Quickstart
Use this UI-first flow to get from a fresh OpenVibely instance to a running AI coding task.
1. Start OpenVibely
./start.shOpen http://localhost:3001 in your browser.
2. Add A Model In The App
Open Models from the sidebar and create a model configuration. For the shortest Codex path, add gpt-5.6-sol with medium reasoning effort and set it as the default model.
| Provider Path | Typical Setup |
|---|---|
| OpenAI | OAuth or API key auth for Codex models. Suggested starting model: gpt-5.6-sol; if its model config does not specify a reasoning effort, the runtime uses medium. |
| Anthropic | OAuth or API key auth for Claude models. Useful as a secondary coding model. |
| Ollama | Local model server, defaulting to http://localhost:11434 when no base URL is set. |
| OpenAI-compatible | Use a built-in preset or custom endpoint for OpenAI-style Chat Completions providers and gateways. Presets include OpenRouter, NVIDIA NIM, DeepInfra, Fireworks, Groq, Mistral, Cerebras, Together, Hugging Face Router, DeepSeek, Qwen / DashScope, Z.AI / GLM, Tencent TokenHub, LiteLLM, vLLM, LM Studio, SGLang, and more. |
Use the preset list when your provider appears there; OpenVibely can fill the expected endpoint pattern and try model discovery. Use Custom OpenAI-Compatible for a gateway or local runtime that needs its own base URL, auth header, extra headers/body, or manual model ID.
Set a default model if you want new tasks to work without choosing a model every time. See Model Providers for the full provider setup guide.
3. Create A Project From The Sidebar
Use the project selector area in the sidebar to create a project. A project gives Chat, Tasks, Schedule, Workers, Alerts, Memory, Insights, and Channels their shared workspace.
Choose either a local repository path or a repository URL. Local paths are enabled by default in desktop mode, but server mode requires OPENVIBELY_ENABLE_LOCAL_REPO_PATH.
4. Start From Chat Or Tasks
| If You Know The Exact Work | If You Need To Explore First |
|---|---|
Open Tasks, click + Add Task, enter a title and prompt, choose a model or agent if needed, then create the task. | Open Chat, ask about the project or describe the request, attach files if useful, then let the conversation guide task creation. |
5. Run And Watch Progress
On the Tasks page, run the task and watch it move through queued/running/completed or failed states. The UI updates from live events, and Alerts can notify you when failures or follow-up events happen.
6. Review Before Shipping
Open the task detail view to inspect the prompt, thread, execution output, attachments, changed files, review comments, worktree state, and pull request options when a repository is attached.
Next Steps
| Want To... | Read Next |
|---|---|
| Understand the main surfaces | Projects, Chat, and Tasks |
| Configure reusable AI behavior | Agents and Models |
| Automate recurring work | Schedule and Scheduled Task Runs |
| Review code safely | Git Worktrees & Merge Safety and Task Lifecycle |