Out of quota?
Switch mid-chat.
Pick another model from the same composer. Your messages, the replies and every file path come with you, and reasoning goes all the way up. When a limit resets, the agent picks up on its own.
Every agent,
one window.
Claude, Codex, Cursor, Grok and more, on the subscriptions you already pay for. Switch models mid-chat, run agents as a team, and keep going from your phone.
- 7
- coding agents, one window
- $0
- to use it, MIT licensed
- 1 click
- to switch models mid-chat
- 3
- surfaces: desktop, web, phone
Ask for a team.
Get one.
A lead spawns agents on Claude, Codex or Grok, each in its own thread you can open. They message each other, every answer finds its way back to the lead, and if one fails you switch its model and it carries on.
Ask without
interrupting.
Need a quick answer while the agent works? Ask any model in a side chat. It already knows what the agent has done, and the main thread never stops.
Answers
you can see.
Ask for the architecture and the agent draws it as a diagram you can zoom and pan. Need a chart or a report? It builds a live page, right in the chat.
Agents test
what they build.
Agents open your app in ViewCode's own browser, click through it and check that it works, while you watch. Take over any time.
A second opinion
in one click.
Quick, deep or a security audit on any pull request or uncommitted change. Every finding sits on its line with a fix an agent can apply.
Your agents work.
You watch.
LowMediumHighMaxLowFable 5.1
Fable 5.1LowMediumHighMax
Context window200k1M
Not tied to
your laptop.
Switch agents and models, write code, answer their questions, review your changes and send the fixes, all from your phone. Your computer does the work; you work from anywhere.
The rest of it,
without the tour.
Usage rings. Every plan's limits on the rail, with the exact reset time.
Resumes after limits. Hit a usage limit and the agent continues a minute after it resets.
Checkpoints. Every turn is a hidden git ref you can diff or revert.
Terminal. A real terminal under the chat, with server-side scrollback.
PR watch. An agent wakes when checks fail, pass, or someone reviews.
Computer use. Agents see and drive desktop apps, with per-action approval.
Session import. Bring past sessions from Claude Code, Codex or T3 Code.
Source control. Commit, push and open a PR in one button, with worktrees per thread.
Permission modes. Supervised to full access, chosen per thread.
26 themes. Tinted glass themes, light and dark, with window transparency.
No open port. The desktop app opens a port only when you turn on phone access.
Read aloud. An on-device voice reads replies. Nothing leaves the machine.
Your logins stay
on your machine.
No ViewCode account. Agents run on your computer with your own CLI sign-ins, and your code goes only to the provider you picked.
Free.
You already pay for the agents.
Free forever
ViewCode
Every feature, desktop, web and phone. Open source.
$0 MIT licensed
Download- Every provider in one window
- Handoff, child agents and side chats
- Reviews, diagrams and the agent browser
- Android app included
Bring your own
Your agents
The plans you already have, signed in with their own CLIs.
- Claude Code
- Codex
- Cursor
- Grok
- OpenCode
- Antigravity
- Command Code
Questions, answered.
Do I need a ViewCode subscription?
No. ViewCode is free and MIT licensed. It runs the agents you already pay for, on your own machine, with your own logins.
Which agents does it support?
Claude Code, Codex, Cursor, Grok, OpenCode, Antigravity and Command Code. Sign in to each with its own CLI, then they all appear in one picker.
What does a handoff actually send?
Your messages verbatim, a recap of the replies, and key facts such as links, branches, file paths and PR numbers, under a fixed budget. No model summarizes anything.
Does my code leave my machine?
Only through the provider you chose, exactly as it would from that provider's own CLI. ViewCode itself talks to nothing else. Phone notifications carry titles only.
Can I use it on Windows or Linux?
Yes. The desktop app builds for macOS, Windows and Linux. Computer use is macOS and Linux for now.
Is it a fork of T3 Code?
Yes, and it tracks upstream. ViewCode adds handoff, child agents, agent messaging, Command Code, a no-port desktop mode and its own look, and keeps everything T3 Code ships.