v1.12 · macOS · Linux · Windows

The coding agentfor every model,in your terminal.

It reads your code, edits it, runs the commands and checks its own work. Any model.

$npm i -g @ccode-si/cli
Start free

Free with GitHub · $10 a month

CCode proposes an edit to src/price.ts, shows it as a diff and waits for an answer: yes, yes and don't ask again, switch to auto mode, or no. Once allowed it makes the edit and runs the tests: all 16 pass.

Permissions

Ask, auto or plan. You decide how far it goes alone.

Three agents are sent off in the background: one maps the routes, one writes tests for the cart, one reviews the last commit. A list under the prompt shows each one working, and each reports back as it finishes.

Agents

Split the job across agents working side by side.

CCode changes one line and shows it as a diff. The /review command then lists what it found in the uncommitted work, each with a case and a fix, and /rewind restores the code and the conversation to before an earlier turn.

Review

Every change as a diff first. Undo any of it.

CCode starts the dev server, opens the sign-in page in a browser it drives, types the email and password, and confirms the dashboard loads and every request succeeded.

Browser

It opens the page, clicks through and checks the result.

A goal is set with /goal: npm test passes and npm run lint has no warnings. CCode runs both checks and reports the goal achieved. It then adds a dark mode toggle and tries it in the browser.

Goals

Set a goal. It keeps going until the checks pass.

Bring your own keys

Use the keys and local models you already have. Or one CCode key for all of them.

Anthropic
OpenAI
Google Gemini
DeepSeek
Z.ai
Moonshot AI
Qwen
OpenRouter
Ollama
LM Studio
vLLM
CCode key

Models

Tests behind every release

Free every month

Every model, one key

Claude, GPT and the open models, at each maker's own price.

Get started

$67.27

of $70 · Pro

  • Claude Opus 5.5

    Anthropic

    $20.79

    29.7%

  • GPT-6.2 Sol

    OpenAI

    $17.36

    24.8%

  • Kimi K3.1 Pro

    Moonshot AI

    $13.86

    19.8%

  • DeepSeek V4.1 Flash

    DeepSeek

    $9.73

    13.9%

  • GLM 5.3

    Z.ai

    $5.53

    7.9%

Mix and match

Frontier models for the hard parts, open models for the rest. One balance.

See pricing

Free

$0per month

  • $10 of usage a month
  • Six open models
  • No card
Start free

Starter

$1per month

  • $5 of usage a month
  • All 26 models
  • 8 sessions at once
Get Starter

Pro

$10per month

  • $70 of usage a month
  • All 26 models
  • 8 sessions at once
Get Pro

Included models

FreeStarterPro· 26 models

ModelRequests

Frequently asked questions

Quick answers

  • A coding agent for your terminal. It reads your project, edits files, runs commands and checks its work, with permission modes that decide how much it does on its own.

  • Anthropic, OpenAI, Google Gemini, DeepSeek, Moonshot AI, Qwen, OpenRouter, the Z.ai coding plan, or anything that speaks the OpenAI API, including Ollama, LM Studio, vLLM and llama.cpp. Paste a key in ctrl+p → Connect provider.

  • No. Bring your own keys and CCode works with them. A plan gives you one key for Claude, GPT, GLM and DeepSeek models, charged at each model maker's official price.

  • Sign up with GitHub, or connect GitHub to an email account, and you get $10 of credits a month on six open models: GLM 5.3, GLM 5.3 Flash, Kimi K3.1 Pro, Qwen 4 Max, Qwen 3.8 Max and DeepSeek V4.1 Flash. Each GitHub account unlocks it for one CCode account. Plan details.

  • macOS, Linux and Windows, on Intel and ARM. One npm command installs it; all you need is Node.js 18 or newer.

  • No. A project you clone can't change how CCode behaves: it can't switch the mode or the model, run its own tools, or reach your keys. CCode also asks before it first opens a folder.

// changelog

Shipped, and still shipping.

npm
  1. v1.12.0

    2026-11-16

    • NewBackground sessions. /background sends a session to the background and frees your terminal while it keeps working. Pick it up again with /session or ccode --resume.
    • NewA sandbox for commands (macOS). Turn it on with /sandbox and the agent's commands can only write inside your project and the folders you allow.
    • NewWork across folders. /add-dir gives a session more folders to work in.
    • NewSee a change before you approve it. Permission prompts show the edit as a diff, or the new file's code.
    • New/permissions, rebuilt: see what was blocked recently, and manage allow, ask and deny rules for this project or all of them.
    • ImprovedThe agent's task list stays with it through long sessions.
    • ImprovedPlan mode can search your code before it writes the plan.
    • Improved/thinking shows or hides the model's thinking.
    • ImprovedThe browser handles dialogs, and catches messages that pop up after a click, like toasts and form errors.
    • FixedA key you were already pressing no longer answers a question the moment it appears. What you type meanwhile waits until you've answered.
  2. v1.11.0

    2026-11-12

    • NewWeb search that just works, with the results a browser gets and nothing to install.
    • NewEach session keeps its own model. Reopen a session and it carries on with the model it was using; new sessions start on your last pick.
    • ImprovedLighter on long sessions: CCode uses much less memory after hours of work.
    • FixedWindows: no command window pops up while CCode updates itself.
  3. v1.10.0

    2026-11-12

    • NewStop any agent, resume it later. Stop a background agent, along with the agents it started, then message it to pick up where it left off.
    • NewAgents can't pass for you or for CCode. Messages between you, the main agent and its agents can't slip in instructions that look like CCode's own.
    • FixedThe browser no longer hangs on pages that keep a connection open, such as live updates or streams.

Start in your terminal

$npm i -g @ccode-si/cli
Start free