mcpField
L
lokei

lokei

github.com/ulpi-ioCloud & DevOps5132 installsUpdated 9w ago
Repository

About lokei

lokei is a workflow skill for using the lokei CLI to serve local projects over named HTTPS domains, create public tunnels, manage local environments, and diagnose networking issues. It emphasizes checking installation and setup health and getting approval before one-time setup or uninstall actions.

Key features

  1. 01Check CLI availability and setup health
  2. 02Run apps on named HTTPS domains
  3. 03Create public tunnels when requested
  4. 04Orchestrate Docker and multi-service projects
  5. 05Diagnose proxy and networking issues

Use cases

  1. 01Expose a local app over HTTPS for development
  2. 02Share a local service through a public tunnel
  3. 03Start a Docker Compose or multi-service environment
  4. 04Investigate local proxy or DNS setup failures

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/ulpi-io/skills --skill lokei

After installing, type/lokeiin Claude Code to confirm it loaded.

Bundled filesreferences/command-reference.md

Details

From the project README
`lokei` affects the local networking environment and can trigger setup flows.

Non-negotiable rules:

  1. Verify lokei availability before relying on it.
  2. Do not auto-run one-time setup or uninstall flows without approval.
  3. Resolve whether the user wants local-only HTTPS, a public tunnel, or multi-service orchestration.
  4. Prefer the simplest command that matches the job.
  5. Report the resulting URLs and diagnostic blockers clearly.

lokei

Inputs

  • $request: The local HTTPS, share, Docker, service, or diagnostics task

Goal

Use lokei to:

  • run a local project behind a named HTTPS .test domain
  • expose a local service through a tunnel when requested
  • orchestrate Docker or multi-service local environments
  • diagnose and explain local proxy/setup problems

Step 0: Verify installation and setup health

Check:

  • lokei --version
  • lokei doctor

If lokei is missing, tell the user:

lokei is not installed. Install with:

npm i -g lokei && lokei setup

lokei setup runs once -- generates a local CA, trusts it in the OS keychain, configures DNS for .test, and sets up port forwarding.

Do not install or run setup automatically. Wait for user confirmation.

If setup is unhealthy:

  • explain the failing component
  • common fixes: CA not trusted or DNS resolver missing -> re-run lokei setup
  • ask before suggesting or running one-time setup actions

Success criteria: lokei is available and its setup state is understood.

Step 1: Resolve the requested mode

Determine whether the user wants:

  • lokei run for one local app
  • lokei share or lokei run --expose for public access
  • lokei docker up for Docker Compose
  • lokei up for .test.yaml orchestration
  • lokei routes ls, inspect, or logs for diagnosis

Also resolve whether the project name, command, or port needs to be overridden.

Success criteria: The correct lokei mode is chosen before execution.

Step 2: Run the smallest correct command

Prefer:

  • lokei run for normal local development
  • lokei share --port <port> for ad-hoc sharing
  • lokei run --expose when the same session needs local HTTPS and a public tunnel
  • lokei docker up or lokei up only when the project structure actually calls for them

Avoid service-install and uninstall flows unless the user explicitly asked for those operations.

Success criteria: The executed command matches the requested workflow without overreaching.

Step 3: Report URLs and health clearly

After execution, report:

  • local .test URL
  • public tunnel URL if one was created
  • project name and route details if relevant
  • any doctor/setup blockers if the command failed

Success criteria: The user knows exactly where the app is reachable and what failed if it is not.

Guardrails

  • Do not auto-run lokei setup, service install, or uninstall without approval.
  • Do not guess that the user wants a public tunnel when local-only HTTPS is enough.
  • Do not hide diagnostics behind generic "setup failed" summaries.
  • Do not add disable-model-invocation; this is a normal environment workflow skill.

When To Load References

  • references/command-reference.md Use for the full command catalog, framework detection table, architecture details, .test.yaml examples, git worktree behavior, and scenario quick reference.

Output Contract

Report:

  1. mode chosen
  2. command run
  3. resulting local or public URLs
  4. any detected setup or networking blocker
  5. next action if user approval is required

Comments

Markdown supported