AI & Machine Learning · Modelry + Claude
Modelry MCP with Claude: Deploy, Monitor and Manage ML Models Through Plain Language
The short answer
Claude connects to Modelry through the Model Context Protocol (MCP), an open standard that acts as a secure adapter between the two. Once connected, Claude can read training runs, datasets and performance metrics, and carry out approved actions such as deploying a model, all through natural-language requests. Access is controlled by your organisation, so people stay in charge of consequential changes.
What is Modelry?
Modelry is a platform for building, deploying and managing machine learning models. It supports the full lifecycle, from preparing datasets and running training to releasing models and tracking how they perform. Data science teams, machine learning engineers and technology leaders use it to ship models faster and more reliably. Organisations that run several models in production often use it to keep their work organised and visible.
How Claude connects to Modelry through MCP
The Model Context Protocol (MCP) is an open standard that lets Claude use other software securely, reading information and taking actions with permissions your organisation controls. Think of it as a universal adapter between Claude and the tool. For Modelry, the connection is made using an API key, so access is tied to a credential your team issues and can revoke at any time.
Once connected, Claude can typically retrieve information such as the status of training runs, the datasets available to a project, and real-time performance metrics for deployed models. Staff can ask questions in everyday language, such as which model's accuracy dropped this week, and receive a clear summary. Claude can also prepare actions, for example starting a deployment or updating a dataset, using the capabilities the connection allows.
Permissions and approvals keep people in charge. Access is limited to the actions your team chooses to allow, and consequential steps such as releasing a model to production can require a named person to approve before anything happens. Activity is recorded, so every request and action can be traced back later.
What you can automate
Daily model health check
A machine learning lead asks Claude each morning for a summary of every model in production. Claude pulls the latest performance metrics from Modelry and highlights any model whose accuracy or response behaviour has moved outside the expected range. The team starts the day knowing exactly where to look, rather than opening several dashboards.
Training run follow-up
An engineer launches a long training job and wants to know when it finishes and how it compares with the previous version. Claude checks the run's status and compares its results with earlier runs, then reports the outcome in plain language. This saves time spent refreshing pages and comparing spreadsheets by hand.
Dataset readiness review
Before a new project starts, a project manager asks which datasets are current, complete and approved for use. Claude reviews the datasets available to the project and lists those that are ready, outdated or missing key information. Decisions about which data to use stay with the data owners.
Controlled model release
A product team wants to move a tested model into production. Claude prepares the release details and summarises the test results for review. A named approver confirms the release before Claude carries out the deployment, so no model goes live without sign-off.
Stakeholder reporting
A department head needs a simple update on how the company's models are performing for a monthly meeting. Claude gathers the key metrics and turns them into a short, readable summary without technical language. The report is accurate, quick to produce, and easy for non-specialists to follow.
An end-to-end workflow: Monitored model release with ticket tracking
TriggerA data science team finishes testing a new version of a customer churn prediction model.
The team member asks Claude to compare the new model's test results with the current production version.
Claude retrieves the relevant training runs and performance metrics from Modelry and prepares a comparison summary.
Claude logs a release request in the organisation's ticketing system, attaching the summary and the test results.
The engineering lead reviews the request and approves the release in the ticket.
Claude carries out the approved deployment in Modelry and confirms the new version is live.
Claude checks performance metrics over the following days and posts any unusual changes to the ticket.
OutcomeReleases move forward with a clear record of testing, approval and post-launch monitoring. Teams spend less time gathering information and reviewers can see the full history in one place.
How we set it up
Free automation audit
We review how your teams use Modelry today and identify the workflows most worth automating, such as monitoring, dataset checks or release preparation. There is no cost and no obligation.
Secure connection
We connect Claude to Modelry using an API key, issued by your team and stored securely. The key is scoped to the access you approve, so Claude can only do what you allow.
Workflow design
We design the end-to-end workflow across Modelry and your other systems, such as Microsoft 365, your CRM, ERP or ticketing tool. We decide together which actions run automatically and which need a person's approval.
Testing with real cases
We test the workflow against past training runs, datasets and releases from your own history. This shows how Claude handles your real situations before anything goes live.
Launch, monitoring and training
We launch with monitoring and audit logs in place and train your team to use the new workflow confidently. We stay available to adjust the setup as your needs change.
Who benefits
- Machine learning engineers
- Spend less time checking dashboards and more time improving models, with training results summarised on request.
- Data science leads
- Get a clear view of model health across projects, with release decisions backed by a traceable record.
- IT and platform managers
- Keep control of access through scoped API keys and audit logs, while letting teams use AI in daily work.
- Business and operations heads
- Receive plain-language updates on how models are performing, without needing technical knowledge.
Security and control
- Access is limited by an API key scoped to only the actions your team approves, and can be revoked at any time.
- Consequential actions such as production deployments can require a named person to approve first.
- Every request and action is logged, giving you an audit trail for review.
- Claude works from the data and permissions you provide and does not take actions outside the connection you set up.
- We test thoroughly against real cases and launch with monitoring in place before the workflow is relied upon.
For teams in the US, UK, India and the Middle East
Organisations in the United States and the United Kingdom can host the solution in their own region to meet local data handling expectations. In India, we design around the Digital Personal Data Protection Act, 2023, so personal data is handled lawfully and transparently. Organisations in the Middle East, including the UAE and Saudi Arabia, can follow their local data protection laws, and we choose the hosting region to meet those requirements.
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 Modelry MCP server do?
The Modelry MCP server is the bridge that lets Claude talk to Modelry in a standard, secure way. It allows Claude to read information such as training runs, datasets and performance metrics, and to carry out actions you have approved. It acts as a universal adapter, so the connection works without custom coding for each request.
Is connecting Claude to Modelry secure?
Security depends on how the connection is set up, and we design it carefully. Access uses an API key that your team issues and can revoke. Permissions are limited to the actions you approve, consequential steps can require human sign-off, and all activity is logged so you can review it.
What can Claude do with Modelry once connected?
Claude can typically check the status of training runs, review datasets, retrieve real-time performance metrics and summarise results in plain language. It can also prepare or carry out approved actions, such as starting a deployment, within the limits you set.
How long does setup take?
Timing depends on how many workflows you want to automate and how many systems they touch. A focused connection to Modelry can often be set up and tested in a few weeks. We confirm the timeline during the free automation audit, once we understand your needs.
What access or plan is needed to connect Modelry?
In general, you need an API key from your Modelry account with permissions suited to the workflows you plan to automate. The exact level of access depends on your account arrangements, so we confirm what your team needs during the audit and recommend the least-privilege setup.