Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

cursor-agents

Make Cursor Claude Again

Turn Cursor into a multi-agent powerhouse. Autonomous workflows, parallel research, specialized subagents — all running natively inside your IDE. No external process. No API key.

For developers who can't access Claude Code (or prefer Cursor), this project bridges the gap by bringing Claude-style autonomous agent capabilities directly into your existing IDE.

Core Features

Multi-Agent Orchestration

The agent decomposes complex tasks, spawns specialized subagents in parallel, synthesizes results, and delivers polished output — just like Claude Code's autonomous mode.

Skill Packs (Extensible)

Domain-specific capabilities installed as plug-and-play modules. Install what you need, ignore what you don't.

Zero Config

One bash install.sh and everything works in every project. No docker, no daemon, no API keys.

Quick Start

git clone https://github.com/cocolwy/cursor-agents.git
cd cursor-agents
bash deploy/install.sh

Open any project in Cursor and start using commands.


Use It in 30 Seconds

After install, open any project in Cursor and type these prompts directly in the chat:

Build a feature end-to-end:

/code Add user authentication with JWT — register, login, and refresh token endpoints

→ Planner designs the approach → Coder implements → Tester writes & runs tests → Reviewer checks quality → auto-fix if needed.

Refactor existing code:

/refactor Extract the database logic from controllers into a repository layer

Research then execute:

/goal Evaluate whether we should migrate from Express to Fastify, then do it if it makes sense

→ Defines success criteria → parallel research → plans → executes → verifies the goal is actually achieved → if not, iterates with a different approach (up to 5 rounds). Doesn't stop until the goal is met or proven infeasible.

Financial modeling:

/comps Tesla NIO BYD — comparable company analysis
/dcf Apple — DCF valuation, assume 12% revenue growth over 5 years

Bootstrap a new project:

/init

→ Scans your codebase and generates .cursor/rules/project-context.mdc so Cursor understands your project in every future conversation.

Create a structured PR:

/pr

→ Analyzes your diff, drafts title + summary + test plan, shows you the draft, then creates the PR.

Daily workflow (review → commit):

/review
/commit

/review checks your changes for bugs and security issues. /commit generates a quality commit message and commits.

Generate tests:

/test src/utils/parser.ts

→ Analyzes the file, writes comprehensive tests (happy path + edge cases + errors), runs them.

That's it. No config files, no setup, no API keys — just prompt and go.


Core Commands

These work in any project, for any task:

Command What it does
/code <task> Plan → Implement → Test → Review → Fix
/refactor <task> Analyze → Refactor → Verify → Review
/goal <task> Research → Plan → Execute → Verify until done
/init Scan project → Generate context rules for Cursor
/doctor Check installation health → Report issues
/pr Analyze diff → Draft PR → Create via gh CLI
/review Review uncommitted changes → Structured feedback
/commit Analyze diff → Generate message → Commit
/test <file> Analyze code → Write tests → Run them

Examples

/code Add WebSocket support to my Express server with reconnection logic

Spawns a planner agent to design the approach, a coder agent to implement, runs tests, then a reviewer agent checks for bugs and security issues.

/refactor Extract the authentication logic into a reusable middleware

Analyzes dependencies, refactors safely, verifies tests pass, reviews for regressions.

/goal Research the best approach for migrating from REST to GraphQL in our codebase

Launches 3 parallel research agents (schema design, migration strategy, performance implications), synthesizes findings into an actionable plan.


Skill Packs

Financial Analysis Pack

Institutional-grade financial modeling with Excel output. Adapted from Anthropic's financial-services.

Command What it does
/comps <companies> Comparable company analysis → .xlsx
/dcf <company> DCF valuation model → .xlsx
/lbo <company> Leveraged buyout model → .xlsx
/competitive <company> Competitive landscape analysis
/comps NVIDIA vs AMD vs Intel

Parallel research agents gather financials from SEC filings, then a financial analyst agent builds a formatted Excel comps table with operating metrics, valuation multiples, and statistical summary.

/dcf AAPL

Builds a 5-year DCF model with WACC, terminal value, and sensitivity tables — all as live Excel formulas.

More Packs Coming

  • DevOps/deploy, /monitor, /incident (planned)
  • Data Science/eda, /pipeline, /visualize (planned)
  • Content/blog, /docs, /translate (planned)

Want to build a skill pack? See Contributing.


Architecture

~/.cursor/
├── skills/                    # Workflow definitions (SKILL.md files)
│   ├── agent-system/          # Master orchestrator (core)
│   ├── comps-analysis/        # ── Financial Pack ──
│   ├── dcf-model/             #
│   ├── lbo-model/             #
│   ├── competitive-analysis/  #
│   └── xlsx-author/           # Excel generation utility
└── agents/                    # Subagent role definitions
    ├── planner.md             # Architecture & decomposition
    ├── coder.md               # Implementation specialist
    ├── reviewer.md            # Code review & quality
    ├── researcher.md          # Research & data gathering
    └── financial-analyst.md   # Financial modeling (Finance Pack)

How it works:

  1. You type a command (e.g., /code add caching)
  2. Cursor's agent reads the agent-system skill → becomes the supervisor
  3. Supervisor spawns specialized subagents via Cursor's built-in Task tool
  4. Subagents work in parallel where possible
  5. Supervisor synthesizes results and presents final output

No external runtime. No daemon. The intelligence lives in the skill definitions.

Installation

Prerequisites

  • Cursor IDE installed
  • Any Cursor subscription (skills use your existing model access)

Install

bash deploy/install.sh

Uninstall

rm -rf ~/.cursor/skills ~/.cursor/agents

Advanced: Programmatic Interface

For CI/CD pipelines or external scripts, the project also includes a TypeScript orchestrator using @cursor/sdk:

npm install
export CURSOR_API_KEY=your_key
npx tsx src/main.ts code "Add rate limiting to the API"

This is optional — the primary use case is the native in-IDE experience.

Contributing

Contributions welcome! Here's how to add a new Skill Pack:

  1. Create a directory under deploy/skills/your-skill-name/
  2. Write a SKILL.md with YAML frontmatter (name, description) and workflow instructions
  3. If your pack needs a specialist agent, add it to deploy/agents/
  4. Register your commands in deploy/skills/agent-system/SKILL.md
  5. Run bash deploy/install.sh to test
  6. Submit a PR

Areas where help is needed:

  • New skill packs (DevOps, data science, content creation)
  • Better data sourcing for financial skills
  • Localization (Chinese docs, A-share market support)
  • Real-world workflow testing

License

MIT


Star History Chart

About

MCCA (Make your Cursor Claude Again)

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages