cm-context
Ships with cm: "I can't write code. But in 6 months, I shipped 12 real products using AI. CodyMaster is everything I learned — so you don't have to repeat my mistakes." — Tody Le, Head of…
Claude Code
> /plugin marketplace add tody-agent/codymaster> /plugin install cm@codymaster
Codex
$ codex mcp add cm-context --env CM_PROJECT_PATH=YOUR_VALUE -- node ./dist/mcp-context-server.js
Cursor
{ "mcpServers": { "cm-context": { "command": "node", "args": [ "./dist/mcp-context-server.js" ], "env": { "CM_PROJECT_PATH": "YOUR_VALUE" } } } }
Add it to mcp.json.
Gemini CLI
{ "mcpServers": { "cm-context": { "command": "node", "args": [ "./dist/mcp-context-server.js" ], "env": { "CM_PROJECT_PATH": "YOUR_VALUE" } } } }
Add it to settings.json.
GitHub Copilot
$ copilot mcp add cm-context --env CM_PROJECT_PATH=YOUR_VALUE -- node ./dist/mcp-context-server.js
OpenCode
{ "mcp": { "cm-context": { "type": "local", "command": [ "node", "./dist/mcp-context-server.js" ], "environment": { "CM_PROJECT_PATH": "YOUR_VALUE" } } } }
Add it to opencode.json.
Windsurf
{ "mcpServers": { "cm-context": { "command": "node", "args": [ "./dist/mcp-context-server.js" ], "env": { "CM_PROJECT_PATH": "YOUR_VALUE" } } } }
Add it to mcp_config.json.
Security report
Not enough data
Known flaws
Not published
Signed build
Not published
Maintained
Aug 25, 2026
Last push
Licence
Not published
Runs as
Local process
On your machine
Deprecated
Not published
Signals from public registries, not an audit. Checked Nov 11, 2026.
cm README
CodyMaster
"I can't write code. But in 6 months, I shipped 12 real products using AI. CodyMaster is everything I learned — so you don't have to repeat my mistakes." — Tody Le, Head of Product, Creator of CodyMaster
50+ skills. One install. Your AI coding agent becomes a full team.
v7.0.2 — Browse Hybrid Bridge: AI-native browser automation with a11y snapshots, error collection, video recording.
( . \ --- / . )
/ ^ ^ \
( u )
| \ ___ / |
'--w---w--'
Meet Cody 🐹
The Problem
You installed an AI coding agent. It writes code fast. But then:
- It designs differently every time — your brand looks like 3 different companies
- It fixes one thing, breaks five others — you redo the same work over and over
- It forgets everything between sessions — you re-explain your project every morning
- It writes zero tests, zero docs — your codebase becomes a ticking time bomb
- You install 15 skills — none of them talk to each other
"AI gave me 100 hands. Without discipline, those hands created chaos."
The Solution: A Full Senior Team in One Install
CodyMaster isn't a collection of random skills. It's an operating system for your AI agent — 50+ skills that work together like a real team.
When you install CodyMaster, you hire:
| Role | What They Do | Key Skills |
|---|---|---|
| Senior Developer | Write tests first, debug systematically, review code | cm-tdd cm-debugging cm-code-review |
| UX Designer | Extract design systems, preview UI before coding | cm-design-system cm-ux-master cm-open-design |
| Product Manager | Plan features, brainstorm ideas, understand users | cm-planning cm-brainstorm-idea |
| DevOps Engineer | Safe deploys, secret scanning, right-account checks | cm-safe-deploy cm-identity-guard |
| Tech Writer | Auto-generate docs, API refs, knowledge base from code | cm-dockit cm-codeintell |
| Automation Lead | Sprint pipelines, autopilot workflows, status tracking | cm-sprint-bus cm-autopilot |
Quick Start — 30 Seconds
Step 1: Install
npm install -g codymaster && cm
A wizard detects your AI tools (Claude Code, Cursor, Gemini, etc.) and installs skills to all of them.
Step 2: Start coding
Open your AI agent and say what you want:
Build a landing page for my coffee shop
Step 3: Watch it work
CodyMaster automatically chains the right skills:
cm-brainstorm-idea → cm-design-system → cm-execution → cm-safe-deploy
That's it. No config files. No setup ritual. Just describe what you want.
