RolyCode
RolyCode
One Tool to Rule Them All: RolyCode for AI Coding Agents
Live Free, Vibe Hard
RolyCode is a local web control surface for the coding-agent CLIs you already use, such as Codex, Claude Code, Gemini, Grok, Cursor, and local model chat. It runs the backend on your machine and gives you one desktop/mobile interface for starting threads, sending prompts, reviewing output, and driving repo workflows.
It uses local workflow-model calls, such as Ollama, llama.cpp, LM Studio, or another local OpenAI-compatible server, to refine prompts, generate thread titles and commit messages, and parse commands from private Telegram and Discord bots.
Use the Android companion app on Android, or use Telegram and Discord bots from iPhone or Android. The desktop server can stay local, run on your LAN, or be reached over Tailscale when you configure a private auth token.
Developed by DataWiseguys, LLC.
Join our Discord or email us to ask questions or give feedback.
Use the Android mobile companion app to replicate your desktop's interface on the go.
Talk to your private bots using a short list of commands to streamline your vibe coding experience.
Talk to your private bots using a short list of commands to streamline your vibe coding experience.
Use one thread to manage an entire workflow or update, start to finish.
Supports coding-agent CLIs such as Codex, Claude Code, Gemini, Grok, and Cursor, plus local model chat through engines such as Ollama, llama.cpp, LM Studio, and local OpenAI-compatible servers.
Push raw text to an agent CLI when you know exactly what you want, or route it through the local workflow model when it needs refinement or planning.
Keep the repo workflow inside the thread: sync to main, make changes, commit, push, and create GitHub pull requests.
Use the larger desktop surface at your desk, then switch between devices and pick up where you left off using mobile while relaxing on your couch.
Android gets the native companion app. iPhone and Android users can also send private bot commands such as /new, /pull, and /pr through Telegram or Discord.
Use the desktop file picker to add a repository, then name it with an alias so it is easy to find in the project list.
Paste screenshots straight with Ctrl+V or the + button when the conversation needs visual context.
See live or estimated token and budget usage in one panel so you can keep an eye on resets and current burn across coding agents. 🔥😈
Keep an eye on every local inference endpoint from one place, so you know a model is reachable before you route a prompt to it.
RolyCode probes each configured local endpoint, such as Ollama, llama.cpp, LM Studio, or another OpenAI-compatible server, and shows whether it is connected or unavailable, the endpoint host, the active model, when it was last checked, and any connection error.
Schedule recurring or one-time thread runs so RolyCode keeps working while you are away, then use the timing to squeeze more out of every usage reset.
Pass existing context to a new thread and agent
Fork this thread's context into a new editable thread, then hand it to another agent: route simple tasks straight through, or fork big tasks after planning so the follow-up agent starts with the full plan.
Start a new thread by voice
Open Vibe mode to start a new thread and interact with RolyCode directly through voice chat.
Resume a thread by voice
Start Vibe on any existing thread to pick up where you left off with hands-free voice chat.
The desktop package is the required base install because it owns the local backend, launcher restart, .rolycode.env, Telegram & Discord settings, auth token, local workflow-model configuration, and the coding-agent CLIs on your machine. Use the Windows EXE on Windows or the Linux binary on Linux. The Android APK is optional and connects back to the desktop server.
Reach your desktop backend from anywhere without hand-editing firewall rules or running Administrator terminals. Simplicity is king.
Usage terms for the software, your responsibilities, acceptable use boundaries, and liability limitations.
How the app handles local data, optional third-party integrations, and what remains under your control.
What generated output can and cannot be trusted for, and where human review is still required.