devlensOSS: local MCP server that maps code into a queryable graph
devlensOSS, from Devlensio, is an open-source MCP server that turns local repositories into a queryable graph for AI coding agents. It indexes components, classes and functions as nodes with typed edges so agents can traverse relationships instead of scanning files. Key functions include native semantic parsing, per-node AI summaries, security ranking, impact analysis and Mermaid diagram export. Engineers and technical leads who use AI-assisted coding gain clearer architectural context and fewer guessing errors.
What tasks can you actually use the tool for?
devlensOSS converts a codebase into a graph where every component, class and function becomes a node connected by typed edges such as CALLS, IMPORTS or IMPLEMENTS. Use cases include semantic search across the project, dependency visualization with Mermaid diagrams, and letting MCP-aware assistants query structure directly rather than reading files sequentially. The server exposes a purpose-built Skill endpoint so agents can ask targeted questions about code relationships.
How accurate are the generated summaries and analyses?
The tool uses native semantic parsing instead of basic regex or tree-sitter heuristics, which produces deeper structural representations of code. Per-node AI summaries supply both business-logic and implementation notes, and integrated security assessments rank nodes by severity. Impact analysis estimates a change's blast radius before editing, so generated guidance is grounded in explicit graph links rather than isolated text snippets.
What inputs and environments does it require?
The server runs on systems with Node.js and integrates with MCP-compatible clients such as Claude Desktop, Cursor, Windsurf and Claude Code. Language support includes JavaScript, TypeScript, Python, Java, Go and Rust, with framework-aware mapping for frameworks like React and Next.js. The distribution includes a suite of 21 MCP tools for discovery, search, traversal and security tasks, so repository indexing and queries happen locally.
Does it fit into existing developer workflows and privacy needs?
The tool adopts a local-first architecture that keeps repository data on the user's machine, which suits teams concerned about exposing source code. It targets software engineers and technical leads who run AI-assisted workflows and need architectural context for onboarding or complex refactors. The project is open-source and positioned to integrate into MCP-centric toolchains rather than replace manual code review.
A practical choice for teams that use MCP agents and require project-aware context
devlensOSS is a practical option for engineering teams that rely on MCP-capable assistants and need clearer, machine-queryable architecture; its open-source, local-first approach supports code privacy and community contribution. Expect an initial setup and an operational learning curve for MCP integration, and treat security rankings and automated summaries as inputs to be validated by human reviewers before making production changes.
