Vela Intelligence
Vela Intelligence
Structural code intelligence for AI agents. A high-performance MCP server powered by Rust and Tree-sitter.
$ curl -fsSL https://github.com/rsalesas/zephlyn/releases/latest/download/install.sh | sh
Then register: zephlyn install claude-code
Release binaries come from GitHub Releases.
Search symbols, text, and APIs across your whole codebase in milliseconds. See where things are defined, who uses them, and how they connect.
See file structure, type definitions with fields, public APIs, and test coverage — all without reading source code. The AI reads only what it needs.
Kanban board for work items, session state that persists across conversations, file summaries that survive restarts. Pick up exactly where you left off.
AST-aware rename, body edit, and scoped refactor across 16 languages. Preview a unified diff before writing. Skips strings, comments, and substring matches.
File watcher + SQLite symbol index. Changes detected and indexed automatically. Multiple AI sessions can share the same project safely.
Diff against any commit to see what changed, what's tested, what's not. Switch between projects in a monorepo with their own config and knowledge.
outline Every function, type, variable, and constant in a file with line numbers — without reading the whole file.find_symbol Find a symbol definition across the workspace. Uses a persistent background index for instant results.scope_search Text search that returns matches with their enclosing function or class. See context, not just the line.get_api Public symbols with signatures. Structs, classes, and interfaces include the full body — fields visible without reading the file.get_dependencies Import and dependency extraction. Handles Go multi-line imports, Rust use, Python, JS/TS, Java, C/C++.get_tests Find test functions and related test files. Supports Go sibling _test packages, Rust #[test], Python, JS/TS.test_context Analyze test coverage for changed symbols. Uses git diff to find what changed, maps tested vs untested symbols. The AI decides what to read and where to write tests.tree Compact file tree — full paths, language, line counts. Default 500 entry limit. Set json=true for full metadata.hot Most recently modified files, sorted by modification time.changes File changes since a sequence number, powered by the file watcher. Know exactly what changed since your last check.rename_symbol AST-aware workspace rename with unified-diff preview. Skips strings, comments, and substring matches. Set apply=true to write.edit_symbol Replace a symbol's full AST range — signature and body — with new source text. Returns a unified diff; set apply=true to write.smart_refactor Scoped rename: rename an identifier only within the body of a named enclosing function, class, or method. AST-bounded.create_summary Store an LLM-generated file analysis: purpose, key symbols, dependencies, complexity. Persists in the knowledge DB.get_summary Retrieve a stored summary. If stale (minhash similarity), returns file content with instructions to regenerate.snapshot Save, revert, delete, and list project snapshots. APFS copy-on-write clones — instant and zero-cost on macOS.board Kanban board with now/next/later lanes. Track work items with context. Shared or private boards.state Save and restore AI session state — goals, progress, context, open questions. Pick up where you left off across sessions.menu Navigate a project command palette from .zephlyn/menu.json. Run commands, read files, or follow instructions.edit_menu Add, remove, or reorder menu items. Shared and personal (per-user) menu files.config Workspace config with 7 actions: discover (where am I, parent/sibling repos), init, get, set, add/remove paths, reindex.status Project root, version, indexed files/symbols, knowledge entries, board counts, snapshots. Call this first to orient.