Triggers and logic
Workflows start from schedules, system events, or user requests, passing data through conditional branches and custom code runtimes.
Builder path
This path is designed for technical contributors, automation engineers, and department creators who want to build, test, and publish custom functions, agents, and multi-step workflows.
1. Design principles and workflow architecture
Section titled “1. Design principles and workflow architecture”Before building, understand how Rival structures business logic and automation sequences.
Modular functions
Build standalone, reusable functions in Python, JavaScript, Lua, or Storm DSL to handle specific system tasks or data transformations.
Human-in-the-loop safeguards
Design explicit approval checkpoints into workflows before high-risk actions, such as financial steps or external system updates, execute.
2. Connectors and authentication setup
Section titled “2. Connectors and authentication setup”Connectors give your custom agents and workflows secure access to internal systems:
- System access — authenticate internal apps and APIs using secure workspace connectors.
- Credential isolation — never hardcode API keys or credentials into scripts. Use centralized workspace secrets and admin-approved connectors. You can enter these under settings in the Organization Settings section.
- Environment testing — test custom integrations inside a sandbox environment before submitting them for team production use.
3. Building and testing agents, workflows, and functions
Section titled “3. Building and testing agents, workflows, and functions”Transform requirements into functional automations using manual setup or AI-assisted drafting:
- Drafting with RivalBot — use RivalBot to generate agents, functions, initial code drafts, schema definitions, and workflow steps from plain-language prompts.
- Building workflows — create workflows visually with the workflow canvas or through conversational prompts using RivalBot.
- Defining input and output schemas — set clear JSON parameter schemas for your functions so agents know exactly what data to pass during execution.
- Sandbox execution — test functions interactively in the sandbox viewer to debug logic, inspect execution latency, and verify output payloads.
4. Versioning, governance, and Marketplace publishing
Section titled “4. Versioning, governance, and Marketplace publishing”Publish your finished automations so your team or organization can adopt them safely:
- Immutable versioning — apply semantic version tags, such as 1.0.0, to your assets so updates never break existing production workflows.
- Submit for admin review — route new functions and agents to the administrative requests queue for security and policy review.
- Catalog publishing — once approved, publish your asset to the internal Marketplace with clear descriptions and usage guidance for end users. If your administrator requires approval for published assets, it is automatically submitted to their requests queue for review.
Builder milestone checklist
Section titled “Builder milestone checklist”| Phase | Core task | Primary outcome |
|---|---|---|
| Architecture | Map business logic and approval gates | Workflow diagram and schema plan |
| Connectors | Link external APIs and workspace secrets | System access authenticated safely |
| Build and test | Draft logic with RivalBot and test in the sandbox or with the workflow canvas | Functional, bug-free function or agent |
| Publish | Tag a semantic version and submit for admin review | Published asset in the Marketplace |
Next steps
Section titled “Next steps” Developer guide Walk through building custom code tools.
Publishing and governance guide Submit finished assets for review.
Building workflows Design multi-step processes on the workflow canvas.
Building agents Configure custom agents with their own tools and guardrails.