Quickstart¶

This guide walks you through building and running your first component on the Manifest Platform. You will create an AI agent, test it, and deploy it as a hosted service — all from the platform. No local installation required.

Prerequisites

You need an Manifest Platform account with access to at least one organization and workspace. Log in to the platform console to get started.

Step 1: Open the Platform Console¶

Navigate to your workspace in the Manifest Platform console. Confirm you can see the Solution Builder in the left navigation.

SDK in Code Blocks

Every Code Block on the platform has flow-sdk pre-installed. When you write Python in a Code Block, you can immediately from flow_sdk import FlowSDK without any setup.

Step 2: Create a Solution¶

Navigate to Solution Builder in the console and create a new solution. Give it a name like "My First Solution" and an optional description.

Solutions are containers for related components — they group the agents, workflows, code blocks, and configuration that together solve a business problem.

Step 3: Create an Agent¶

Agents are AI-powered components that can use tools, follow instructions, and produce structured output.

Inside your solution, click Add Component > Agent
Name it "Research Assistant"
Select a model (e.g., GPT-4o or Claude Sonnet)
Add a system prompt:

You are a helpful research assistant. Summarize topics clearly and cite sources when possible.
Save the agent

Step 4: Test in the Playground¶

Navigate to Agents > Playground and select your Research Assistant. Test it interactively:

Try: "What are the three laws of thermodynamics?"
Adjust the system prompt or model parameters as needed
When you are happy with the results, move on to deployment

You can also test the agent from a Code Block using the SDK:

from flow_sdk import FlowSDK

flow = FlowSDK()
result = await flow.agents.run(
    agent_id="<your-agent-uuid>",
    input_data={"message": "What are the three laws of thermodynamics?"},
)
print(result["output"])

Step 5: Deploy as a Hosted Service¶

Turn your agent into an HTTP-accessible service. Hosted services give your agent a stable endpoint that other systems can call.

Navigate to Hosted Services > Create Service
Select your Research Assistant agent as the backing component
Choose persistent execution mode
Deploy to the dev ring

Once deployed, you can invoke the service from a Code Block:

from flow_sdk import FlowSDK

flow = FlowSDK()
response = await flow.services.post(
    "/invoke",
    body={"message": "Explain quantum entanglement in simple terms"},
    service_id="<your-service-uuid>",
)
print(response)

What You Built¶

Here is the architecture of what you just created:

graph LR
    REQ["Your Code / curl"] --> GW["API Gateway"]
    GW --> SVC["Hosted Service<br/>research-assistant-svc"]
    SVC --> AGT["Research Assistant<br/>(Agent)"]
    AGT --> MDL["AI Gateway<br/>gpt-4o"]

Your request flows through the API gateway to the hosted service, which invokes the agent, which calls the AI model through the platform's AI Gateway (handling model routing, rate limiting, and cost tracking).

Next Steps¶

Now that you have a working agent deployed as a service, explore these topics:

Topic	Description
Solution Builder	Learn about all component types and how to compose them
Agents	Add tools, guardrails, and human-in-the-loop to your agents
Hosted Services	Configure endpoints, async jobs, and the skills system
Connectors	Connect your agent to external data sources and APIs
SDK Reference	Full SDK and CLI documentation

Explore the playground

The Manifest Platform web console includes an Agent Playground where you can test agents interactively, adjust system prompts, and experiment with different models before deploying. Navigate to Agents > Playground in the console.