AI Fabrix - Builder
Open developer runtime for AI Fabrix ecosystem. Install the AI Fabrix platform and test it locally. Then add external integrations or build your own applications.
← Full documentation: docs/README.md (table of contents for all guides)
Contributor quick start (miso ↔ dataplane ↔ builder): QUICK-START.md
Why AI Fabrix Builder?
- Build perspective: Everything is driven by declarative config and JSON schemas—no hidden logic, AI assistant–friendly.
- Industry standards and security: Follow industry standards and high security (e.g. ISO 27k); no secrets in version control.
- Full lifecycle in your version control: Configuration, apps, and integrations live in your own VCS (GitHub, GitLab, Azure DevOps).
- One tool from day one: Single CLI for local infra, app and integration creation, build, run, and deploy—same workflow for apps and integrations.
- Consistency and production readiness: Schema-driven; deploy apps and integrations to the same controller/dataplane; production-ready secrets with
kv://and Azure Key Vault. - Application development: Use miso-client (TypeScript and Python) to talk to the dataplane and controller. The repo includes both TypeScript and Python SDKs; see templates/applications/dataplane/README.md and the repo for usage.
Prerequisites
- Node.js 22+ – Supported CLI floor; Node.js 24.22.1 is the recommended current runtime.
- Docker Desktop – Full access to Docker (containers for infra and platform).
- Secrets before platform: Add an AI tool key (OpenAI or Azure OpenAI) in setup or with
aifabrix secret setbefore the platform needs them. See Infrastructure.
Install the CLI
npm install -g @aifabrix/builder
Alias: You can use af instead of aifabrix in any command.
Customer integration repositories can install the Agent Kit (Cursor, Codex, and Claude Code) with:
aifabrix agent-kit install
See Agent Kit commands.
Goal 1: Install the platform (one command)
Get Postgres, Redis, Keycloak, Miso Controller, and Dataplane running on your machine.
aifabrix setup
That is the whole story for most developers. The command detects whether infra is already running:
| Situation | What happens |
|---|---|
| Fresh machine | Short wizard → pull images → up-infra → up-platform |
| Already installed | Mode menu (3 choices): re-install (destroys volumes), wipe data (DBs only), or update images (keeps data) — each path pulls fresh Docker images before restart |
Grab a coffee — setup pulls infrastructure and platform images (a few minutes on first run or after image updates). When setup finishes, the CLI stores platform-controller in config.yaml (absolute Miso URL for your machine). Sign in with username admin and your admin password. If you were already logged in to that controller URL, setup keeps your session; otherwise it prompts for device login like a fresh install.
Controller URL: Local Docker without Traefik uses http://localhost: + port 3000 + developerId×100. With Traefik and remote-server, use the public URL shown in the Platform Ready footer (path /miso/*). See Infrastructure guide.
→ Step-by-step detail, ports, and dev / pro installation profiles: Infrastructure guide
Dev vs pro installation (setup profile)
These names match how we talk about environments elsewhere (dev, tst, pro). Here they mean how you install locally, not which cloud environment you deploy to later.
| Profile | Best for | Admin email | Admin passwords |
|---|---|---|---|
| dev (default) | Local development, training, labs | You enter once | One password for Postgres, pgAdmin, Keycloak, and platform login |
| pro | Production-style or hardened installs | Same | Autogenerate strong passwords (shown once) or enter manually |
Default: dev (one email, one password). Pro is available in the setup wizard or non-interactively:
aifabrix setup --installation pro --pro-password-mode autogen
Use --pro-password-mode manual to enter passwords yourself (one password for all roles, or separate infra / Keycloak / platform passwords via up-infra flags).
Admin credentials are stored in ~/.aifabrix/admin-secrets.env (encrypted when you configure a secrets encryption key). App and integration secrets stay in ~/.aifabrix/secrets.local.yaml.
Configure an LLM provider
aifabrix setup does not collect LLM credentials. Store the selected provider value
under its standard address in your local or shared secrets store:
aifabrix secret set openai/apiKey <your-openai-secret-key> --shared
The same pattern applies to azure-openai/apiKey, azure-openai/endpoint, and
anthropic/apiKey. When an integration manifest references the matching kv://
address, aifabrix upload resolves the value from local or shared storage and pushes
it to the Dataplane without placing the value in the manifest. Miso onboarding can
also configure these standard addresses directly.
Remove the platform completely: aifabrix teardown. Command details: Infrastructure commands.
Goal 2: External system integration
Create and deploy an external system (e.g. HubSpot): wizard or manual setup, then validate and deploy.
Example: HubSpot:
- Create:
aifabrix create hubspot-test(external is the default; oraifabrix wizardfor guided setup). For a web app useaifabrix create my-app --type webapp. - Configure auth and datasources under
integration/hubspot-test/. - Validate:
aifabrix validate hubspot-test - Deploy:
aifabrix deploy hubspot-test
→ External systems guide · Wizard
Goal 3: Build your own application
Create, configure, and run your own AI Fabrix application locally or deploy it (create app → configure → build → run / deploy).
Documentation
All guides and references are listed in docs/README.md (table of contents).
- CLI Commands Reference – All commands
- Infrastructure – Install the platform, services, dev/pro setup
- Configuration reference – Config files (deployment key, application.yaml, env.template, secrets)
Requirements
- Docker Desktop – For running containers
- Node.js 22+ – Supported CLI floor; Node.js 24.22.1 is recommended
- Azure CLI – For deploying to Azure (optional)
License
© eSystems Nordic Ltd 2026 - All Rights Reserved
@aifabrix/builder is part of the AI Fabrix platform.