Agent onboarding
Clanker Cloud supports local-first onboarding for AI agents. The agent can create a Clanker Cloud app account and connect to the running desktop app, while provider credentials stay on the user's machine.
macOS bootstrap
On macOS, an agent can install and open the desktop app with:
curl -fsSL "https://clankercloud.ai/install.sh" | shIf the user wants an email attached for later claiming, pass it as local environment data:
CLANKER_USER_EMAIL="user@example.com" curl -fsSL "https://clankercloud.ai/install.sh" | shThe script downloads the latest macOS release, opens the app, waits for the local backend, and asks the local app to create a Clanker Cloud agent account.
By downloading, installing, opening, creating an account, or using Clanker Cloud, the user agrees to the Clanker Cloud Terms of Service and Privacy Policy:
What goes to Clanker Cloud
The cloud auth service receives only Clanker Cloud account data:
- generated agent user id
- optional human email for later claiming
- account kind
agent - terms version and account metadata
- app auth token records
It does not receive AWS, GCP, Azure, Cloudflare, Vercel, Railway, database, kubeconfig, or LLM provider secrets.
Provider setup
Agents should run the local MCP setup check after connecting:
clanker_cloud_setup_checkThe setup check reports:
- app and backend status
- MCP status
- Clanker Cloud account state
- bundled
clankerCLI status - local provider CLI availability
- dependency install commands for macOS, Linux, and Windows
- credential presence
- redacted saved settings
- exact next actions
Secret-looking values are reported as configured or missing. Raw secret values are not returned.
First-install dependencies
On a fresh machine, agents should use the setup doctor as the dependency contract. The tools and dependencies sections say which binaries are missing, which providers they support, how to install them on the current OS, how to verify them, and where the vendor docs live.
Install only what the user needs for their providers:
| Provider or workflow | Common dependency |
|---|---|
| AWS, EKS | aws CLI |
| GCP, GKE | gcloud CLI |
| Azure, AKS | az CLI |
| Kubernetes | kubectl |
| Terraform or maker/deploy flows | terraform |
| Local image builds | docker |
| GitHub and GitHub Actions | gh CLI |
| Cloudflare | wrangler |
| DigitalOcean | doctl |
| Hetzner | hcloud |
| Supabase | supabase CLI |
| Vercel | vercel CLI |
| Railway | railway CLI |
After installing a missing dependency, rerun clanker_cloud_setup_check before asking Clanker Cloud to inspect or modify that provider.
Agent workflow
- Install/open Clanker Cloud.
- Create or import the agent account through the local app.
- Discover MCP at
http://127.0.0.1:{8080-8084}/mcp. - Call
clanker_cloud_setup_check. - Install missing provider dependencies from the doctor output only for providers the user wants to use.
- Ask the user to run local login commands or open local wizard panels for missing providers.
- Call
clanker_cloud_startup_snapshotafter setup status is clear. - Keep write, delete, deploy, and apply operations user-approved.
Local credential commands
Use local provider flows instead of collecting secrets in chat:
aws sso login
gcloud auth application-default login
az login
gh auth loginProvider tokens that cannot be discovered through CLI login should be entered in the Clanker Cloud desktop app settings or exported locally before the app starts.