6. Terraform MCP-Server
Despite alternatives, HashiCorp Terraform is still the first address when it comes to Infrastructure as Code (IaC). Accordingly, the official MCP server is an interesting option for generating and managing Terraform configurations via AI agents. The MCP server can be integrated into both the registry APIs and Terraform’s Enterprise/HCP services. This enables AI agents to:
- query module and provider metadata,
- check the status of workspaces, and
- Trigger tasks (with human permission).
A command like “Generate Terraform code for a new run” could do this create_runoperation, whereupon the AI agent validates and schedules the configuration before applying it.
The Terraform MCP server ships with the readme file AGENTS.md. This makes it easier for agents to interpret tools. Currently (as of December 2025) the Terraform MCP server is only available for local use. It is expressly not intended for remote or hosted deployments.
7. GitLab MCP-Server
The GitLab platform also provides an MCP server to its Premium and Ultimate customers. This is currently in beta and enables AI agents to collect project information and securely execute operations via GitLab APIs.
The GitLab MCP server allows some status changes, such as creating issues or merge requests. The other functions are mainly used to query data – for example, to query information about issues, merge requests, commits, diffs and pipelines. A general search tool is also included.
The GitLab MCP server documentation is very detailed and contains numerous examples of natural language expressions that can be processed. The server also supports dynamic client registration via OAuth 2.0.
8. Snyk MCP-Server
Snyk offers a security platform for developers – and an MCP server. This can be used to scan for vulnerabilities and fix them with the help of AI agents (IaC) code, open source dependencies, containers and SBOMs or AIBOMs. Integrating the Snyk MCP server is therefore suitable for carrying out security scans automatically as part of a CI/CD workflow with AI agents. These scans can even be coordinated across other MCP servers, for example by retrieving repository details via the GitHub MCP server before starting a Snyk scan.
A prompt like “Scan the Authentication microservice repo for vulnerabilities” could instruct an agent to locate the repository using GitHub MCP and then Snyk tools like snyk_sca_scan or snyk_code_scan to identify known vulnerabilities, leaked credentials, and other risks.
This MCP server runs locally and uses the Snyk CLI to execute commands like these via authenticated API calls. The company does not offer a hosted remote version of its MCP server.
9. AWS MCP-Server
Cloud hyperscalers have been particularly busy working to quickly spin up MCP servers that integrate into their ecosystems. For example, Amazon Web Services (AWS) has introduced dozens of specialized MCP servers that enable AI agents to interact with all types of AWS services. Some of these are offered as fully managed services, while others can only be run locally.
- For example, AI agents can list and call Lambda functions via the Lambda Tool MCP server.
- The AWS S3 Tables MCP server, on the other hand, can be used by an agent to query S3 buckets or create new tables from CSV files.
- The AWS Knowledge MCP server connects agents to the latest AWS documentation, API references, and architecture guides.
A query to the latter knowledge server could contain the instruction to call an API reference for the Prometheus tool managed by AWS. This would provide the right up-to-date information – optimized for use by AI agents.
10. Pulumi MCP-Server
Pulumi is another popular IaC option – and has also introduced an official MCP server. This enables AI agents to
- to query Pulumi registries,
- access cloud resources and infrastructure, and
- Execute Pulumi commands.
Pulumi explains in detailed step-by-step instructions how developers can use this MCP server to provide an Azure Kubernetes Service (AKS) cluster. (fm)
This article originally appeared at our sister publication Infoworld.com.
