Set up your backend from your coding agent

Paste one prompt into your coding agent. It installs the Blocks skills, signs you in and sets up your project.

Read https://seliseblocks.com/bootstrap and follow it to bootstrap this repo, then help me sign in and get started with SELISE Blocks.

What your agent does with the prompt

The prompt points your agent at the Blocks bootstrap runbook. It follows the same steps every time, and asks you before anything that installs software or creates a project.

  1. Asks which agents you use

    So it writes files only for the coding agents on your team.

  2. Installs the skills and rules

    The Blocks skills go into .agents/skills/ and the rules into AGENTS.md. Pointer files such as CLAUDE.md are added only for agents that need them.

  3. Asks before installing the CLI

    It runs npm install -g @seliseblocks/cli-os@latest only after you say yes.

  4. Signs you in through your browser

    blocks login shows a device code that you approve in your browser, so the agent never sees your password.

  5. Picks or creates your project

    It selects an existing project with blocks use. Creating a new one means accepting the Blocks terms, so it asks you first.

  6. Scaffolds your app

    blocks new web creates a React app that is already wired to your project, with hosted login and route guards.

21 skills, one for each part of Blocks

Your agent reads the skill for the task in front of it, so it uses the Blocks CLI and SDK the way the docs describe instead of guessing.

Start1
blocks-bootstrap
Data4
blocks-data-gateway-configurationblocks-data-gateway-crudblocks-data-storageblocks-storage-configuration
Identity and access8
blocks-iam-accountblocks-iam-usersblocks-iam-access-controlblocks-iam-organizationsblocks-iam-mfablocks-iam-sso-oidc-configurationblocks-iam-sso-oidc-implementationblocks-captcha
Localization2
blocks-localization-configurationblocks-localization-implementation
Messaging3
blocks-mailblocks-notifierblocks-notification
Platform operations2
blocks-release-deploymentblocks-secrets
Local development1
blocks-frontend-local-https

Guardrails in every skill

A dry run before every change

Every command that changes your project runs with --dry-run first. The agent shows you the plan and runs --yes only after you approve it.

Your password stays with you

Sign-in uses a device code approved in your browser.

Nothing installed without asking

The agent asks before it installs the Blocks CLI.

Open source, reporting your choice

The skills are MIT licensed. Anonymous issue reports to the Blocks team are opt-in or opt-out, and your answer is saved in .agents/skills/.blocks-reporting.

Works with 13 coding agents

The skills are plain Markdown files, so any agent on this list can read them.

  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI
  • GitHub Copilot
  • OpenCode
  • Amp
  • Zed
  • Windsurf
  • Kilo Code
  • Kimi Code
  • Qwen Code
  • Kiro

Questions about building with coding agents

Which coding agents are supported?

Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode, Amp, Zed, Windsurf, Kilo Code, Kimi Code, Qwen Code and Kiro.

Does the agent see my password?

No. blocks login shows a device code that you approve in your browser, so the agent never handles your password.

What does the agent change in my repository?

It adds the Blocks skills under .agents/skills/ and the rules in AGENTS.md, plus pointer files such as CLAUDE.md for agents that need them. Selecting a project writes blocks.json, and scaffolding creates your app folder.

Can I do the same steps without an agent?

Yes. Install the CLI with npm install -g @seliseblocks/cli-os@latest, then run blocks login, choose a project with blocks use, and scaffold with blocks new web.

Are the skills open source?

Yes, under the MIT license, and the source is on GitHub.

Start with a free project

Sign up at os.seliseblocks.com and create a project, or paste one prompt into Claude Code, Codex or Cursor and let your agent set it up through the Blocks CLI.