mcp-cs
MCS (mcp-cs) is a universal Model Context Protocol server and developer toolkit that lets AI assistants, a terminal CLI, and a visual dashboard perform competitive programming, project diagnostics, database modeling, web performance, security, API testing, code hygiene, and Git/CI automation.
Algorithm execution & competitive programming: run sandboxed code in Python/JS/TS/C++/C/Go, execute WASM modules, differential stress-test solutions against brute force, check code plagiarism via AST token similarity, and generate adversarial edge cases.
Project & environment diagnostics: diagnose project setup issues, compare
.envvs.env.example, and inspect which processes/PIDs occupy network ports.Database & ERD generation: parse SQL DDL and generate Mermaid entity-relationship diagrams.
Web performance auditing: find oversized raster images and estimate WebP/AVIF savings, and audit HTTP compression, caching, and security headers.
Security & secrets management: scan repositories for leaked API keys/tokens and auto-sanitize hardcoded secrets into
process.envreferences.API testing & mock data: send live HTTP requests with latency measurement and generate realistic mock datasets for users, products, posts, and transactions.
Documentation & code hygiene: check/fix broken markdown links, extract code snippets, locate TODO/FIXME/HACK/BUG comments, and flag heavy npm dependencies.
Git, release & CI/CD workflows: generate changelogs, check PR readiness, create PR descriptions, fetch Codeforces problems, and generate GitHub Actions workflows.
Agent skills & interactive tools: install 30 skills.sh-compatible operational runbooks, use slash commands like
/diagnose-repoand/prepare-pr, and access live resources for ports and git status.
Runs C++ code in a sandboxed process with strict timeout and memory limits for algorithmic testing.
Generates GitHub Actions CI/CD workflows for Docker containers and images.
Generates changelogs, checks PR readiness, and generates PR titles and descriptions from git repository data.
Supports PR workflow by validating repository health and generating pull request descriptions, as well as generating GitHub Actions workflows.
Generates production GitHub Actions CI/CD workflows for various technologies including Next.js, Node.js, Docker, and Python.
Checks for broken links in Markdown documentation, auto-fixes them, and extracts code snippets from Markdown files.
Generates interactive Mermaid entity-relationship diagrams from SQL DDL schemas for visual database exploration.
Generates GitHub Actions CI/CD workflows for Next.js projects.
Executes JavaScript/TypeScript code in a sandboxed environment and generates CI/CD workflows for Node.js projects.
Runs Python code in a sandboxed process with strict time and memory limits, and generates CI/CD workflows for Python projects.
Runs raw WebAssembly modules in-process with sub-millisecond execution for algorithmic evaluation.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-cs stress test my quicksort implementation against bubble sort"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
⚡ MCS (mcp-cs)
The Universal Campus Developer, Diagnostics & AlgoJudge Competitive Execution Suite
MCS (Model Context Server for Computer Society) is a full-featured Model Context Protocol (MCP) server, standalone developer CLI, and visual console created for developers, university students, competitive programmers, and engineering teams.
Installation • Terminal CLI • Visual Console • Client Configuration • Tool Catalog
🌟 Why MCS?
MCS connects your AI assistant (Antigravity, Cursor, Claude Desktop, Windsurf, Cline) directly to powerful local execution, repository diagnostics, database modeling, and competitive algorithm tools with zero manual configuration.
⚡ AlgoJudge Engine: Run sandboxed code (Python, JS/TS, C++, Go, Docker micro-containers), perform automated differential stress-testing, and check AST code similarity.
🩺 Environment Doctor: Diagnose project setups, detect
.envdisparities, and inspect port conflicts with process PIDs.🗄️ Database & ERD: Parse SQL schemas and auto-generate Mermaid Entity-Relationship diagrams.
⚡ Web Performance: Scan for oversized raster assets and audit HTTP compression & security headers.
🛡️ Security Guard: Detect leaked API keys, tokens, and hardcoded credentials before committing.
🌐 API Playground: Live HTTP test client and realistic synthetic mock data generator.
