Ask, edit, verify
Project inspection, checkpointed file edits, tests, diffs, task steering, pinned chats and local review checks.
Native Windows preview · 0.1.3TalkToAi Code.
TalkToAi Code.
Your projects. Your AI.
A local-first coding and game-development workspace. Describe the outcome, let the agent inspect and act, then review the evidence.
No OpenAI subscription required for local models · no GPU required · use your own AMD server when available
One workspace for code, games and evidence
Build games
Godot checks, Blender launch and scripts, Unity project workflows, game capture and screenshot evidence.
Choose your runtime
AMD server · Qwen coder 30B / 3B activePC · Qwen3.8-27BCPU · Qwen3.5 4B
Download TalkToAi Code 0.1.3
Windows installer
The installer downloads the native runtime, creates shortcuts and adds a safe uninstall entry that preserves local data by default.
Download Setup EXEPortable Windows
Use the native folder build when you want to keep the app in a chosen directory without an installer.
Download Portable ZIPLinux and macOS
Use the source installers in the GitHub release. The app uses your local runtime or your own compatible server.
Open release filesThe app contains no private server, model weights, API keys or personal account data. Local inference has no hosted usage meter; hardware, electricity and optional provider terms still apply.
Privacy and control
Local models stay on your machine. AMD mode connects to your own server. ZeroThink is optional account-backed inference; when selected, task context is sent to that service. SSH credentials remain in your own OpenSSH setup.