Getting Started with VibeFlow
This guide walks you through setting up your first VibeFlow project and connecting an AI agent to start building.
Prerequisites
- An Axiom Studio account (sign up at axiomstudio.ai)
- A git repository for your project
- An AI coding agent (Claude Code CLI, Cursor, Gemini, or similar)
Step 1: Configure Object Storage
Before using VibeFlow, configure where your documents and assets are stored:
- Navigate to Storage in the sidebar
- Select your storage provider (S3, MinIO, Cloudflare R2, or Google Cloud Storage)
- Enter your endpoint URL, bucket name, region, and credentials
- Click Test Connection to verify
- Save your configuration
Step 2: Create Your First Project
- Go to VibeFlow in the sidebar
- Click + New Project
- Enter a project name (e.g., "my-app") and optional description
- Set the default git branch (usually
main) - Click Create
Your project is now in in_review status, ready for planning.
Step 3: Add Features and Work Items
Create a Feature
Features are major areas of work within a project:
- Open your project
- Click + New Feature
- Enter a name (e.g., "User Authentication"), description, and priority
- The feature starts in
in_reviewstatus
Create Todos
Todos are specific tasks within a feature:
- Open a feature
- Click + New Todo
- Enter a title (e.g., "Add login form with email/password"), description, and priority
- Set the target git branch
- Click Create
Create Issues
Issues are standalone bugs or tasks at the project level:
- From the project view, click + New Issue
- Enter title, description, and priority
- Optionally link to a feature
- Click Create
Step 4: Prepare Work Items
Move work items through the planning stages:
in_review→ Review the item with stakeholdersplanning→ Plan the implementation approachready_to_implement→ Mark as ready for an agent to pick up
Items in ready_to_implement status will be picked up by connected agent sessions.
Step 5: Connect an AI Agent
Using Claude Code CLI
- Go to your project's Setup tab
- Copy the MCP server configuration for Claude CLI
- Add the configuration to your Claude Code settings
- Start a new Claude Code session with the VibeFlow persona prompt
Using Other Agents
VibeFlow supports multiple agent types:
- Claude Desktop — Copy the MCP config to your Claude Desktop settings
- Cursor — Add the MCP server URL to your Cursor configuration
- Gemini — Configure via the Gemini MCP integration
- OpenAI Codex — Connect via the MCP bridge
Each agent type has specific setup instructions in the Setup tab.
Step 6: Watch Agents Work
Once connected, agents autonomously:
- Poll for ready work items on your branch
- Claim the highest-priority item
- Plan the implementation (read docs, analyze code)
- Implement the code changes
- Test the implementation
- Commit to git with traceability
- Mark done and move to the next item
You can watch real-time progress in the Session Chat — agents publish execution logs and ask you questions when they need input.
Step 7: Review and Verify
After agents complete work:
- Security Review — The security lead persona reviews for vulnerabilities
- QA Verification — The QA lead persona tests functionality
- Both reviews must pass before a work item is fully complete
Next Steps
- Projects — Learn about project management in detail
- Personas — Set up a multi-persona agent team
- Agent Sessions — Deep dive into autonomous agent workflows