Related MCP server: FastApply MCP Server
💻 Standalone Terminal CLI Commands
In addition to serving AI assistants, MCS works as a standalone terminal productivity CLI:
# 1. Run instant project diagnostics & .env synchronization check
mcs doctor
# 2. Inspect active network ports and process PIDs
mcs ports
# 3. Scan repository for leaked secrets and API keys
mcs scan
# 4. Generate Mermaid ER diagram from a SQL file
mcs erd schema.sql
# 5. Launch the embedded visual web dashboard
mcs ui
# 6. Launch the official Model Context Protocol Inspector
mcs inspector🖥️ Interactive Visual Dashboard
Launch the embedded graphical visualizer locally with either command:
mcs ui
# or
npx mcp-cs --ui📦 Installation Guide
Method 1: Instant Run via npx (Recommended — No install needed)
{
"mcpServers": {
"mcs": {
"command": "npx",
"args": ["-y", "mcp-cs"]
}
}
}Method 2: Homebrew Installation (macOS & Linux)
brew tap ieeecsopen/tap
brew install mcsMethod 3: Global NPM Installation
npm install -g mcp-cs🎛️ Modular Tool Filtering
To save prompt tokens in your AI assistants, enable only the specific tool modules you need:
# Enable only algorithm and doctor tools
MCS_MODULES=algo,doctor npx mcp-cs
# Or via CLI arguments
mcp-cs --modules=algo,security,dbSupported module tags: algo, doctor, security, db, perf, api, docs, code, git, problem, ci, wasm.
🔌 Client Configuration
1. Antigravity IDE
Add mcs to your global configuration file at ~/.gemini/config/mcp_config.json:
{
"mcpServers": {
"mcs": {
"command": "npx",
"args": ["-y", "mcp-cs"]
}
}
}2. Claude Desktop
Add to your Claude configuration file (~/Library/Application Support/Claude/claude_desktop_config.json on Mac or %APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"mcs": {
"command": "npx",
"args": ["-y", "mcp-cs"]
}
}
}3. Cursor IDE
Open Settings (
Cmd + ,on Mac orCtrl + ,on Windows).Go to Features $\to$ MCP.
Click + Add New MCP Server:
Name:
mcsType:
commandCommand:
npx -y mcp-cs
🛠️ Complete Tool Catalog (25 Tools)
⚡ 1. AlgoJudge Competitive & Algorithm Engine
Tool | Description | Example Prompt |
| Runs code in Python, JS, TS, C++, C, or Go in an isolated process with strict time and memory limits. | "Run this C++ solution with the test input |
| Runs code inside a zero-trust ephemeral Docker micro-container ( | "Run this untrusted script inside a Docker sandbox." |
| Automated Differential Tester: Compares an optimal algorithm against a brute-force baseline on randomized inputs to find failing edge cases. | "Stress-test my Dijkstra implementation against a brute-force solution to find where it fails." |
| Token-based AST code similarity checker that compares two code submissions and detects suspicious duplicated logic. | "Check if |
| Generates adversarial corner cases (min $N=1$, max constraints, identical items, extreme negatives, disconnected graphs). | "Generate edge testcases for a binary tree problem." |
🗄️ 2. Database & Schema Suite
Tool | Description | Example Prompt |
| Parses SQL | "Generate a Mermaid ERD diagram from this |
⚡ 3. Web Performance Suite
Tool | Description | Example Prompt |
| Scans directory for uncompressed raster images (>250KB) and calculates bandwidth savings from WebP/AVIF conversion. | "Audit the images in |
| Audits HTTP response headers of a web application for Gzip/Brotli compression, Cache-Control, and security headers. | "Check the performance and security headers on |
🩺 4. Environment & System Doctor
Tool | Description | Example Prompt |
| Auto-detects runtime (Node, Next.js, Python, Docker), missing | "Diagnose why this repo won't start." |
| Diffs | "Check if my |
| Identifies which background processes and PIDs are locking ports ( | "Check why port 6379 is occupied." |
🛡️ 5. Security & Secrets Guard
Tool | Description | Example Prompt |
| Scans codebase for leaked OpenAI keys, AWS access tokens, GitHub PATs, and private keys. | "Scan my repository for accidentally committed secrets." |
| Auto-replaces detected secrets with | "Sanitize all hardcoded keys across the project." |
🌐 6. API Client & Mock Generator
Tool | Description | Example Prompt |
| Executes live HTTP calls ( | "Send a GET request to |
| Generates realistic synthetic mock datasets ( | "Generate 10 mock user records with IDs and emails." |
📚 7. Documentation & Code Hygiene
Tool | Description | Example Prompt |
| Scans all markdown files ( | "Check all markdown files in the repo for broken links." |
| Automatically updates and rewrites broken markdown links when target files have moved. | "Fix broken links in the |
| Extracts all fenced code blocks from markdown documentation for quick inspection. | "Extract all code snippets in |
