How iamstoick/argus works
Argus builds a per-project structural index of supported source files, then exposes codebase search and analysis to AI coding agents over MCP. The principal workflow is workspace sync → language parsing and relationship extraction → SQLite storage → tool queries and results. Serve mode also offers an authenticated admin UI/API; hybrid search can add an optional Ollama semantic tier. The README documents additional tool and UI capabilities; some implementation details were not sampled, so uncertain internal links are omitted.
Main entry points: http.ts HTTP server and API · projects.ts Project manager · indexer.ts File sync and watcher · tools.ts Dictionary and analysis tools
Components 14
Indexing
- Project manager
src/projects.ts - File sync and watcher
src/indexer.ts - Polyglot parser
src/parser.ts - Language query packs
src/queries.ts - Project SQLite index
src/db.ts
Query and analysis
- Dictionary and analysis tools
src/tools.ts
Agent and admin access
- Runtime entry
src/index.ts - MCP tool server
src/server.ts - HTTP server and API
src/http.ts - Admin web UI
src/admin/index.html
Semantic search
- Embedding provider and sync
src/embeddings.ts
Other components
- AI coding agent
- Project workspace
- Ollama
Connections 20
- AI coding agent → MCP tool server: sends MCP requests
- AI coding agent → HTTP server and API: uses admin/API
- Project workspace → File sync and watcher: provides source files
- Runtime entry → Project manager: opens projects
- Runtime entry → MCP tool server: starts stdio MCP
- Runtime entry → HTTP server and API: starts HTTP
- Project manager → File sync and watcher: syncs and watches
- File sync and watcher → Polyglot parser: parses changed files
- Polyglot parser → Language query packs: uses language queries
- File sync and watcher → Project SQLite index: writes index data
- Project manager → Project SQLite index: opens project index
- Project manager → Embedding provider and sync: syncs vectors
- MCP tool server → Dictionary and analysis tools: dispatches MCP tools
- HTTP server and API → MCP tool server: creates MCP endpoint
- HTTP server and API → Dictionary and analysis tools: serves analysis data
- HTTP server and API → Admin web UI: serves UI
- Dictionary and analysis tools → Project SQLite index: queries index
- Dictionary and analysis tools → Embedding provider and sync: requests semantic matches
- Embedding provider and sync → Project SQLite index: reads and writes vectors
- Embedding provider and sync → Ollama: probes and embeds