliance officer or engineering director may ask who changed a flag, when, and why. Claude can pull together the flag history and usage details into a short summary in plain language. This speeds up audits and reviews without requiring someone to search through dashboards manually.
An end-to-end workflow: Safe release of a new payments feature
TriggerA developer marks a payments feature as ready for testing in a Jira ticket.
Claude reads the ticket in Jira and creates a matching LaunchDarkly flag in the staging environment.
The QA lead asks Claude to enable the flag for an internal test group, and the change is logged.
After testing passes, Claude prepares a proposed rollout to 10 percent of production users and sends it to the release manager in Microsoft Teams for approval.
Once approved, Claude applies the rollout and posts a status update to the Jira ticket.
If error alerts rise, the on-call engineer asks Claude to switch the flag off, and the incident is recorded in the ticket.
OutcomeTeams release payments changes in controlled steps, with approvals recorded and a fast way to switch off a problem feature. Typical results include fewer manual updates between tools and a clearer audit trail for each release.
How we set it up
Free automation audit
We review how your teams release software today, including which LaunchDarkly environments and projects you use, and identify the workflows worth automating first.
Secure connection
We connect Claude to LaunchDarkly through an MCP server or direct API, using API key or OAuth sign-in as your setup supports, with least-privilege access for each role.
Workflow design
We design the end-to-end process across LaunchDarkly and your other systems, such as Jira, Microsoft 365 or your ticketing tool, and define which actions need human approval.
Testing with real cases
We test the workflow against past releases and incidents to confirm it behaves as expected, then adjust permissions and approval steps.
Launch and training
We launch with monitoring and audit logs enabled, and train your developers, release managers and operations staff to use Claude confidently in their daily work.
Who benefits
- Engineering leaders
- Gain a clearer view of release risk, with approvals and audit trails that support governance without slowing delivery.
- Release and DevOps engineers
- Make controlled rollouts and quick switch-offs in plain language, reducing manual steps during busy release windows.
- Product managers
- Set up and check feature flags without waiting for an engineer to run each step, speeding up experiments and launches.
- Compliance and audit teams
- Request flag history and change summaries quickly, making reviews faster and more consistent.
Security and control
- Least-privilege access, so Claude only reaches the projects and environments each role needs.
- Human approval for consequential changes, such as production rollouts or switch-offs.
- Audit logs of every action Claude takes, so changes can be traced and reviewed.
- Secure sign-in using API key or OAuth, as supported by the tool, with credentials managed by your IT team.
- Regular review of permissions and workflows as your team and release process change.
For teams in the US, UK, India and the Middle East
Organisations in the United States and the United Kingdom can host the Claude integration in a region that supports their data handling requirements. Companies in India must consider the Digital Personal Data Protection Act, 2023, and we help set up data flows accordingly. Organisations in the UAE and Saudi Arabia are subject to local data protection laws, and we design the workflow and hosting region to meet those requirements. Hosting region is chosen to meet each client's legal and contractual needs.
Learn more about Claude and MCP
Anthropic's AI model and assistant.
Secure links between Claude and business tools.
How engineers build and change software with Claude.
Hosted agents that run multi-step work.
The open standard Claude uses to connect to tools.
Frequently asked questions
What does the LaunchDarkly MCP server do?
The LaunchDarkly MCP server acts as a secure bridge between Claude and your LaunchDarkly account. It lets Claude read flag information and, where permitted, make changes such as creating flags or adjusting rollouts. It uses the access you grant, so Claude works only within the limits your team sets.
Is connecting Claude to LaunchDarkly secure?
Security depends on how the connection is set up. We use least-privilege permissions, supported sign-in methods such as API key or OAuth, human approval for important changes, and audit logs. Your IT team keeps control of credentials and can review or revoke access at any time.
What can Claude do with LaunchDarkly once connected?
Typically, Claude can create feature flags, check which environments they exist in, adjust percentage rollouts, toggle flags on or off, manage environments and review flag usage. Requests are made in plain language, and the exact actions available depend on the permissions you grant.
How long does setup take?
A simple setup for one team can often be completed in a few weeks, including the free audit, secure connection, testing and training. Larger setups with several projects, approval chains and connections to other systems usually take longer. We agree a timeline after the initial audit.
Which LaunchDarkly plan or permissions are needed?
The permissions needed depend on the actions you want Claude to take, such as reading flags or changing rollouts. Generally, the API key or OAuth account must have access to the relevant projects and environments. We check your current setup during the audit and advise on any changes required.