VT Code CLI
A terminal coding agent explores projects, edits files and runs checks using your chosen model and command policies
- 898 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The agent edits files and runs policy-permitted tools; automatic classifier review can misjudge a request.
- The chosen model receives code and tool results. exec needs full_auto and workspace trust, while explicit deny and forced ask rules remain constraints.
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/vt-codeDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent edits files and runs policy-permitted tools; automatic classifier review can misjudge a request.
- The chosen model receives code and tool results. exec needs full_auto and workspace trust, while explicit deny and forced ask rules remain constraints.
This is third-party code. Review the repository files before installing.
What it does
VT Code is a standalone coding agent in the terminal. Choose a model, open a project and give it a task. The agent reads code, proposes edits and runs available checks through its environment’s tools. `ask` supports a preliminary conversation without tools. Interactive mode lets you discuss changes, while `review` examines uncommitted edits. Longer sessions can resume later with an execution log. Configuration defines access rules for commands and files. Decisions can require a human prompt or classifier review. The separate `exec` mode performs tasks without the ordinary conversation and needs automation settings plus workspace trust.
Who it is for. Terminal-agent users choosing a model and defining access rules for commands and project files.
Good fit when
- You need to explore a repository and discuss a plan before editing
- You want a bounded change with terminal verification
- You need to resume a long session from saved history
- You want explicit command and project-path policies
Not a fit when
- You need a skill for another installed coding agent
- Project code cannot reach the chosen hosted-model provider
- You need unattended execution without configuring permissions and trust
Example request
Explain authentication in this project and propose a change plan. Leave files untouched and list commands needed for verification.Limitations
Use a chosen model’s key or supported sign-in; your own OpenAI API is billed separately. macOS and Linux are supported, with separate Windows/WSL limits. ChatGPT sign-in unofficially uses Codex’s public client ID, with no availability guarantee from the author. Classifier review can misjudge commands; exec separately needs full_auto and workspace trust, while tool allow-lists and explicit denies still apply.
How to disable. Stop the session and any VT Code schedules. For Cargo installation, run `cargo uninstall vtcode`; otherwise uninstall with the chosen package manager. Revoke unused provider keys. Signing out of VT Code does not remove saved Codex credentials; inspect credential sources separately.
FAQ
How do I install VT Code through Cargo?
Run `cargo install vtcode` with a suitable Rust version. The current source manifest requires 1.98.1 or later; prebuilts use a different installation route.
What does `vtcode init` create?
Project configuration and `AGENTS.md`. Review them before committing, especially if the repository already has agent instructions.
How do I ask without tools?
Use `vtcode ask "your question"`. It returns an answer and exits without a tool loop.
Does automatic review guarantee a safe command?
No. `permissions.auto` uses a classifier. Keep explicit denies and environment restrictions because model-based review can be wrong.
Must Codex be installed for VT Code?
Not for ordinary API-key use. Separate ChatGPT compatibility can use its own OAuth flow or detected Codex credentials; those routes differ from direct API access.
What license and extra costs apply?
First-party code uses MIT OR Apache-2.0. Optional output checks through OpenAI API can incur separate charges.
Discussion
No questions or reviews yet.
Related
CLI toolsA terminal agent explores projects, prepares plans and makes edits or runs commands under a chosen approval policy.
An open coding agent for the terminal, browser and desktop that works with models from OpenAI and Claude to Ollama
Command Code is a terminal agent that uses accepted, rejected and revised edits as feedback for its Taste preference profile.
An open-source coding agent for VS Code, JetBrains and the terminal with hundreds of models to choose from
An open source coding agent for the terminal and desktop with build and plan modes
A teaching terminal coding agent in Python, built as a "nanoGPT for agents": read it, take it apart and fork it
README in short
The project provides a native terminal interface and automation modes. It supports extensions and MCP, whose servers add their own permissions. Client configuration exposes history and access policies.