Skip to main content

runtm deploy

Deploy your agent’s work to a live HTTPS URL. Your agent can use this URL to test, share, and iterate.

Arguments

Options

Examples

Basic Deployment

Machine Tiers

Auto-Fix Issues

Quick Config Changes

Non-Blocking Deploy

JSON Output for AI Agents

Output (each line is valid JSON):
On failure:

Deployment Flow

Redeployment

When you deploy a project with the same name, Runtime automatically redeploys:
  • URL stays the same
  • Secrets are preserved
  • Custom domains are preserved
  • Version number increments
This is perfect for agent workflows - your agent can deploy, see logs, fix bugs, and redeploy in a tight loop:

Force New Deployment

Use --new to create a completely new deployment:
--new will:
  • Create a new URL
  • NOT transfer custom domains
  • NOT transfer secrets

Machine Tiers

All tiers use suspended state - machines pause when idle and resume quickly on requests.
The web-app template requires standard or higher.

Config-Only Deploys

Use --config-only for fast updates that don’t change code:
--config-only fails if source code has changed since last deploy.

Validation

The deploy command validates your project:
  1. Manifest - runtm.yaml is valid
  2. Lockfile - Dependencies are locked
  3. Imports - Python imports work (optional)

Skip Validation

Force Re-validation

Output

Successful deployment shows:

Troubleshooting

Fix automatically:
Or manually:
Update .runtmignore to exclude:
Large dependencies can cause timeouts. Consider:
  • Using smaller base images
  • Caching dependencies
  • Removing unused packages
Ensure /health endpoint:
  • Returns 200 status code
  • Responds within 100ms
  • Works without authentication
Check pyproject.toml has runtime dependencies in [project] dependencies, not [project.optional-dependencies] dev.