SuperAPI AI Directory
Partner with us中文AI newsFavorites
DirectoryNews and tutorials

MiniMax Code CLI - MiniMax's open-source command-line AI coding tool

Author: AI小集 · Original source: AI工具集

What is MiniMax Code CLI

MiniMax Code CLI is an open-source command-line AI coding toollaunched by MiniMax. As the core component of the MiniMax Code client, it has been open-sourced under the MIT license. In the FrontierHarness Eval benchmark, it achieved a task pass rate of 76.7%, with a median time of 4 minutes 33 seconds for successful tasks, outperforming public baselines. The tool supports one-click installation via curl, can be integrated into IDEs, and focuses on efficient tool calling, permission handling, and enterprise-grade reliability.

MiniMax Code CLI

Main features of MiniMax Code CLI

  • Code editing and verification: Reads files, generates diffs, runs shell commands and tests in the terminal, and manages tool execution through permission control and sandbox mechanisms.
  • Free choice of models: You can use a MiniMax account / Token Plan, or use BYOK to connect custom models compatible with the OpenAI or Anthropic API format.
  • Multimodal tools: Built-in web search and media tools (mcode-tools), and supports MCP and hosted connectors, covering search and media processing needs beyond code.
  • Session continuation and task planning: Supports resuming historical sessions (--continue/--session), Plan Mode, and subagents to advance tasks in parallel.
  • Plugins and skill extensions: Extend Agent capabilities through official, local, or GitHub plugins and built-in skills.
  • Three usage entry points: Interactive TUI (mcode [prompt]), headless mode (mcode exec, suitable for scripts/CI/batch evaluation), ACP protocol (mcode acp, for integration into IDE editors).
  • Project-level guidance：mcode init . can generate AGENTS.md, describing the expected results, allowed modification scope, and verification method for the project.
Dig into GitHub

Follow WeChat and reply "Open Source" to join theAI Open Source Project Exchange Group

How to use MiniMax Code CLI

  • Installation: On macOS / Linux / WSL, run curl -fsSL https://filecdn.minimax.chat/public/install.sh | bash(on Windows, run with PowerShell irm https://filecdn.minimax.chat/public/install.ps1 | iex), to install the CLI and a compatible Node.js environment in one step.
  • Verify installation: After reopening the terminal, run mcode --version to confirm the version, and use mcode --help to view all commands.
  • Log in or configure keys: For domestic accounts, run mcode login; for overseas accounts, run mcode login --region global. After completing login in the browser, use /status to check the account, and/provider to select a model.
  • Connect your own model (BYOK): Without MiniMax login, set the MCODE_PROVIDER_API_KEY environment variable, then use mcode provider add --name my-provider --base-url ... --api-key-env MCODE_PROVIDER_API_KEY --use to add a custom provider.
  • Start and assign tasks: Enter the project directory cd /path/to/your/project, directly run mcode to open the interactive interface and describe the task, or submit it with a one-line command:mcode "Find a failing test, fix the implementation, and run the relevant tests."。
  • (Optional) Generate project guidance: Run mcode init . to generate or update AGENTS.md, specifying the task expectations, allowed modification scope, and verification method.
  • Interactive operations：Enter to send a message,@ to reference a file,Shift+Tab to switch plan mode,Alt+M Switch permission modes,Esc interrupt tasks,/sessions and recover historical sessions.
  • Advanced usage: use headless mode for scripts/CI mcode exec [prompt], use mcode acpfor IDE integration, and interrupted work can be resumed with mcode --continue or --session .

Core advantages of MiniMax Code CLI

  • Outstanding performance: In the FrontierHarness Eval benchmark, the task pass rate reached 76.7%, and the median time for successful tasks was only 4 minutes 33 seconds; both metrics outperform the public baseline, and the official positioning is "continuously unleashing model capabilities through excellent harness design."
  • Truly MIT open source: Since v0.4.12, all first-party source code has been released under the MIT license; developers can examine tool-calling and permission-handling logic, and the community can participate in discovering and fixing issues, making it suitable for building enterprise-grade applications.
  • Zero model lock-in: It supports both MiniMax accounts / Token Plan and BYOK access to any custom model compatible with the OpenAI or Anthropic API format, without being tied to a single vendor.
  • Flexible three-mode architecture: The same toolset provides an interactive TUI, headless mode, and the ACP protocol, covering the full range of scenarios from personal development to engineering deployment.
  • Comprehensive permission and security mechanisms: Built-in permission modes and sandbox controls allow reviewing diffs and confirming tool calls one by one.

MiniMax Code CLI project address

  • GitHub repository：https://github.com/MiniMax-AI/minimax-code

Comparison of MiniMax Code CLI with similar competitors

Comparison dimension MiniMax Code CLI Claude Code
Provider MiniMax (Xiyu Technology) Anthropic
Open source status Fully open source under the MIT license (v0.4.12), source code is auditable Closed source, only the CLI tool is available for download and use
Model support MiniMax account / Token Plan, or BYOK access to any model with an OpenAI / Anthropic compatible API Only supports Claude models, locked into the Anthropic ecosystem, with no option to bring your own model
Benchmark performance FrontierHarness Eval pass rate 76.7%, median time 4 minutes 33 seconds SWE-bench Verified 80.9% (Opus 4.6, highest recorded score)
Pricing model The tool is free and open source; pay based on MiniMax Token Plan or your own API usage Subscription-based: Pro $20/month, Max $100–200/month, or API token-based billing (Sonnet 5 $2/$10 per MTok)
Access points Interactive TUI, headless mode (CI/batch), ACP protocol for IDE integration Terminal CLI, plus editor, desktop, web, and mobile access points
Extensibility Plugins, skills, MCP, subagents, session continuation MCP, subagents (nestable up to 3 levels, 20 concurrent), dynamic workflow orchestration

Application scenarios of MiniMax Code CLI

  • Everyday coding and debugging: In the terminal, let the Agent locate failing tests, analyze root causes, modify code, and automatically rerun tests, forming a "code change—verification" loop.
  • CI/CD and batch tasks: Use headless mode mcode exec to script tasks and integrate them into continuous integration pipelines, nightly batch refactoring, or large-scale code evaluation.
  • Deep IDE Integration: Integrate with editors such as Zed that support the Agent Client Protocol via the ACP protocol, and use Agent capabilities within your familiar IDE environment.
  • Enterprise-Grade Security Review and Compliant Development: Use MIT open-source features to review tool invocation and permission handling logic, and combine sandboxing and permission modes to build auditable enterprise-grade applications within controlled boundaries.