1. Conduid
  2. Blockchain
  3. Cortex Suite
MCP server · Blockchain

Cortex Suite

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.

Unclaimed blockchain
44Fair

Scored 18 days ago · breakdown

About Cortex Suite

Cortex Suite is an MCP server published by Artistsyn in the Blockchain category: 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.

The repository has 84 stars and 20 forks, with the last commit 14 days ago. Recent commit activity is the strongest signal that reported issues will get attention.

Install

From source
git clone https://github.com/Artistsyn/cortex_suite

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about Cortex Suite

Powered by Claude · Grounded in docs

I know everything about Cortex Suite. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

Static checks derived from the repository 2 days ago. Heuristics on the source, not an audit.

  • ✓README presentThe repository ships documentation.
  • !License declaredNo license file. Legally you have no right to use it in a product.
  • ✓Tests presentA test suite exists in the repository.
  • ✓Dependencies pinnedA lockfile pins dependency versions.
  • !No dynamic code executionSource scan found eval or shell execution. Review how tool inputs reach it before granting credentials.

1 check has not been run yet.

Releases

v1.0.0Cortex Suite: Frontier Thinking Update · 5 Oct 2026Note: Cortex Suite is currently a source-build project. It is not distributed as a native binary for macOS or Windows, and users must build and configure it for their platform. After extensive iteration, the core workflows are now…

README

Cortex Suite

Gives your AI coding assistant a memory and a map of your code.

  • Remembers what worked and what broke, so mistakes aren't repeated.
  • Looks up your real code instead of guessing names and types.
  • Stays current: change a file and it already knows.

Works with Claude Code, VS Code Copilot and Codex, in 10 languages. Everything stays on your computer.

Install

Easiest: let your AI install it

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.

Or by hand

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).

Is it working?

In any project you've worked in:

~/cortex_suite/cortex/target/debug/cortex scoreboard

Update

git pull in cortex_suite, then run the install again.


Details: TECHNICAL.md · Help: SETUP_HANDOFF.md

README fetched from the repository 2 days ago. The original is authoritative.

Questions

About Cortex Suite

How do I install Cortex Suite?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Cortex Suite safe to use with an AI agent?

Its trust score is 44 out of 100 (fair). It passes 3 of the 5 static security checks run so far; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Cortex Suite still maintained?

Yes — the latest release is v1.0.0 (5 Oct 2026), and the last commit was 14 days ago. The repository has 84 stars.