ChatGPT plugins for DevOps & Cloud
119 plugins, apps and MCP servers in DevOps & Cloud, ranked by official status, traction and completeness. Page 1 of 2.
In DevOps work, ChatGPT plugins are mostly used to investigate rather than to change things: explain a failed deploy, pull the logs and metrics behind an alert, find which release introduced an error, or look up the right Terraform provider or AWS API. Some can also act, by triggering deploys, changing environment variables or running cloud CLI commands, and that is where most of the care goes.
The listings split three ways. Plugins from platforms such as Vercel and Cloudflare bundle skills (deployment guidance, framework know-how) with a connection to your account. Apps such as Datadog and Render connect ChatGPT to one service. MCP servers cover most cloud providers and infrastructure tools. Many of them, including the AWS API, Azure and Grafana servers, run locally with your own credentials, so you add them in Codex rather than installing them from the Plugin Directory.
Before connecting production accounts:
- Use read-only modes where they exist. The AWS API server has a read-only setting, OpenAI's Sentry plugin is built around a read-only skill, and Google's BigQuery server has a separate read-only SQL tool.
- Use dedicated credentials. Give local servers a least-privilege IAM role, scoped API token or service account, not your admin profile.
- Keep confirmations on for anything that deploys, deletes or changes configuration. In Codex you can set approvals per server or per tool in
config.toml. - Check plan limits. Adding your own MCP server to ChatGPT needs developer mode, which is limited to Plus, Pro, Business, Enterprise and Education accounts.
Top picks for DevOps & Cloud
Vercel's plugin is one of the largest in the directory: skills for Next.js, the AI SDK, AI Gateway, deployments and CI/CD, environment variables, domains, caching and feature flags, plus the Vercel app connection. Use it to audit a repo for deployment risks or work out why a build fails. It suits teams already deploying on Vercel.
Cloudflare's plugin combines curated skills for Workers, Wrangler, Durable Objects, the Agents SDK and the Sandbox SDK with the official Cloudflare API MCP server. It helps you choose the right Cloudflare product and deploy a Worker. If you only need documentation lookups, the separate Cloudflare Docs server needs no authentication.
Datadog's app lets ChatGPT investigate observability data: chart metrics, logs, traces and RUM events, inspect dashboards and notebooks, and review incidents and monitors. It suits on-call engineers and anyone writing up an incident. For Codex, Datadog's MCP server offers the same data with selectable toolsets and requires
mcp_read/mcp_writerole permissions.Sentry's hosted server at
mcp.sentry.devsearches errors, events and traces and helps triage issues, with OAuth access that can be scoped to one organization or project. Ask it which errors spiked after a release. If you only want to inspect recent issues, OpenAI's Sentry plugin wraps a read-only skill.AWS Labs' server turns natural-language requests into validated AWS CLI commands and runs them with your local AWS profile, so IAM decides what it can do. It suits engineers who already work in the AWS CLI. Start in read-only mode with a limited profile; for docs and regional availability without credentials, use AWS Knowledge.
All 119 in DevOps & Cloud
AWS KnowledgeSearch current AWS docs, API references and regional availabilityMCP server · by Amazon Web Services · 9.7k stars · RemoteAWS APIRun validated AWS CLI commands from natural languageMCP server · by Amazon Web Services · 9.7k stars
KubernetesNative Go MCP server for Kubernetes and OpenShift clustersMCP server · by containers (Red Hat) · 2.1k starsDevOps & Cloud plugins: common questions
- Can ChatGPT deploy to AWS, Azure or Google Cloud?
Through MCP servers, yes. The AWS API server runs AWS CLI commands with your local profile, Microsoft's Azure MCP Server covers more than 45 Azure services using Azure Identity, and Google's gcloud CLI server runs gcloud commands with safety restrictions and is in preview. All three run locally, so you add them in Codex. Each acts with whatever credentials you give it, so use a least-privilege account.
- Which ChatGPT plugins work with Terraform and Kubernetes?
HashiCorp's Terraform MCP server searches the public Terraform Registry for providers, modules and policies, and can manage HCP Terraform or Terraform Enterprise workspaces, runs and variables with a token. For clusters, the community Kubernetes server from the containers project talks to the Kubernetes API directly and handles pods, resources, events and Helm. It can also exec into pods and install charts, so point it at a kubeconfig with limited rights.
- Is it safe to connect production infrastructure to ChatGPT?
Treat it like giving a new teammate access. Many incident questions only need read access to tools like Sentry, Datadog or Grafana. Use dedicated, scoped credentials for local servers, keep confirmations on for deploys and deletions, and in Codex set
default_tools_approval_modeor disable specific tools per server. Be wary of agents reading logs or tickets that contain text from outside your team. More in our security checklist.- Do DevOps MCP servers work in ChatGPT on the web?
Only the hosted ones. Servers with a public URL, such as Sentry or AWS Knowledge, can be added to ChatGPT through developer mode. Servers that run as a local command, including AWS API, Azure and Grafana, need Codex. Plugins from the directory, like Vercel and Cloudflare, are listed in both ChatGPT and Codex. Steps: Add an MCP server to ChatGPT.
Guide updated Oct 24, 2026.