| Scans the codebase for | "Find all |
| Inspects | "Check if we have heavy dependencies slowing down our bundle." |
🚀 8. Git, Release & CI/CD Suite
Tool | Description | Example Prompt |
| Parses conventional git commit history between tags and outputs a clean markdown changelog. | "Generate a changelog for all commits since |
| Verifies uncommitted files, unpushed commits, and branch health before opening a Pull Request. | "Check if my branch is ready for a Pull Request." |
| Generates a structured markdown Pull Request description summarizing branch commits and diffs. | "Generate a structured PR description for my current branch." |
| Fetches problem statements, sample inputs/outputs, tags, and contest limits from Codeforces. | "Fetch the details for Codeforces problem 2060A." |
| Generates production-ready GitHub Actions CI/CD workflows for Next.js, NPM Publishing, Docker, or Python. | "Generate a GitHub Action workflow to automatically publish this package to NPM." |
🧠 Agent Skills Suite (30 Skills • Compatible with skills.sh)
MCS bundles 30 production agent skills adhering to the open skills.sh standard. These provide AI assistants (Antigravity, Cursor, Claude Code, Windsurf) with specialized operational runbooks:
# 1-Click Install all 30 skills into any project (.agents/skills/)
mcs skills install
# Or install for Cursor (.cursor/rules/)
mcs skills install --cursor
# Or install via skills.sh open package manager
npx skills add ieeecsopen/mcp-cs🎓 1. Campus & Student Engineering (8 Skills)
Skill Name | What It Instructs the AI Agent to Do |
| Plans and documents university Final Year Projects (FYP) with C4 architecture diagrams, tech comparisons, and IEEE Std 830 SRS chapters. |
| Formats online software, journals, DOIs, and papers into strict IEEE referencing format. |
| Auto-generates structured IEEE 2-column practical lab reports and experiment write-ups. |
| Guides student group projects on feature branching, rebase strategies, and conventional commits. |
| Generates 10-minute slide outlines and word-for-word viva defense speaking scripts. |
| Pre-flight LMS assignment validator (strips |
| 10-minute rapid prototyping architecture for 12h/24h hackathons (SLIITXtreme). |
| Generates step-by-step ASCII memory traces for LinkedLists, BST trees, DP tables, and Graphs. |
⚡ 2. Competitive Programming & Algorithms (4 Skills)
Skill Name | Description |
| Differential testing against brute-force baselines to find edge-case failures. |
| Generates corner cases ($N=1$, bounds, extremes, disconnected graphs). |
| AST token similarity analysis between code submissions. |
| Asymptotic Big-O time and space complexity auditing. |
🎨 3. Frontend & UI Engineering (2 Skills)
Skill Name | Description |
| Modern, high-contrast, accessible SaaS UI design with Tailwind and dark mode tokens. |
| State architecture using Zustand, TanStack Query, and React Server Components. |
🗄️ 4. Backend, Databases & APIs (4 Skills)
Skill Name | Description |
| Production REST API design with Zod schemas, status codes, and cursor pagination. |
| Zero-downtime database schema migrations, indexes, and safe alterations. |
| Parses SQL DDL statements into visual Mermaid ER diagrams. |
| Generates realistic synthetic datasets (users, products, transactions). |
🤖 5. AI, Testing & DevOps (12 Skills)
Skill Name | Description |
| Retrieval-Augmented Generation (RAG) pipelines with vector embeddings and reranking. |
| System prompt engineering, red-teaming, and structured JSON output validation. |
| Vitest / Jest unit test suites with parameterized edge cases and 100% coverage. |
| Resilient Playwright browser test specs with user-facing accessibility locators. |
| Secure multi-stage Dockerfiles with layer caching and non-root users. |
| Cloud hosting configuration for Vercel, Supabase, and Render. |
| Identifies broken runtime dependencies, missing packages, and setup errors. |
| Compares and synchronizes |
| Resolves socket collisions ( |
| Generates GitHub Actions CI/CD workflows for Node, Docker, and Python. |
| Pre-flight scanner to detect and auto-sanitize leaked API credentials. |
| Scans markdown files for broken internal links and dead references. |
🧩 Interactive Prompts & Resources
Slash Commands / Prompts:
/diagnose-repo: Runs full system doctor diagnostics, checks environment variables, scans ports, and generates an onboarding report./stress-test-solution: Interactive algorithm stress-testing assistant to find counter-example edge cases./prepare-pr: Pre-flight PR creation checklist and summary.
