Command Code

Command Code is a coding agent built around open models. It runs as a command-line tool, a desktop app, and sandboxed cloud sessions, and it can read a codebase, plan a change, edit files, run commands, review diffs, and open pull requests. Two ideas set it apart: tool-call repair that fixes malformed calls on every supported model, and taste-1, a model that turns your accepts, rejects, and edits into reusable project skills. It also supports hooks, MCP servers, plan review, session sharing, and a model catalog that mixes open and premium options.

Key Features

• A CLI installed with a single npm command, plus a desktop app for macOS, Linux, and Windows.
• taste-1, which learns preferences from every accept, reject, and edit and writes them into project-level skills.
• Tool-call repair and context engineering aimed at making open models reliable on real repositories.
• Interactive, headless, and sandboxed agents, plus cloud sessions billed by the second.
• Hooks that block unsafe commands and enforce project rules, and plan review before edits land.
• Skills, slash commands, MCP servers, plugins called mods, and session sharing or forking.
• API access through OpenAI-compatible and Anthropic-compatible endpoints on every plan except Go.

Who is it for

• Developers who want to run coding agents on open models without losing reliability.
• Teams that want shared conventions enforced through skills, hooks, and plan review.
• Engineers who automate reviews, tests, migrations, and repository-wide refactors.
• Cost-conscious builders comparing subscription plans against raw API pricing.
• Command-line users who prefer a terminal agent over a full editor.

What can you do with Command Code

• Describe a feature and let the agent implement it across files, then review the diff before committing.
• Turn a repeated review or migration process into a skill the whole team can pull.
• Connect MCP servers so the agent can query data, use tools, and automate workflows.
• Run headless or cloud sessions for longer jobs while you work on something else.
• Capture your preferences once and have later sessions follow the same libraries, test runner, and command style.
• Share a session with a teammate or fork it to explore a different approach.

How does Command Code work

1. Install the CLI with npm, or download the desktop app for your operating system.
2. Sign in, choose a plan, and pick a model from the available open and premium options.
3. Describe the task in the terminal or desktop chat; the agent can plan first, then edit and run commands.
4. Approve the plan and changes, with hooks and permission modes guarding commands that could be unsafe.
5. Push your learned taste to the team, or keep it local, and let later sessions reuse the same conventions.

Pricing

• Go: $1 a month plus a processing fee, with $10 in credits and access to the taste-1 model.
• GOAT: $10 a month plus a processing fee, with $80 in credits, API access, and access to more than 30 additional models.
• Pro: $20 a month, with $80 in credits and premium models included in the shared pool.
• Max 10x: $100 a month, with $150 in credits, higher rate limits, and priority support.
• Max 20x: $200 a month, with $300 in credits and the highest limits.
• Provider: $15 a month for API access with no markup; Teams: $40 a month with pooled credits.
• Enterprise pricing is custom and includes unlimited seats, a dedicated support engineer, and SLA guarantees.

FAQ

Is there a free plan?

The site says pricing is free to start, but the published plan list begins at $1 a month for Go with $10 in credits included. Check the billing page for any trial or starter credit that applies to new accounts.

Which models can I use?

The catalog includes open models such as DeepSeek, GLM, Kimi, MiniMax, and Qwen variants, plus selected closed models. The plan table lists per-model allowances, and API access is included on every plan except Go.

Does Command Code train on my code?

No. The company states that it does not train on your code and that taste learning data is stored on your machine. The privacy policy has the details.

Can I use my own API key?

The pricing page mentions bring-your-own-key support, and API access through OpenAI and Anthropic compatible endpoints is included on plans above Go. Check the docs for the current setup steps.

Can I cancel or change plans?

Yes. The company says you can upgrade, downgrade, or cancel from the billing area, keep access until the end of the billing cycle, and roll unused credits over.