Getting Started¶
Manifest Platform is an enterprise operating system for orchestrating data, agents, and workflows. It provides a unified control plane where teams build solutions from composable components, deploy them through managed infrastructure, and govern everything with role-based access control and audit logging.
This section walks you through getting started on the platform and using the SDK.
What is Manifest Platform?¶
Manifest Platform sits between your data sources, AI models, and business processes. Instead of stitching together dozens of services, you define your logic as components — agents, workflows, code blocks, connectors — and the platform handles orchestration, scaling, security, and observability.
graph TD
U["Your Team"] --> FP["Manifest Platform"]
FP --> A["AI Agents"]
FP --> W["Automated Workflows"]
FP --> H["Hosted Services"]
FP --> D["Managed Data Pipelines"]
Key Concepts¶
Organizations¶
An organization is the top-level tenant in Manifest Platform. It owns all resources, users, billing, and security policies. Most teams map one organization to one company or business unit.
Workspaces¶
Workspaces partition resources within an organization. Use them to separate environments (development, staging, production), teams, or projects. Each workspace has its own set of solutions, connectors, and deployments.
Solutions¶
A solution is a container for related components. Think of it as a project: it groups the agents, workflows, code blocks, and configuration that together solve a business problem. Solutions are the unit of deployment — you promote an entire solution through deployment rings.
Components¶
Components are the building blocks inside a solution. Manifest Platform defines several component types:
| Component | Purpose |
|---|---|
| Agent | An AI agent with model access, tools, and guardrails |
| Workflow | A multi-step automation that chains components together |
| Code Block | A reusable Python function that runs in a managed runtime |
| Tool | A callable capability exposed to agents (API calls, data lookups, actions) |
| Connector | An integration with an external system (database, SaaS API, file store) |
| Model | A configured AI model with provider, parameters, and routing rules |
| Prompt | A versioned prompt template with variable interpolation |
| Schema | A Pydantic model that defines structured input/output contracts |
| Transform | A data transformation step for reshaping payloads between components |
| Validation | A rule that enforces data quality or business logic constraints |
Everything is a component
The component model means you can version, share, test, and reuse any piece of your solution independently. Agents can call tools. Workflows can invoke agents. Code blocks can be shared across solutions.
What Can You Build?¶
AI-powered customer support — Build agents with access to your knowledge base, ticketing system, and CRM. Add human-in-the-loop approval for sensitive actions like issuing refunds.
Automated data pipelines — Connect to source systems with connectors, transform data with code blocks, and load it into datasets. Schedule the pipeline as a workflow with monitoring and alerting.
Internal tools and services — Deploy Python services as hosted services with HTTP endpoints. The platform manages scaling, authentication, and logging.
Compliance automation — Chain agents and workflows to continuously monitor regulatory requirements, generate audit reports, and flag policy violations.
Next Steps¶
-
Quickstart
Build and deploy your first agent on the platform.
-
Installation
Learn about SDK availability on the platform and upcoming local developer tools.
-
Authentication
Configure API keys or SSO to access the platform.