Dynamic Context Resources:
resource://system/ports: Live socket feed of active network ports.resource://git/status: Real-time repository branch and sync state.
💻 Local Development
# 1. Clone repository
git clone https://github.com/ieeecsopen/mcp-cs.git
cd mcp-cs
# 2. Install dependencies
npm install
# 3. Run automated tests (23 tests)
npm test
# 4. Build TypeScript
npm run build
# 5. Start in development mode with UI
npm run dev -- --ui📄 License & Community
Distributed under the MIT License.
Maintained with ❤️ by the IEEE Computer Society of SLIIT.
Available Tools
28 toolsalgo_check_plagiarismA
Compares two code submissions using token-level AST analysis and returns similarity percentage and verdict
Inputs: codeA (required), codeB (required).
algo_generate_edge_casesB
Generates adversarial corner cases (min N=1, max bounds, identical items, extreme negatives, disconnected graphs)
Inputs: maxN, type.
algo_run_dockerA
Executes code inside a zero-trust ephemeral Docker container (python:alpine, node:alpine, gcc:alpine) with no network access
Inputs: code (required), stdin, language, timeoutMs.
algo_run_sandboxedB
Executes code in an isolated process with strict timeout and memory limits (supports python, javascript, typescript, cpp, c, go)
Inputs: code (required), stdin, language, timeoutMs.
algo_stress_testA
Automated differential tester: runs a fast solution against a brute-force baseline using randomized inputs until a failing edge case is found
Inputs: solutionCode (required), generatorCode (required), bruteForceCode (required), language, +1 more.
api_generate_mock_dataC
Generates realistic synthetic mock data (users, products, posts, transactions) for testing
Inputs: type, count.
api_test_requestC
Executes an HTTP request (GET, POST, PUT, DELETE, PATCH) and measures latency & formats response data
Inputs: url (required), body, method, headers, +1 more.
ci_generate_workflowB
Generates production-ready GitHub Actions CI/CD workflows for Next.js, NPM Publishing, Docker, or Python
Inputs: type (required).
code_find_todosA
Scans the codebase for TODO, FIXME, HACK, and BUG comments with line numbers and file paths
Inputs: targetDir.
code_inspect_heavy_dependenciesA
Inspects package.json to identify bloated dependencies (moment, lodash, monaco, katex) with optimization tips
Inputs: projectPath.
db_generate_erdA
Parses SQL table definitions and generates a clean Mermaid Entity-Relationship (ER) Diagram
Inputs: schemaSql (required).
docs_auto_fix_linksB
Automatically updates and rewrites broken markdown links when target files have moved
Inputs: dryRun, targetDir.
docs_check_broken_linksA
Scans all markdown files (.md, .mdx) in the project to detect broken internal links and dead file references
Inputs: targetDir.
docs_extract_code_snippetsB
Extracts all fenced code blocks from markdown documentation for quick inspection and review
Inputs: targetDir.
doctor_check_envA
Compares .env against .env.example and flags missing or extra environment keys
Inputs: projectPath.
doctor_diagnose_projectA
Scans project structure, detects language/framework, missing dependencies, and setup issues
Inputs: projectPath.
doctor_port_inspectA
Checks if specific network ports (e.g. 3000, 6379, 5432) are occupied and by which process PID
Inputs: ports (required).
git_generate_changelogA
Parses conventional git commit history between tags and outputs a clean markdown changelog
Inputs: toTag, fromTag.
git_generate_pr_descriptionA
Generates a structured markdown Pull Request description summarizing branch commits and diffs
Inputs: targetBranch.
