The rivano init and rivano deploy commands let you define agents as code and deploy them declaratively from a rivano.yaml file — the same file used to configure the gateway.

Initialize a project

Run rivano init in the root of your project to create a starter rivano.yaml:

rivano init
# ✔ Created rivano.yaml

Generated file:

# rivano.yaml

agents:
  - name: my-agent
    description: Describe what this agent does
    modelProvider: openai
    modelName: gpt-4o
    environment: production

gateway:
  port: 8080
  providers:
    openai:
      apiKey: "${OPENAI_API_KEY}"
  middleware:
    pii: true
    injection: true
    audit: stdout
  cloud:
    apiKey: "${RIVANO_API_KEY}"
    baseUrl: "https://api.rivano.ai"

Edit the file to define your actual agents and provider configuration.

rivano.yaml — agents section

agents:
  - name: contract-summarizer
    description: Summarizes legal contracts for review teams
    modelProvider: openai
    modelName: gpt-4o
    environment: production

  - name: support-classifier
    description: Classifies support tickets by category and urgency
    modelProvider: anthropic
    modelName: claude-3-5-sonnet-20241022
    environment: production

  - name: code-reviewer
    description: Reviews pull request diffs for security issues
    modelProvider: openai
    modelName: gpt-4-turbo
    environment: staging

Agent fields

Field	Required	Description
`name`	Yes	Unique agent identifier within the tenant
`description`	No	Human-readable description
`modelProvider`	No	LLM provider: `openai`, `anthropic`, `google`, etc.
`modelName`	No	Model identifier
`environment`	No	`production`, `staging`, or `development`

Deploy

rivano deploy

Output:

Validating rivano.yaml...
✔ Schema valid
✔ 3 agents found

Deploying...
✔ contract-summarizer   → v4 (updated model: gpt-4o)
✔ support-classifier    → v2 (no changes, skipped)
✔ code-reviewer         → v1 (new)

3 agents processed. 2 deployed, 1 unchanged.

Dry run

Preview what would be deployed without making any changes:

rivano deploy --dry-run

Output:

Dry run — no changes will be made.

Validating rivano.yaml...
✔ Schema valid
✔ 3 agents found

What would change:
  contract-summarizer   UPDATED (model: gpt-4-turbo → gpt-4o)
  support-classifier    UNCHANGED
  code-reviewer         NEW

2 agents would be deployed.

Specify a config file

rivano deploy --file ./config/rivano.prod.yaml

CI/CD example

Deploy from a GitHub Actions workflow:

- name: Deploy Rivano agents
  run: rivano deploy --no-color
  env:
    RIVANO_API_KEY: ${{ secrets.RIVANO_API_KEY }}

💡

Run rivano deploy --dry-run in pull request checks and rivano deploy on merge to main. This gives you a preview of agent changes before they go to production.

Validation

rivano deploy validates rivano.yaml before deploying. Validation checks:

YAML syntax is valid
All required name fields are present
modelProvider values are recognized provider keys
No duplicate agent names

If validation fails, the deploy is aborted with an error message and a non-zero exit code.

rivano deploy
# ✗ Validation failed:
#   agents[1].name is required
# Aborted. No agents deployed.
echo $? # 1

CLI Agents — Inspect deployed agents and roll back
Gateway Configuration — The full rivano.yaml reference
SDK Agents — Programmatic agent management

CLI — Deploy