Quick Start

This guide takes you from sign-in to a working GeniSpace experience. You can start with a ready-made industry solution or build individual resources in Console.

1. Sign in and select a space

1. Open www.genispace.ai.
2. Sign in or create an account.
3. Use the space selector in the header to choose the destination space.

Your active space controls which agents, data, applications, permissions, and billing account are used. Confirm it before creating resources.

2. Choose a starting path

Fastest path: deploy an industry solution

1. Open Explore in the Platform Hub.
2. Select Industry Solutions.
3. Review the target users, business pain, expected outcome, and included resources.
4. Choose whether to create a new space or deploy into an existing space you own.
5. Confirm deployment and follow the progress dialog.

Deployment continues in the background if you close the dialog. When it finishes, open the generated Workbench and test the included agents and workflows. See Industry Solutions.

Build manually in Console

Open Console from the app launcher and create only the resources you need.

Goal	Start here
Interactive AI assistant	Console → Agents → Create → Conversational
Structured AI processing	Console → Agents → Create → Task
Business automation	Console → Workflows, then Tasks
Database access	Console → Data → Data Sources
Semantic record search	Console → Data → Datasets
Document-grounded answers	Console → Knowledge Base
Low-code business UI	Workbench → Create Workbench

3. Create and test an agent

1. Go to Console → Agents and select Create Agent.
2. Choose Conversational for multi-turn interaction or Task for structured input and output.
3. Enter a name, description, model, and system instructions.
4. Optionally add knowledge bases, memory, datasets, tools, or MCP servers.
5. Save the agent.
6. Test a conversational agent in Chat or a task agent in its structured test panel.

Use Agent mode when you want a conversational agent to call tools and work through a multi-step request. Use Ask mode for a direct answer without delegated actions. See Agent Overview and Chat.

4. Add data or knowledge only when needed

You do not need to configure data before creating every agent.

• Use a data source when the agent or workflow must read or update a business database.
• Use a dataset for managed records with structured, full-text, or semantic search.
• Use a knowledge base for documents that should be searched and cited.

Grant the narrowest practical permissions and test with non-sensitive sample data first.

5. Build and publish a Workbench

1. Open Workbench.
2. Create a blank Workbench or start from a template.
3. Add pages, navigation, and components.
4. Bind components to a data source, dataset, workflow, or task where required.
5. Preview desktop and mobile layouts.
6. Save the draft, review it, and publish when ready.

Workbench Copilot can modify the current draft, but its changes are not published until you review and save them. See Workbench.

6. Share safely

Before sharing an agent or application:

1. Confirm the active space.
2. Review space and application members.
3. Verify tool, data-source, dataset, and knowledge-base access.
4. Test actions that create, update, or delete data.
5. Publish only the reviewed version.

If something goes wrong

• A deployment can be reopened from its running-job status and retried if a step fails.
• A long Chat session shows context usage and can be compressed with /compact.
• A failed tool call can be inspected from the execution timeline and Activity Panel.
• A Workbench version can be reviewed or restored from version history.

See Troubleshooting for common recovery steps.