Use Elula with your AI

AI coding agents can deploy to Elula for you. They run the same elula CLI you would, in your project folder. You paste one prompt, sign in once in the browser, and the agent gives you back a live URL.

This works with any agent that can run terminal commands, such as Claude Code, Codex, Cursor's agent and Antigravity.

The short prompt

Open your agent in the project folder and paste:

Deploy this to Elula. Install the CLI with `curl -fsSL https://elula.com/install | bash`, run `elula login`, then `elula init` and `elula deploy --no-wait`. A deploy takes a few minutes: run `elula status --wait 60` until it says Live, and tell me which step it's on each time. If anything is unclear, run `elula docs` first. Give me the live URL when it's done.

The step-by-step prompt

The dashboard shows a longer prompt when your workspace has no apps yet. It spells out each step, which helps agents that tend to improvise:

Deploy this project to Elula.

1. Install the CLI: curl -fsSL https://elula.com/install | bash
2. Run `elula login` — a browser opens; I'll sign in with my Google account.
3. Run `elula init` to create the app (accept sensible defaults; add --database if this project needs Postgres).
4. Run `elula deploy --no-wait` to start the deploy. It takes a few minutes.
5. Run `elula status --wait 60` until it says Live. After each run, tell me which step it's on and how long it has taken.
6. When it's live, give me the URL.

If anything is unclear, run `elula docs` for the full platform reference before asking me.

What happens

1. The agent installs the CLI. It may need your password for sudo.
2. The agent runs elula login. A browser window opens. You sign in with Google. This is the only step that needs you.
3. The agent runs elula init, which creates the app in your active workspace and writes .elula.json.
4. The agent starts the deploy with elula deploy --no-wait, then checks on it with elula status --wait 60 and tells you each step: queued, building the image, updating Cloud Run, live. A first deploy usually takes 3 to 6 minutes.
5. The agent reports the URL.

Why agents shouldn't wait on one long command

Most agents show nothing while a command runs. If the agent runs elula deploy as one command (or pipes it through tail), you see "Running…" for several minutes with no idea what's happening. elula status --wait 60 returns at least once a minute with the current step, so the agent can keep you posted.

When elula deploy's output isn't a terminal, it prints only the steps, a "still building… 2:30" line every 30 seconds, and, if the build fails, the end of the build log. Add --logs to get the full build log.

Give your agent the full reference

elula docs prints a plain-text reference written for AI assistants. It covers app types, monorepos, jobs and schedules, environment variables, logs, databases, deployments, access, scaling and config files. Agents can read it on their own when the prompt tells them to.

elula docs          # print it
elula docs --copy   # copy it to the clipboard, to paste into a chat

You must be logged in to run elula docs.

If your agent supports project instructions (for example CLAUDE.md for Claude Code or AGENTS.md for Codex), you can save the output there so every session starts with it:

elula docs > ELULA.md

Commands that work well for agents

Most read commands take --json, so an agent can parse the result instead of reading tables:

elula status --json
elula deployments --json
elula http-logs --status 5xx --json
elula metrics --hours 6 --json
elula health --json
elula env list --json

Some commands return exit codes an agent can check:

• elula deploy exits with code 1 when the deploy fails.
• elula status --wait 60 exits with code 1 when the latest deploy failed, and 0 when it's live or still running (the output says which). elula status --json includes deploy_progress with the step, done, elapsed_seconds and typical_seconds.
• elula job run exits with code 1 when the run fails or is cancelled.
• elula health exits with code 1 when the app is down.

Commands that ask for confirmation have a -y or --yes flag for non-interactive use, for example elula rollback -y and elula project delete --yes.

Example requests

Once an app is linked, you can ask your agent things like:

• "Why is the app returning 500s?" The agent can run elula http-logs --status 5xx and elula runtime-logs --search error.
• "The last deploy broke something. Roll it back." The agent can run elula rollback.
• "Add a Postgres database." The agent can run elula db provision, then elula deploy.
• "Set the Stripe key." The agent can run elula env secret STRIPE_KEY=....
• "Run this job every weekday at 9am Kuala Lumpur time." The agent can run elula job schedule add --cron "0 9 * * 1-5" --timezone "Asia/Kuala_Lumpur".

Limits to keep in mind

• The agent acts as you. It can only do what your workspace role allows.
• Private apps need a verified team domain. Until an owner verifies one, new apps are public. See Private and public apps.
• Making an app public may need an owner or admin to approve it. The agent can file the request with elula access request-public.
• Your workspace needs an active subscription to create apps and deploy. See Billing.

See the Command reference for every command.