Clanker CLI
A Go CLI: plain-language questions to AWS, GCP, Azure, Cloudflare and Kubernetes, change plans and its own MCP server
- 380 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- In maker mode the CLI creates real cloud resources and, with `--destroyer`, deletes them
- Access comes with the permissions of local provider and cluster profiles
- A model drafts the change plan, so it must be read before applying
Install
- Works in
Manual install
brew tap clankercloud/tap && brew install clankerThe Homebrew build can lag; `make install` from the repository gives the latest (Go is required).
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- In maker mode the CLI creates real cloud resources and, with `--destroyer`, deletes them
- Access comes with the permissions of local provider and cluster profiles
- A model drafts the change plan, so it must be read before applying
This is third-party code. Review the repository files before installing.
What it does
Clanker is an open-source CLI for working with cloud infrastructure, part of the Clanker Cloud ecosystem. The `clanker ask` command takes a plain-language question and answers from the data of connected providers: AWS, GCP, Azure, Cloudflare, Kubernetes, DigitalOcean, Hetzner, Tencent Cloud, Fly.io and Verda. For a quick resource listing with no model involved there are static commands such as `clanker aws list resources`. Changes go through maker mode. The `--maker` flag asks the model for an action plan as JSON, `--apply` executes an approved plan and `--destroyer` allows destructive operations. When applying, the CLI retries safe operations, waits for async steps such as CloudFormation and may adjust an AWS CLI command on common errors. Besides questions, `clanker security` runs a protection check on the data the providers expose, `clanker sre` starts a light SRE bot with environment discovery, and `clanker onboarding scan` finds missing provider CLIs and suggests installs. The `clanker mcp` command raises its own MCP server over stdio or HTTP: an agent gets the version, question routing, running clanker commands and `clanker_k8s_*` tools for Kubernetes clusters.
Who it is for. For DevOps engineers and SREs who question and change AWS, GCP, Azure and Kubernetes clusters from a terminal or MCP client and are ready to review plans before applying them.
Good fit when
- You want to quickly learn which lambdas, databases or services an account holds without recalling cloud CLI commands
- You need a plan to create a small environment, such as an instance and a database, and review it before applying
- You need a static resource list of a provider with no model call, for example for an inventory
- An agent in an MCP client should answer questions about a Kubernetes cluster through `clanker_k8s_*` tools
- You want a protection summary: public APIs, IAM permissions and places without authorization
Not a fit when
- You must delete and change resources in a production cloud with no manual plan review: the model can err and `--destroyer` allows deletion
- You need infrastructure as code with review and change history: Clanker runs provider commands from a model's plan
- No provider profiles are configured: the CLI uses local profiles, SSO and keys and grants no access itself
- You only need the Clanker Cloud browser workspace: installing this CLI is not required for it
Example request
Show which lambdas and RDS instances exist in the dev profile, then draft a plan for creating a small EC2 instance without applying it.Limitations
You need Go to build from source and the CLIs of the providers you work with. For AWS the CLI uses local profiles, preferably through SSO, and keeps no keys in its own config. The model comes from `~/.clanker.yaml` or from the `OPENAI_API_KEY`, `GEMINI_API_KEY` and `COHERE_API_KEY` variables. Without a config file the default is OpenAI with the gpt-5 model.
Installing through Homebrew can lag behind the repository, and the author suggests building from the master branch with `make install` and updating with `clanker update`. Some features, such as deploying `clanker cloud apps`, need a CLANKER_CLOUD_API_KEY from the Clanker Cloud platform.
Change commands create and delete real resources, and the model can err. The `clanker sre` bot sends discovery and heartbeat events to Cerebro of the Clanker Cloud platform. Availability of cloud services and model APIs from Russia was not checked when this card was prepared.
How to disable. Remove the CLI with `brew uninstall clanker`, or delete the binary if you built from source, then delete `~/.clanker.yaml`. If you connected MCP, remove the clanker block from the client config. Resources created in the cloud are removed with the provider's own tools.
FAQ
How do I install Clanker?
Through Homebrew: `brew tap clankercloud/tap`, then `brew install clanker`. That build can lag, so building from source is safer: `make install`. `clanker update` upgrades it, and the main channel for fresh commits turns on with `clanker config init --update-channel main`.
How do I connect Clanker to Claude Code as an MCP server?
Run `clanker mcp --transport stdio` as an MCP client command, or `clanker mcp --transport http --listen 127.0.0.1:39393` for HTTP. The agent gets tools for the version, question routing, running clanker commands and questions to Kubernetes clusters.
Can Clanker change my infrastructure by itself?
Only in maker mode. The `--maker` flag builds a plan as JSON, `--apply` applies an approved plan and `--destroyer` allows destructive operations. Nothing is created without `--apply`, so you can review the plan calmly.
Which providers can Clanker read?
AWS, GCP, Azure, Cloudflare, Kubernetes, DigitalOcean, Hetzner Cloud, Tencent Cloud, Fly.io and Verda Cloud, plus Linear and Notion integrations. `clanker onboarding scan` shows which providers are already found on the system.
Do I need a Clanker Cloud account to use the CLI?
Not for questions and plans: a model key and local provider profiles are enough. An account and the CLANKER_CLOUD_API_KEY key are needed only for platform features such as hosting apps with `clanker cloud apps`.
Discussion
No questions or reviews yet.
Related
CLI toolsAn AI assistant for Kubernetes that turns plain-language requests into kubectl operations and doubles as an MCP server
A Kubernetes panel with an MCP server: Claude Code reads pod logs, edits resources and rolls back releases in connected clusters.
A CLI and Kubernetes controller for developing, building and deploying MCP servers. For platform teams, written in Go
clawk is a virtual-environment CLI that snapshots a Linux machine’s memory and disk so running processes can be resumed later.
An MCP server and CLI that lets an agent query and provision cloud, SaaS and API resources with plain SQL
A Go MCP server for Kubernetes and OpenShift: resources, pods, logs, Helm and Tekton straight through the cluster API
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Clanker (this card) | no | 380 | ||
| kubectl-ai | yes | not required | 7,598 | an assistant for Kubernetes only that turns requests into kubectl operations |
| Pulumi Agent Skills | no | 78 | official Pulumi skills for writing infrastructure as code |
README in short
The README is very detailed: install, configuration, per-provider examples and separate chapters on Kubernetes, DigitalOcean, Hetzner, Tencent, Fly.io and Verda. The opening covers the link to the Clanker Cloud platform and app deployment, followed by MCP, the protection check, the SRE bot and the logic of applying plans. It ends with debugging tips.