Omen

You can run Omen’s MCP server to expose all analyzers as tools for large language models. This enables AI assistants to query code health, surface hotspots, and perform automated code analysis through standardized tool calls.

How to use

To use the MCP server, start the Omen CLI with the mcp command and connect your MCP-enabled client (such as Claude Desktop or Claude Code) to the exposed tools. You can request analyses like the most complex functions, hotspot files, defect risk, or a full repository health score. Tools return structured, compact results that help you reason about code quality without manual sifting.

How to install

Prerequisites: you need Rust installed to build from source, or you can install the prebuilt binaries. You also need a terminal with Git and a network connection for fetching dependencies.

Install the CLI via Cargo if you want to build from source or use the prebuilt binaries for quick setup.

# Install via Cargo (build from source)
cargo install omen-cli

# Alternatively, download a prebuilt binary and place it in your PATH

Additional notes

The MCP server is designed to be used with MCP clients that can issue tool calls in a standard format. Your client should be configured to communicate with the MCP server using the command and arguments shown below.

MCP server configuration (stdio)

The MCP server runs locally as a standard I/O server. The runtime command to start the MCP server is the Omen executable with the mcp argument.

Tools available through the MCP server

The MCP server exposes a broad set of analyzers as tools for LLMs. Each tool returns a structured result and guidance on interpretation. Available tools include but are not limited to complexity, SATD, dead code detection, churn, clones, defect probability, change risk, diff risk, TDG scores, dependency graph, hotspot analysis, temporal coupling, ownership metrics, CK metrics, repository map, architectural smells, code ownership, flags, health score, and semantic search. Use them to build a comprehensive understanding of a codebase directly from your MCP client.

MCP server usage example with Claude</heading,{

Configure your MCP client to connect to the server at the defined stdio endpoint. Then ask questions such as: which functions are most complex, where are hotspots, or what is the overall health score of the repository.

Available tools

complexity

Returns cyclomatic and cognitive complexity for functions or files and guidance on reducing complexity.

satd

Detects self-admitted technical debt in comments and groups findings by category such as Design, Defect, Requirement, Test, Performance, and Security.

deadcode

Identifies dead code such as unused functions, variables, or unreachable branches to clean up.

churn

Analyzes git history to measure file change frequency and identify hotspots.

clones

Detects code clones across the repository to highlight duplication patterns.

defect

Predicts file-level defect probability using PMAT-weighted signals.

changes

Performs Change Risk Analysis at the commit level, including JIT factors.

diff

Evaluates branch diffs against a target branch to assess overall risk.

tdg

Calculates Technical Debt Gradient scores for files to guide refactoring.

graph

Generates a dependency graph showing module connectivity and centrality metrics.

hotspot

Identifies hotspots where high churn combines with high complexity.

temporal

Finds files that change together, revealing temporal coupling and hidden dependencies.

ownership

Assesses code ownership and bus factor by analyzing authorship and contributor counts.

cohesion

Computes CK metrics related to coupling and cohesion in OO codebases.

repomap

Provides a PageRank-ranked symbol map to help prompt context in LLMs.

smells

Detects architectural smells relative to codebase size to guide architecture reviews.

flags

Detects feature flags and staleness across providers and custom config.

score

Delivers a composite health score for the repository with adjustable thresholds.

semantic_search

Performs natural language code search using semantic embeddings.

mutation

Runs mutation testing and reports mutation score, operators, and survivors.

flags

Detects and analyzes feature flags across the codebase.