Quick Start Guide

Deploy Your First Agent

Learn how to deploy your first AI agent using Hikigai's config-based deployment. Start in your personal workspace, or set up a team organization with shared cloud accounts, LLM keys, and members. Choose GCP or AWS as the cloud target. No coding required.

Beginner Friendly5 Min SetupNo CodeGCP & AWS

New to Hikigai?

Solo builders can stay in the personal workspace (Free / Pro). Teams create an organization, link cloud accounts, add LLM keys, invite members by role, then deploy — usually in a few minutes.

1

Create Your Account

First, you'll need a Hikigai account. Signing in creates your personal workspace automatically.

•

Visit login and sign in securely with your Google account.

•

Access the developer console immediately. Use the org switcher in the sidebar to move between your personal workspace and any team organizations you join.

Personal account rate limits (Free)

New personal workspaces start on Free. Exceeding a limit returns HTTP 429 until the window resets. Hikigai admin can grant Pro for higher hosted caps.

  • 1 project · 2 agents · 0 apps / MCP servers
  • 10 requests/min · 20 requests/day · 5 agent invokes/day
  • Deploys use Hikigai's platform cloud (no team BYOC required)
  • Personal workspaces cannot invite members

Full Free / Pro / Enterprise matrix: Usage Limits.

Staying personal?

Skip to Step 5: Create Project and deploy from your personal workspace. Continue below when you need a shared team org.

2

Create an Organization

Team organizations share projects, members, cloud accounts, and LLM keys. They run as Enterprise and deploy into your own cloud (not Hikigai's platform account).

1.

Open Create an organization (or use the org switcher → create).

2.

Enter an organization name (a URL-safe slug is generated for you). You become the Owner.

3.

Choose member project access:

  • Inherit — members can work on every project at their org role level.
  • Explicit — members only reach projects they are added to. Owners always see everything.
4.

Optionally add members by email and role now (they must already have a Hikigai account — they join immediately, no invite link). You can also add them later in Step 4.

5.

Click "Create organization", then switch to it in the sidebar org switcher.

Enterprise on your cloud

Team projects cannot deploy until an organization (or project) cloud account is linked. Complete Step 3 before deploying.

3

Link Cloud Accounts & Add LLM Keys

In a team org, configure shared defaults under Organization Settings. Owners and admins can edit these.

Link cloud accounts (BYOC)

1.

Open Organization Settings → Cloud accounts.

2.

Add a GCP service-account JSON key or AWS credentials. Org accounts become the default for every project in the organization; a project can still attach its own override.

3.

Validate the credential, then use it when deploying. Setup details: Cloud accounts (GCP / AWS).

Add organization LLM keys (BYOK)

1.

In Organization Settings → LLM API keys, add keys for OpenAI, Google, Anthropic, and/or OpenRouter.

2.

These become organization defaults. Project members can still set a project-level override in project settings.

Personal vs team

Personal projects can deploy on Hikigai's platform cloud without BYOC. Team organizations must use an org or project cloud account. LLM keys are optional but recommended so billing and quotas stay with your providers.

4

Add Members & Access

Manage the roster from Organization Members or Organization Settings. Assign each person a role that matches what they should do.

Owner

Full control: members, settings, plan, delete organization.

Admin

Manage projects, members, cloud accounts, and LLM keys.

Release Engineer

Builds and deploys agents and apps.

Agentic / App Developer

Agentic: agents. App Developer: apps and consuming agents.

Tester

Read and invoke only — good for QA and stakeholders.

1.

Add a member by email and choose their org role. They must already have a Hikigai account — they join immediately (no invitation link). Ask them to sign in first if they are new.

2.

If access is Explicit, also add them to each project they need (Owners always reach all projects).

3.

Change or remove roles anytime from the members page. Org roles govern the workspace; project roles still apply inside a single project when used.

Match role to access: Testers for invoke-only, developers for build/deploy, Admins for roster and org credentials. Prefer Explicit access when not everyone should see every project.

5

Create a Project

Projects organize agents, apps, and MCP servers. Create them in the workspace selected in the org switcher (personal or team).

1.

Navigate to the Projects page in your console

2.

Click "New Project"

3.

Enter a project name and description

4.

