接入你的智能体
从发布者处安装。不是 PlugMCP 端点。
远程 URL 还是本机命令？ 两种安装方式的区别
来自发布者
Gives your AI coding assistant a memory and a map of your code.
Local MCP servers that give coding agents a reliable project memory. quartz-ctx serves live code structure in 10 languages, cross-language calls included; cortex keeps learned patterns and traps, pushes them into the agent's context at edits and failing builds, and measures outcomes and token cost from real sessions. No network, no API keys.
Works with Claude Code, VS Code Copilot and Codex, in 10 languages. Everything stays on your computer.
Don't have an AI coding assistant yet? Install Claude Code.
Then copy the prompt for your assistant and paste it in. It installs Cortex Suite for all your projects.
Install Cortex Suite for all of my projects.
1. If `cargo` isn't installed, install Rust from https://rustup.rs first.
2. Clone https://github.com/Artistsyn/cortex_suite into ~/cortex_suite
(if it's already there, run `git pull` in it instead).
3. From inside ~/cortex_suite, run ./scripts/install-global.sh
If I have more than one Claude Code account (look in my shell profile for
aliases that set CLAUDE_CONFIG_DIR), add `--claude-config <folder>` for each
folder that isn't already a ~/.claude-* folder.
4. Show me the cortex and quartz-ctx lines from `claude mcp list`, then tell me
to restart Claude Code.
Install Cortex Suite so you can use it in all of my projects. Run each step in
the terminal and check it worked before moving on.
1. Check the tools it needs: `git --version`, `cargo --version`, and a C
compiler (on a Mac: `xcode-select -p`; on Linux: `cc --version`).
- If cargo is missing, install Rust from https://rustup.rs, then open a new
terminal so cargo is on the PATH.
- If the C compiler is missing: on a Mac run `xcode-select --install`; on
Linux install build-essential.
2. Clone https://github.com/Artistsyn/cortex_suite into ~/cortex_suite
(if it's already there, run `git pull` in it instead).
3. From inside ~/cortex_suite, run ./scripts/install-global.sh
It builds two MCP servers (cortex and quartz-ctx) and adds them to my VS Code
user MCP config (mcp.json in VS Code's User folder), so they work in every
workspace. The build takes a few minutes.
4. Open that user mcp.json (Command Palette: "MCP: Open User Configuration")
and confirm it has "cortex" and "quartz-ctx" entries.
5. Tell me to reload the window, then run "MCP: List Servers" and start both,
and to turn their tools on in the agent's tool picker.
6. If this workspace is a git repo, also run
~/cortex_suite/cortex/target/debug/cortex hooks-init --vscode
so cortex can warn you about known mistakes as you work.
I'm on Windows? The install script needs macOS or Linux. Instead, run
.\scripts\setup.ps1 -Workspace <project folder> from ~/cortex_suite for each
project.
I want to install Cortex Suite (https://github.com/Artistsyn/cortex_suite).
It's a pair of local MCP servers, "cortex" (memory) and "quartz-ctx" (code
map), that give AI coding assistants like Claude Code, VS Code Copilot and the
Codex CLI a memory and a map of my code. You can't run commands on my computer,
so walk me through it: give me ONE step at a time, tell me exactly what to type
and what I should see, and wait for me to paste the result before the next step.
If something fails, help me fix it before moving on.
First ask me which operating system I use (macOS, Linux or Windows) and which
assistants I use (Claude Code, VS Code Copilot, Codex CLI).
The steps, for macOS and Linux:
1. Open a terminal.
2. Check git: `git --version`. If it's missing: on a Mac, `xcode-select
--install`; on Linux, install git with the package manager.
3. Check a C compiler: on a Mac `xcode-select -p` (if it errors, run
`xcode-select --install` and wait for it to finish); on Linux
`cc --version` (if missing, `sudo apt install build-essential` or the
distro's equivalent).
4. Check Rust: `cargo --version`. If it's missing, run
`curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh`, accept the
defaults, then close the terminal and open a new one, and check again.
5. Download it: `git clone https://github.com/Artistsyn/cortex_suite ~/cortex_suite`
(if the folder already exists: `cd ~/cortex_suite && git pull`).
6. Install it for every project: `cd ~/cortex_suite && ./scripts/install-global.sh`
The build takes a few minutes. It then registers both servers with every
assistant it finds: each Claude Code account, VS Code's user MCP config, and
~/.codex/config.toml. Its output says which ones it set up. If I use more
than one Claude Code account through CLAUDE_CONFIG_DIR, add
`--claude-config <that folder>` for each.
7. Check it worked, for each assistant I use:
- Claude Code: `claude mcp list` should show cortex and quartz-ctx as
Connected. Then restart Claude Code.
- VS Code Copilot: reload the window, run "MCP: List Servers" from the
Command Palette, start both, and turn their tools on in agent mode.
- Codex CLI: start a new codex session; the servers load from config.toml.
For Windows, the install script doesn't run. Instead, after steps 2-5 (Git for
Windows, Visual Studio Build Tools with the C++ workload, Rust from
https://rustup.rs, and the clone), run in PowerShell from the cortex_suite
folder, once per project:
`.\scripts\setup.ps1 -Workspace C:\path\to\project`
Common problems: "cargo: command not found" means the terminal needs
reopening after installing Rust. A build error about "cc" or "linker" means the
C compiler from step 3 is missing. A warning about mcp.json means VS Code's
config has comments in it; then add the two servers by hand, with commands
~/.cortex-suite/bin/cortex-mcp and ~/.cortex-suite/bin/quartz-ctx-mcp.
You need Rust.
git clone https://github.com/Artistsyn/cortex_suite ~/cortex_suite
~/cortex_suite/scripts/install-global.sh
Then restart your editor. Or, to set up just one project:
~/cortex_suite/scripts/setup.sh ~/code/my-project
(on Windows: .\scripts\setup.ps1 -Workspace C:\code\my-project).
In any project you've worked in:
~/cortex_suite/cortex/target/debug/cortex scoreboard
git pull in cortex_suite, then run the install again.
Details: TECHNICAL.md · Help: SETUP_HANDOFF.md
Local MCP servers that give coding agents a reliable project memory. quartz-ctx serves live code structure in 10 languages, cross-language calls included; cortex keeps learned patterns and traps, push
概览
Local MCP servers that give coding agents a reliable project memory. quartz-ctx serves live code structure in 10 languages, cross-language calls included; cortex keeps learned patterns and traps, pushes them into the agent's context at edits and failing builds, and measures outcomes and token cost from real sessions. No network, no API keys.
工具
暂无工具列表。远程服务器在首次健康检查后会获得工具列表。
常见问题
在本页复制适用于 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 配置。它使用发布者的安装方式，而不是 PlugMCP 代理。
此记录未列出必需的环境变量。
Cortex Suite 通过 stdio 在本地运行。
上次检查时间：2026-11-05。 Local MCP servers that give coding agents a reliable project memory. quartz-ctx serves live code structure in 10 languages, cross-language calls included; cortex keeps learned patterns and traps, pushes them into the agent's context at edits and failing builds, and measures outcomes and token cost from real sessions. No network, no API keys.
