VibeAround

An all-in-one hub for AI coding agents, keeping your vibe flow around.

Desktop Download | CLI npm | Demo | Wiki | Discord | WeChat | 简体中文

Why VibeAround

VibeAround keeps AI coding work together without making you rebuild the environment you already have.

• Work with the coding agents you already use, including Claude Code, Codex CLI, Gemini CLI, Pi, OpenCode, Claude Desktop, Codex Desktop, and more.
• Launch agents directly or through third-party AI APIs, without hand-editing each agent's config files back and forth.
• Bridge different AI API protocols so agents and model providers can work together even when their native APIs do not match.
• Continue the same sessions across desktop, CLI, messaging apps, mobile browsers, web browsers, and a Web Terminal.
• Preview dev servers, Markdown, and HTML remotely while execution stays on your own computer.
• Provide host-side tools like web search when the selected model provider does not offer them natively.
• Add these capabilities around your existing configs, project permissions, and workflows while keeping them as untouched as possible.

Agent Launch

Launch the right agent with the right model.

Pick an AI agent, model profile or API endpoint, and workspace. VibeAround launches Claude Code, Codex CLI, Gemini CLI, Pi, OpenCode, Claude Desktop, Codex Desktop and more with 3rd party AI APIs, without changing each agent's own config files, skills, MCP servers, workflow, or project context.

• Launch AI coding agents and desktop apps like Claude and Codex from one desktop UI.
• Choose agent, model profile, API endpoint, workspace, terminal, and session before launch.
• Start new sessions or continue previous sessions.
• Use direct launch or profile-based launch, including profile overlays for Claude Desktop and Codex Desktop.
• Record and inspect launch-scoped API traffic, including original request, bridge request, raw response, bridge response, and search tool contents.
• Keep each agent's own config files, workflow, and project context.
• VibeAround does not modify original CLI config files. If you use tools such as CC Switch, manually remove conflicting profile fields such as env in ~/.claude/settings.json.

Known Issue

Why do I see Unable to connect to API (ConnectionRefused)?

This usually means the agent is still reading API settings written by CC Switch or another config-switching tool. Those settings can point the agent to a local proxy that is no longer running, so the agent refuses the connection before VibeAround can route the request.

Remove the conflicting fields from the corresponding agent config file, then launch the agent again from VibeAround. For Claude Code, check ~/.claude/settings.json and remove stale provider fields such as an env block that sets a base URL, API key, or proxy endpoint. If you also have project-level Claude config files, check those as well.

API Profiles & Bridge

Use more model providers with the AI agents you prefer. Save provider keys, base URLs, models, and aliases as profiles; when an agent and a provider speak different API shapes, VibeAround's local bridge translates between them.

Agent-facing API shapes             VibeAround API Bridge               Provider API shapes
+-------------------------+         +-----------------------------+      +-------------------------+
| OpenAI Responses        | ----\   | profile-scoped local routes |  /-> | OpenAI Responses        |
| OpenAI Chat Completions | -----\  | model aliases and metadata  | /--> | OpenAI Chat Completions |
| Anthropic Messages      | ------> | request/response translate  | ---> | Anthropic Messages      |
| Gemini Generate Content | -----/  | va-ai-api-bridge (VAAAB)    | \--> | Gemini Generate Content |
+-------------------------+         +-----------------------------+      +-------------------------+

The bridge is powered by va-ai-api-bridge, the standalone VAAAB project behind VibeAround's API translation.

Built-in provider presets include DeepSeek, Alibaba DashScope, Moonshot / Kimi, MiniMax, Xiaomi MiMo, xAI / Grok, NVIDIA NIM, Z.AI / GLM, Google Gemini, OpenRouter, Azure OpenAI, and custom endpoints.

• Save API settings with keys, base URLs, models, aliases, and metadata.
• Bridge OpenAI Responses, OpenAI Chat Completions, Anthropic Messages, and Gemini Generate Content shapes.
• Run the same AI agent with different API profiles across sessions.

Host-side Web Search

Give agents web search even when the selected model provider does not expose native server-side search.

VibeAround can replace provider-native web_search with a local search runtime, then feed normalized results back into the model through the bridge. Search sources are configured in Settings and run through the va-search-tool project. The same search SDK/runtime can also run standalone from the command line for local smoke tests or custom integrations.

va-search-tool

va-search-tool is the standalone search runtime behind VibeAround host-side web search. It can run as a supervised VibeAround plugin over stdio, as a CLI for one-off searches, or as a small local HTTP service with /v1/search. It currently supports Exa, Tavily, Grok / xAI, and Brave Search.

• Use Exa, Tavily, Grok / xAI, and Brave Search as host-side sources.
• Run standalone CLI checks with va-search-tool search ... or expose a local /v1/search service when you want the search runtime outside VibeAround.
• Test search settings before saving from the Web Search settings page.
• Inspect search requests and source-separated results in the Bridge recorder.
• Keep API keys in local settings while the AI provider only receives normalized search results.

Agent as API

Use local AI coding agents as API endpoints for development and local testing.

VibeAround can expose enabled local agents as OpenAI / Anthropic-compatible APIs, so you can test app integrations against Claude Code, Codex CLI, Gemini CLI, OpenCode, and other local agents without deploying a produ