Labsco
MCP SERVER

Before you change a method or a database column, get the blast radius from a dependency graph of the actual system.

Code Intelligence & Repository Indexing✓ Verified
Summary
Answers the question static analysis in one repository cannot: what else breaks, including the database and the people who own it.

The two impact tools are designed to be called before writing code, and the code-to-database one is the unusual half — a column rename is exactly the change whose consequences do not show up in the source tree. The graph tools go further when the graph tier is deployed: bounded traversal from a seed, a path explanation for why the dependency exists, and an owner lookup so the review lands with the right team.

What it is

A server over Lineai's software dependency data: two impact tools for code and database entities, and six graph tools that query the dependency graph directly.

What you get
  • Impact analysis for a specific method within a class, meant to be run before the edit rather than after
  • Impact analysis across the code-to-database boundary, taking a column, table or view and reporting what touches it
  • Graph search over the dependency graph by text query or identity prefix, scoped and limited
  • Bounded blast-radius traversal from seed nodes, with direction — upstream, downstream or both — and a depth you set
  • Path explanation between two nodes, so you can see why one thing depends on another rather than just that it does
  • A change-scope check that reports whether a proposed change looks safe given the seed nodes and your summary of it
  • Owner lookup for a node, which turns an impacted component into the person who should review it
  • A capabilities call that reports which relationship types, limits and flags the workspace supports, and says plainly when the graph tier is not deployed
Requirements

A Lineai server and an account on it: LINEAI_SERVER_HOST pointing at the instance, LINEAI_USERNAME and LINEAI_PASSWORD for the session, and LINEAI_WORKSPACE_NAME to choose the workspace every tool works against. Published on PyPI as lineai-mcp-server and launched with uvx over stdio; on macOS there is a documented workaround using uv when uvx fails to launch inside some IDEs. The graph tools need those routes deployed on your server — when they are not, the call comes back saying so.

Setup effort

One command plus a key — uvx lineai-mcp-server@latest, then supply credentials