git_pr_readiness_checkA
Verifies uncommitted files, unpushed commits, and branch health before opening a Pull Request
Inputs: no named parameters.
perf_audit_assetsA
Scans repository for oversized raster images (>250KB) and calculates bandwidth savings from WebP/AVIF conversion
Inputs: targetDir, thresholdKb.
perf_check_headersA
Audits HTTP response headers of a web application for Gzip/Brotli compression, Cache-Control, and security headers
Inputs: url (required).
problem_fetch_codeforcesA
Fetches problem statement metadata, tags, and contest limits from Codeforces
Inputs: index (required), contestId (required).
security_auto_sanitizeA
Auto-replaces detected hardcoded secrets with process.env references and populates .env.example
Inputs: dryRun, targetDir.
security_scan_secretsA
Scans repository text files for accidentally hardcoded API keys, JWTs, and private tokens
Inputs: targetDir.
skills_installA
Installs official MCS agent skills (algo-stress-testing, repo-doctor, db-architect, security-auditor) into local repository (.agents/skills/ or .cursor/rules/)
Inputs: format, targetDir.
wasm_run_moduleC
Executes pre-compiled WebAssembly (.wasm) binary in memory with exported function calls
Inputs: wasmBase64 (required), functionName (required), args.
wasm_run_vm_sandboxA
Executes JavaScript in an isolated Node.js VM context with strict execution timeout and zero filesystem access
Inputs: code (required), timeoutMs.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 28 tools
Category prefixes like security_, algo_, and docs_ mostly keep tools organized, and most pairings (scan/sanitize, check/fix) have distinct roles. A couple of execution tools are easy to confuse, especially `wasm_run_vm_sandbox` (which runs JavaScript, not WebAssembly) alongside `wasm_run_module`, and `algo_run_sandboxed` vs `algo_run_docker` require a close read.
The overwhelming majority follows a `category_verb_noun` snake_case pattern, e.g. security_scan_secrets, ci_generate_workflow, docs_check_broken_links. Minor deviations like `git_pr_readiness_check`, `doctor_port_inspect`, and `security_auto_sanitize` break the pattern slightly but don't make the set unpredictable.
With 28 tools spanning security, algo, wasm, db, perf, doctor, api, docs, code, git, problem, skills, and ci, this exceeds a reasonable single-server scope and feels like several focused servers merged. Category prefixes mitigate the chaos, but the count is too high for a coherent tool collection.
The server has useful workflows in a few areas, such as secret scan→sanitize, docs check→fix links, and algorithm stress testing, but many domains are represented by only one-off tools (db_generate_erd, problem_fetch_codeforces, skills_install). There are notable lifecycle gaps: git stops at PR description instead of PR creation/merging, and CI generates a workflow but does not validate or update it.
Maintenance
Related MCP Connectors
Code intelligence platform for AI agents. 20 tools for architecture, security & impact analysis.
30+ deterministic dev and delivery tools: linting, metrics, diagrams, model advice, AI-text checks.
Checks AI-written code for leaked keys and made-up packages, on your machine. Never uploads code.
Build Apps and run code in 30 languages — sandboxed, with persistent sessions for agent loops.
Related MCP Servers
- AlicenseNot gradedqualityNot gradedmaintenanceProvides unified development tools including code analysis, debugging, refactoring, documentation, testing, and project automation through multiple LLM providers (KIMI, GLM, OpenRouter). Features agentic audit capabilities with multi-model consensus for finding issues and generating direct fixes.Apache 2.0
- FlicenseCqualityDmaintenanceEnterprise-grade code intelligence platform providing AI-powered code analysis, semantic search, security scanning, and automated refactoring capabilities. Integrates with local AI models for zero-cost operations while delivering comprehensive development workflow automation.28-
- AlicenseNot gradedqualityDmaintenanceProvides AI-powered development tools including code generation, refactoring, debugging, performance optimization, and test generation, along with smart prompts for code analysis and documentation, and a built-in knowledge base of coding best practices.MIT
- AlicenseCqualityDmaintenanceProvides intelligent error detection and debugging capabilities across multiple programming languages with real-time monitoring of build, lint, runtime, console, and test errors. Offers AI-enhanced error analysis with automated resolution suggestions and context-aware debugging.2MIT