Optionally attach project Cloud accounts (AWS or GCP) if you need an override of the org default — team orgs already inherit organization cloud accounts

5.

Click "Create Project"

Example:

Name: Healthcare Assistant

Description: AI agents for clinical documentation and patient care

6

Configure Your Agent

Now you'll configure your agent using our simple form-based interface.

1.

Open your project's Agents tab and click "Deploy New"

2.

Select "Config-Based" deployment mode

3.

Fill out the configuration form with your agent details (including the cloud target at the bottom of the form)

Required Fields:

Agent Name

A unique identifier for your agent (e.g., "clinical-note-summarizer")

Description

A short summary of what the agent does (at least 10 characters). Required before you can preview or deploy.

Instruction

Tell the agent what it should do. Be specific and clear.

You are a clinical documentation assistant. Convert patient encounter transcripts into structured SOAP notes. Extract key medical information including symptoms, vital signs, diagnoses, and treatment plans.

Model

Choose the AI model (gemini-2.5-flash, gpt-4, etc.)

Input Schema

Define what data your agent accepts (JSON Schema format)

{
  "type": "object",
  "properties": {
    "transcription": {
      "type": "string",
      "description": "Raw patient visit transcript"
    },
    "patient_id": {
      "type": "string",
      "description": "Patient identifier"
    }
  },
  "required": ["transcription"]
}

Output Schema

Define what data your agent returns (JSON Schema format)

{
  "type": "object",
  "properties": {
    "subjective": {
      "type": "string",
      "description": "Patient-reported symptoms"
    },
    "objective": {
      "type": "string",
      "description": "Clinical observations"
    },
    "assessment": {
      "type": "string",
      "description": "Diagnosis and analysis"
    },
    "plan": {
      "type": "string",
      "description": "Treatment plan"
    }
  }
}

Cloud target

Choose GCP or AWS on the same form. Personal projects can leave credentials empty to use Hikigai's platform cloud. Team orgs must pick an organization or project cloud account (the org default from Settings → Cloud accounts is used when you leave the picker on default).

Pro Tip

Start with a simple instruction and schema. You can always refine them after testing your agent.

7

Deploy Your Agent

Review your configuration and deploy. The cloud target was already chosen on the configure form (Step 6).

1.

Click "Preview Code" to review the generated code

2.

Verify your agent's instruction, schemas, cloud target, and settings

3.

Click "Deploy Agent"

4.

Wait while your agent is deployed (you'll see real-time progress)

What happens during deployment:

  • Agent code is generated from your configuration
  • Container or managed runtime is provisioned on the selected cloud target
  • Secure endpoint URL is created
  • Agent is registered in your project

Multi-cloud

Agents, apps, and MCP servers share the same catalog-driven cloud targets. See the docs for the full service list.

8

Test & Invoke Your Agent

Once deployed, you can immediately test and invoke your agent.

Using the Console:

1.

Open your project's Playground tab, or navigate to your agent's detail page

2.

Use Form or JSON input (optionally attach multimodal files)

3.

Enter sample input data matching your input schema

4.

View the agent's response and verify it matches your expectations

Using the Python SDK:

from hikigai.appsdk import AppClient

client = AppClient(api_key="your-api-key", project_id="your-project-id")

# Invoke your agent
response = client.agent("clinical-note-summarizer").invoke({
    "transcription": "Patient presents with...",
    "patient_id": "12345"
})

print(response.content)

Using Platform API:

curl -X POST https://backend.hikigaiplatform.io/api/v1/agents/clinical-note-summarizer/invoke \
  -H "Authorization: Bearer hikigai_YOUR_API_KEY" \
  -H "X-Project-ID: YOUR_PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "input": {
      "transcription": "Patient presents with...",
      "patient_id": "12345"
    }
  }'

API Keys

Generate API keys from the API Keys page in your console.

Next Steps

Deploy an App

Ship full-stack apps from GitHub to GCP or AWS

Deploy MCP Servers

Run tool servers from GitHub; auto-register as connectors

Rooms & Event Bus

Realtime WebSocket rooms and project webhooks

Explore the SDK

AgentSDK, AppSDK, multi-cloud, events, and rooms

Platform API Reference

Complete Platform API documentation

Changelog

Rooms, Event Bus, multi-cloud, and more