Project Scope
Project Scope MCP
Semantic code intelligence and safe repository control for AI agents.
Project Scope is a production-grade MCP server that gives AI agents deep, structural understanding of codebases and the ability to modify them safely. It goes beyond file reading by combining AST analysis, context-aware security scanning, safe refactoring, and Git operations.
Built for real-world agentic development workflows.
What it does
Project Scope enables AI agents to:
- Understand code semantically, not textually
- Analyze dependency graphs and change impact
- Perform safe, preview-first refactors
- Scan repositories for real security risks without false positives
- Interact with Git repositories programmatically
- Operate under strict security controls
Core features
Semantic repository analysis Uses Tree-sitter to parse source code into ASTs and build symbol graphs, dependency maps, and reference trees.
Context-aware security scanning Security engine understands where code lives and how it is used.
- Skips source code syntax that resembles attacks
- Strictly validates runtime inputs
- Zero false positives on legitimate code
Safe refactoring tools
- Symbol rename across project via AST
- Function extraction with syntax guarantees
- Mandatory preview before apply
Git-native operations
- Repository status and history
- Branch comparison
- Initialization and inspection
- Designed for agent-driven workflows
Defense-in-depth security
- Prompt injection detection
- Regex + heuristic guards
- Optional LLM Guard integration
- File access and execution boundaries
Why Project Scope
- Eliminates naive regex-based security scanners
- Prevents agents from breaking builds during refactors
- Makes large codebases understandable to LLMs
- Designed for autonomous and semi-autonomous agents
- MCP-native and language-agnostic at the protocol level
Available tools
Repository analysis
get_repo_structure, analyze_impact
Security
scan_repo_for_threats, scan_file_for_threats, validate_shell_input
Refactoring
refactor_rename, refactor_extract_function
Git
git_status, git_commit_history, git_compare_branches, git_init
Security model
Project Scope uses context-sensitive scanning instead of blanket rules.
- Source code is parsed, not pattern-matched
- Comments, imports, and definitions are ignored safely
- Runtime inputs are strictly validated
- Configurable strict and advisory modes
Optional integration with Meta Llama Prompt Guard 2 for advanced prompt injection detection.
Use cases
- Large codebase understanding
- Safe autonomous refactoring
- Secure agent-driven code changes
- Repository audits
- Change impact analysis
- AI-assisted code reviews
Tech stack
- TypeScript
- Node.js
- Tree-sitter
- FastMCP
- Fully local execution
Status
Stable. Security-focused. ISC licensed.
Server Config
{
"mcpServers": {
"project-scope": {
"command": "npx",
"args": [
"-y",
"@adityasasidhar/project-scope-mcp"
]
}
}
}