Disciplined Delivery by Default
CodyMaster now carries one execution contract from planning through verification:
- Rich plans name exact task files and keep every step inside that task scope; TDD work must record RED before GREEN.
- One approved plan grants scoped execution authorization, so in-scope work continues without repeated confirmation loops.
- Mode B uses three distinct sessions per task: a fresh implementer, a spec reviewer, and then a quality reviewer.
- The shared autonomy policy is validated and synchronized across all 14 supported platform distributions.
cm bench --suite workflow-integration --runs 1runs from any project directory, including an installed npm package, and writes results locally.
See It Work
Here's a real conversation — a non-technical founder building a booking site:
You: I want to build a booking website for my spa business.
I have an old website at myspa.com but it looks terrible.
AI: [cm-brainstorm-idea] Let me analyze your current site first...
I found 3 key issues: no mobile booking, unclear pricing,
slow load time. Here are 2 approaches:
A) Redesign from scratch (recommended — your brand needs it)
B) Fix the existing site
You: Let's go with A.
AI: [cm-design-system] I'm extracting your brand colors from
myspa.com... Found: gold + navy. Building a design system.
[cm-planning] Here's the plan: booking page, services page,
about page, contact. Estimated: 30 minutes of AI work.
You: Sounds good, do it.
AI: [cm-tdd] Writing tests first...
[cm-execution] Building 4 pages...
[cm-dockit] Generating service descriptions and FAQ docs...
[cm-quality-gate] All tests pass. Performance: 98/100.
[cm-safe-deploy] Deploying to staging...
✅ Done. Preview: https://staging-myspa.pages.dev
You: The booking button is too small on mobile.
AI: [cm-debugging] Found it — padding was 8px, should be 16px.
[cm-code-review] Fix looks good. No side effects.
[cm-safe-deploy] Deployed fix.
✅ Fixed. Try now.
No code written by the human. No bugs introduced by the AI. The system caught the mobile issue before it reached production.
What Makes This Different
| 😵 Random Skills | 🧠 CodyMaster | |
|---|---|---|
| Integration | Each skill is standalone | 50+ skills that chain and share memory |
| Memory | Forgets everything between sessions | Remembers your project, your style, your mistakes |
| Safety | Deploy and pray | Multi-layer protection: tests → security → staging → production |
| Design | Random UI every time | Extracts your brand, enforces consistency |
| Documentation | "Maybe later" | Auto-generates docs from your code |
| Self-improvement | Static — what you install is what you get | Learns from feedback, gets better over time |
Built For Builders, Not Coders
CodyMaster was built for people who have ideas, not CS degrees.
You're a good fit if you:
- Want to type a message and get a working product back
- Want your AI to learn from its mistakes
- Don't want to manually review every line of AI output
- Want to deploy with confidence, not prayer
You're NOT a good fit if you:
- Enjoy doing the same setup for every project
- Prefer slow, manual deploys with no safety net
- Like re-explaining your codebase every morning
The 10-Second Tour
Instead of listing 50+ skills, here's what CodyMaster does in plain language:
💡 Before You Build
- Thinks before coding — asks questions, challenges your assumptions, saves you from building the wrong thing
- Plans the architecture — diagrams, data flow, edge cases — all before writing code
🎨 While Building
- Enforces your design system — consistent colors, fonts, spacing across every page
- Writes tests first — catches bugs before they exist
- Debugs systematically — traces the root cause, doesn't guess
🚀 Before Shipping
- Scans for secrets — no API keys leaking to GitHub
- Checks the right account — no accidental deploys to the wrong Cloudflare
- Runs quality gates — no deploy without passing tests + performance checks
📈 After Shipping
- Generates documentation — reads your code, writes the docs
- Learns from feedback — gets smarter with every project
- Runs retrospectives — captures what worked, what didn't, improves the process
The Design Pipeline
CodyMaster treats design as a first-class discipline — not an afterthought. Your AI doesn't just write code; it enforces visual consistency across every page.
