Install on macOS
Agents with user approval can run the public installer directly; it opens the app and prints local MCP config.
curl -fsSL "https://clankercloud.ai/install.sh" | shMCP setup
Connect an agent to Clanker Cloud
Open Clanker Cloud on the machine that already has the cloud credentials, kubeconfigs, GitHub context, and provider settings you want the agent to use. The app exposes a local MCP server on one of the app ports.
Give the agent the local discovery command below. It curls the running app, reads the current MCP URL, then configures its MCP client from that local response instead of guessing a port or asking for secrets.
Tell the agent to discover Clanker Cloud through localhost, not through pasted cloud credentials or a public MCP bridge.
Copy for agents
Use the running app instructions endpoint
Direct macOS download
Agents that only need the current artifact can use this redirect without waiting for page JavaScript.
https://clankercloud.ai/download/latest/macosLocal discovery command
Use this exact curl loop in the agent shell on the same machine or remote session where Clanker Cloud is running.
for p in 8080 8081 8082 8083 8084; do
curl -fsS "http://127.0.0.1:$p/mcp/instructions" && break
doneConfig-only endpoint
Use this after discovering the active port when an MCP client wants only the server JSON.
curl -fsS "http://127.0.0.1:8080/mcp/config"MCP config shape
The discovered instructions return a concrete URL for the active port.
{
"mcpServers": {
"clanker-cloud-app": {
"url": "http://127.0.0.1:8080/mcp"
}
}
}Local only
The app-hosted MCP endpoint lives on localhost ports 8080 through 8084 and points to the running desktop backend.
Instruction endpoint
GET /mcp/instructions returns the current MCP URL, backend base URL, and the read-first agent workflow.
Config endpoint
GET /mcp/config returns a JSON MCP server entry for clients that only need configuration.
No secrets in chat
Agents should use saved Clanker Cloud settings and local profiles instead of asking the user to paste credentials.
Supported providers
Works across the environments teams already run
The current product positioning covers cloud providers, Kubernetes, GitHub, and bring-your-own AI keys from one local operating surface.
Supports ->AWSGCPAzureTencent CloudKubernetesCloudflareHetznerDigitalOceanVercelGitHubBYOK
Agent workflow
What the agent should do after connecting
Confirm tools
Reload the MCP client and confirm the clanker-cloud-app tools are visible before answering.
Check account state
Call clanker_cloud_account_status first. If no account is ready and the user asked you to install/use Clanker Cloud, call clanker_cloud_create_account.
Run setup check
Call clanker_cloud_setup_check for app, account, provider credential presence, local tools, dependency install commands, redacted settings, and next actions.
Use hosted LLM by default
Hosted Clanker Cloud LLM is the agent default after account login/account creation. Call clanker_cloud_configure_llm with provider clanker-cloud and test true when you need to verify it.
Install needed dependencies
Install only the missing provider CLIs the user needs, such as aws, gcloud, az, kubectl, terraform, gh, wrangler, doctl, or hcloud, then rerun setup check.
Start scanning and wait
Call clanker_cloud_start_scan once setup status is clear and wait for it to finish. Use mode headless for a fresh backend scan, or mode ui when the user wants the visible app to repaint.
Run research and security
Call clanker_cloud_deep_research for estate-wide findings and clanker_cloud_security_scan for exposure, auth gap, credential leak, and misconfiguration findings only after the scan returns.
Use high-level K8s tools
For cluster work, start with clanker_cloud_k8s_investigate and only drill into narrower tools when evidence is missing.
Stay read-first
Treat write, delete, deploy, and apply operations as user-approved actions with an explanation before the call.
Local endpoints
Routes exposed by the running app
| Route | Purpose | When to use it |
|---|---|---|
| /mcp/instructions | Markdown bootstrap instructions with the active MCP URL. | Give this to any agent that can curl localhost. |
| /mcp/config | JSON MCP server config for the active app port. | Use for scripts or MCP clients that only need server config. |
| /mcp | Streamable HTTP MCP transport. | Configure the MCP client with this URL after discovery. |
| /api/ping | Backend liveness check. | Use only to confirm the app backend is reachable. |
Reference
Read the details behind the setup
MCP infrastructure Q&A
The previous MCP product page now lives here for deeper positioning and FAQ.
Local credentials boundary
How credentials, model-provider payloads, logs, telemetry, BYOK, and MCP localhost access are bounded.
Claude Code MCP example
A concrete app-first workflow for giving an agent live Kubernetes context.
FAQ
Common questions
Is clankercloud.ai/mcp itself the MCP server?
No. This page is the public setup guide. The actual MCP server is exposed locally by the running Clanker Cloud desktop app at http://127.0.0.1:<port>/mcp.
What should I tell Codex or another agent?
Tell it to find Clanker Cloud on localhost ports 8080-8084 and curl /mcp/instructions. The running app returns the exact MCP URL and workflow.
Does the agent need cloud credentials pasted into chat?
No. The agent should use the local Clanker Cloud MCP tools and the saved settings or profiles already available to the app.
Next step
Ready to connect an agent?
Open the desktop app first, then hand the local discovery command to the agent running on the same machine or remote session.