Build
Use Install the agent's pinned dependencies. Commit the lockfile and pin the runtime version so a rebuild installs the same dependency graph.
Review deployment docs →AI Agent · Deployment guide · Hosting
Deploy AI Agent with the runtime made explicit.
Custom agents, tool-using assistants, workflow engines, RAG applications, and self-hosted agent platforms. Darwa AI connects the repository, runs Install the agent's pinned dependencies, deploys the result on always-on agent service, and keeps domains, HTTPS, releases, logs, and billing in the same workspace.
- Always-on agent service
- Deployment model
- Install the agent's pinned dependencies
- Build command
- Agent application plus persistent data
- Expected output
- AI Agent Hosting
- Recommended Darwa product
What a production AI Agent deployment needs.
These are operational requirements, not a generic framework feature list. Confirm them in the repository before the first release.
Start
Production start: Run the documented production entry point. The command must stay in the foreground and report a failed boot with a non-zero exit code.
Explore AI Agent Hosting →Artifact
Expected deployable output: Agent application plus persistent data. Generated artifacts should come from the build rather than being edited after release.
Configure the build →Configuration
Store secrets and environment-specific values in the service environment. Do not commit production credentials or bake them into frontend bundles.
Manage configuration →Health and logs
Expose a lightweight health endpoint, keep startup logs actionable, and verify the public endpoint after every release.
See runtime operations →Known boundary
Persist memory outside the container, scope tool credentials narrowly, require approval for destructive actions, and set resource limits before unattended operation.
Check a workload with Darwa →From AI Agent repository to a repeatable release.
A successful first deploy is not enough. The release must be reproducible, observable, reversible, and safe to scale.
Connect source01
Select the repository and production branch. Keep preview work isolated from the production service.
Detect and confirm02
Confirm always-on agent service, the build command, output, and production start command instead of relying on a lucky default.
Add environment03
Set secrets and public configuration with the correct build-time or runtime scope.
Deploy and verify04
Watch the build, startup, and health checks; then exercise a real application path rather than only the homepage.
Attach the domain05
Point the domain after the generated endpoint is healthy. Managed HTTPS should complete before traffic is moved.
Plan rollback06
Keep the previous successful release available and make database changes backward-compatible before scaling replicas.
Products, plans, and regional guidance.
AI Agent HostingThe recommended Darwa AI product for this AI Agent deployment.→PricingCompare current plan resources and renewal pricing.→Web ServicesAlways-on APIs and server applications.→Frontend CloudStatic build output on global delivery.→Managed databasesPostgreSQL, MySQL, Redis, and compatible data services.→All frameworksCompare the deployment model for another framework.→
Answers before you deploy.
Can Darwa AI host AI Agent?+
Yes. Custom agents, tool-using assistants, workflow engines, RAG applications, and self-hosted agent platforms. The recommended starting product is AI Agent Hosting, subject to the runtime boundary described on this page.
What build command should I use for AI Agent?+
Start with Install the agent's pinned dependencies. Keep the dependency lockfile committed and test the same command in a clean environment before deployment.
What starts AI Agent in production?+
Use Run the documented production entry point. A server deployment must listen on the assigned port, stay in the foreground, and exit when startup fails.
Can I connect a custom domain?+
Yes. Deploy to the generated endpoint first, verify health, then connect the domain and wait for managed HTTPS before moving traffic.
Where should secrets be stored?+
Use the service environment and keep production secrets out of Git, build logs, browser bundles, and image layers.
How do I deploy without downtime?+
Use health checks, backward-compatible data changes, a verified new release, and a retained previous release that can be restored quickly.