Three ways to get a design system:
| Method | When to Use | How |
|---|---|---|
| Extract from URL | You have an existing brand/site | cm-open-design analyzes your site, extracts colors, fonts, spacing |
| Pick from 129 systems | You want a proven aesthetic | Choose from Linear, Stripe, Vercel, Notion, Apple, Tesla, and 123 more |
| Choose a direction | You have no brand yet | Pick from 5 curated styles: Editorial, Modern Minimal, Warm Soft, Tech Utility, Brutalist |
Extract → Tokens → Validate → Build → QA
│ │ │ │ │
│ │ │ │ └─ cm-quality-gate
│ │ │ └─ cm-execution
│ │ └─ cm-ux-master (48 UX laws)
│ └─ cm-design-system (STITCH_TOKENS)
└─ cm-open-design (129 systems + extraction)
📖 Full design pipeline guide →
Real-World Scenarios
| You Say | What Happens |
|---|---|
| "Fix this bug" | cm-debugging finds root cause → cm-tdd writes test → cm-quality-gate verifies |
| "Build a landing page" | cm-brainstorm-idea → cm-design-system → cm-execution → cm-safe-deploy |
| "Make it look like Stripe" | cm-open-design extracts Stripe tokens → cm-design-system applies → cm-execution builds |
| "Deploy to production" | cm-secret-shield → cm-security-gate → cm-identity-guard → cm-safe-deploy |
| "What does this code do?" | cm-codeintell reads the codebase → cm-dockit generates docs |
| "Add Vietnamese support" | cm-safe-i18n extracts strings → translates → validates → ships |
| "Start a new project" | cm-project-bootstrap scaffolds → cm-planning plans → cm-execution builds |
Install
One command
npm install -g codymaster && cm
What the wizard does
- Detects every AI tool you have (Claude Code, Cursor, Gemini CLI, Codex, OpenCode, Windsurf, Cline, Aider, Continue, Kiro, Amazon Q, Amp, Copilot, Claude Desktop)
- Asks which ones you want skills for (pre-checked for detected tools)
- Installs to each platform's native location
- Done — open your AI agent and start building
After install
cm doctor # Check what's installed
cm status # See your tasks and progress
cm dashboard # Open the visual dashboard
For OpenCode users
OpenCode supports both skills and plugins:
git clone https://github.com/tody-agent/codymaster.git ~/.cody-master
# Skills (auto-load)
ln -s ~/.cody-master/.opencode/skills ~/.opencode/skills
# Plugins (optional — adds custom tools)
mkdir -p ~/.opencode/plugins
ln -s ~/.cody-master/.opencode/plugins/cm-brainstorm-idea.ts ~/.opencode/plugins/
Restart OpenCode to activate. The cm-brainstorm-idea plugin adds a strategic analysis tool with auto-detection for brainstorm keywords.
📖 Full OpenCode installation guide →
For Codex users
Fastest path:
git clone https://github.com/tody-agent/codymaster.git ~/.cody-master
cd ~/.cody-master
npm ci
npm run build:platforms
Then tell Codex:
Fetch and follow instructions from https://raw.githubusercontent.com/tody-agent/codymaster/main/.codex/INSTALL.md
Or use the local Codex skill tree directly at .codex/skills/.
No Node.js?
bash <(curl -fsSL https://raw.githubusercontent.com/tody-agent/codymaster/main/install.sh) --all --profile core
What's in the package
- 50+ skills — ready-to-use SKILL.md files for every major AI coding agent
- CLI (
cm) — install wizard, doctor, dashboard, MCP server, browse daemon - Skill profiles — curated bundles:
core,growth,full,knowledge - Multi-platform — installs to Claude Code, Cursor, Gemini, Codex, OpenCode, Windsurf, Cline, Aider, Continue, Kiro, Amazon Q, Amp, Copilot, Claude Desktop
- Zero config — detects your tools, installs skills, you're done
The Dashboard
Visual mission control for your projects:
More MCP servers from other plugins.
Questions about cm-context.
- Which coding agents can use cm-context?
- cm-context works in any agent that reads an MCP config file: Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf. Each agent reads it from its own file; this page names the file for each one.
- How do I add cm-context to Claude Code?
- Run these in order, one at a time. 1. In Claude Code: /plugin marketplace add tody-agent/codymaster 2. In Claude Code: /plugin install cm@codymaster

