Eyrie

Overview

Eyrie is an open-source desktop developer tool that parses local code repositories to generate and maintain a synchronized, interactive architecture map. Developed by software engineer Badhan Sen, the application runs directly on the user's machine under the AGPL-3.0-only license. It constructs a deterministic code knowledge graph using Tree-sitter and incrementally updates visual layouts as source files are edited and saved.

Key Features

• Local Tree-sitter parsing — Extracts syntax facts across 16 language groups from local checkouts, with the deepest support implemented for TypeScript and JavaScript.
• Deterministic edge resolution — Maps structural relationships strictly from parsed syntax facts with explicit confidence scores, avoiding model-based or probabilistic guesses.
• Incremental live synchronization — Recomputes only affected files and dependent export consumers upon save, animating localized updates into the layout and logging them to a change tape.
• Stable node identification — Derives symbol identities by hashing repository, file path, symbol kind, and qualified name to prevent visual reshuffling during internal code edits.
• Architectural derivation lanes — Filters out presentational wrappers and helper hooks to expose primary architecture components across six request-order lanes.
• Multiple analysis lenses — Offers dedicated analytical views across System, Architecture, Data flow, Sequence, and Health perspectives.
• Model Context Protocol integration — Exposes repository architecture and change-impact queries to external AI coding agents via standard MCP interfaces.

Who It's For

Eyrie is intended for software engineers, technical leads, and system architects who need to navigate, document, and inspect complex codebases locally. It is also designed for developers using AI coding assistants who require pre-commit impact analysis and post-edit verification. The tool suits teams and individual contributors working across macOS, Windows, and Linux environments.

Notable Strengths

Eyrie operates completely on-device without cloud dependencies, storing repository graphs in an inspectable, single-file local SQLite cache. Its deterministic parsing model guarantees that identical commits generate identical graphs every time, avoiding the hallucinations common to generative AI diagramming tools. Additionally, native MCP support allows automated agents to query structural repository facts and diffs through a standard interface.