# indxr Roadmap
Planned features and improvements for indxr, roughly in priority order.
## Completed
### Live file watching
Watch filesystem for changes and re-index automatically.
- Standalone: `indxr watch ./project -o INDEX.md`
- MCP integration: `indxr serve --watch` with auto-reindex
### Dependency graph export
File-level and symbol-level dependency graph from existing import/relationship data.
- CLI: `indxr --graph dot|mermaid|json`
- MCP: `get_dependency_graph` tool (scoped by path, file or symbol level, DOT/Mermaid/JSON output)
## Planned
### Complexity metrics / hotspots
Per-function complexity metrics using tree-sitter AST analysis.
- Cyclomatic complexity, max nesting depth, parameter count
- MCP tools: `get_hotspots` (top N most complex functions), `get_health` (codebase-level summary)
- CLI: `--hotspots` flag
### PR-aware structural diffs
Structural diff for GitHub PRs — shows added/removed/modified declarations without reading raw diffs.
- CLI: `indxr diff --pr 42`
- MCP: extend `get_diff_summary` with optional `pr` param
### Cross-file type flow tracking
Track where types flow across function boundaries.
- Extract return types and parameter types from signatures
- Build type-to-functions index (producers and consumers)
- MCP tool: `get_type_flow` — given a type name, show who produces and consumes it
### HTTP+SSE MCP transport
HTTP server with SSE transport alongside existing stdin/stdout JSON-RPC.
- `indxr serve --http :8080`
- Enables multi-client scenarios
### Multi-root / monorepo support
Workspace detection and per-member indexing.
- Detect workspace files (Cargo.toml workspace, package.json workspaces, go.work)
- Scope MCP tools to a specific workspace member
### Semantic code search via embeddings
Optional embedding-based search using a local model (feature-gated).
- Generate embeddings for symbol names, doc comments, and signatures at index time
- MCP tool: `semantic_search` — query by concept, returns ranked symbols
- Fallback to `search_relevant` when embeddings not available