Everything OpenCode Mobile can do
A full-featured Android client for the OpenCode AI coding agent — real-time sessions, multi-model support, private server connectivity, and no cloud intermediary.
Core capabilities
What the app delivers end-to-end, from connection to completed coding task.
Real-time streaming sessions
Streams the agent's tool calls, file edits, and reasoning tokens via Server-Sent Events as they happen — no waiting for a complete response before you see output.
Any model opencode supports
Inherits your server's model configuration: Claude 3.5/4, GPT-4o, Gemini 1.5/2, Ollama local models, and any provider opencode can reach. Switch models server-side; the app follows.
Tailscale & VPN connectivity
Point the app at any private IP or Tailscale MagicDNS hostname. No port-forwarding or public exposure required — your codebase stays on your machine.
Tool-use & diff output
Renders the agent's tool calls inline — file reads, writes, shell commands — so you see exactly what it did to your code, not just a prose summary.
Full session history
Every conversation is stored locally. Scroll back through past sessions, resume interrupted work, or review what the agent changed in a previous run.
Multiple saved servers
Save and switch between multiple server profiles — home lab, work VPN, cloud VM. Each connection stores its own URL and credentials independently.
Password-protected auth
Connects to password-protected opencode serve instances. Credentials are stored in the device keychain, not in plain text.
Built-in diagnostics
Tap "Share diagnostic" to generate a structured connectivity report (DNS, port, health-check, TLS) for faster debugging — without exposing code or message content.
Open source & no telemetry
MIT-licensed, auditable source on GitHub. Crash reporting (Sentry) is opt-in and off by default. No analytics, no account, no cloud relay.
What people use it for
Code review on the go
Ask the agent to summarize a diff, explain a function, or flag potential bugs — from your phone while away from your desk. The session history means you can pick up where you left off.
Quick fixes & patches
Describe the bug in plain language; let the agent locate, patch, and explain the change in your actual codebase. You see every file write as it happens.
Scaffolding new features
Ask the agent to stub out a new module, write tests, or extend an existing pattern — then commit from your laptop when you're back. Start the thinking on the bus, land the code at your desk.
Private local-model sessions
Point opencode at Ollama running locally and your mobile session never touches a third-party server. The app connects directly over your tailnet — fully air-gapped from the internet if you want.
Documentation & explanation
Ask the agent to explain an unfamiliar codebase, write a README section, or generate API docs — useful when you're reading code without an IDE open.
Compared to using Claude.ai or ChatGPT from mobile browser
OpenCode Mobile is a coding-agent client, not a chat UI. The key differences:
- Agent has direct access to your codebase — reads and writes actual files; no copy-pasting snippets.
- Runs against your server — your model config, your API keys, your cost — not a metered consumer tier.
- Streaming tool-use output — see file edits and shell commands as they run, not a finished summary.
- Session persistence — full history stored locally; conversations don't disappear on browser refresh.
- No cloud relay — traffic goes phone → Tailscale → your machine; no intermediary sees your code.
- Self-hosted model support — use Ollama, LM Studio, or any local model opencode supports.
System requirements
- Android 7.0 (API 24) or later
- opencode running on a reachable server (
opencode serve) - Network path from phone to server (LAN, Tailscale VPN, or any tunnel)
- ~15 MB storage for the app; session history grows with use
Get the app
Free, open-source, no account required.
See also: setup guide · troubleshooting · overview