fix(index): honor project configuration in every entry point

README.md

@@ -399,8 +399,15 @@ See [docs/architecture.md](docs/architecture.md#known-limitations) for the full
 
 ## Configuration
 
-The MCP server reads an optional `.codeprism.toml` in the project root when it starts. The CLI
-commands take their options as flags instead (`--languages`, `--workers`, ...).
+The MCP server's startup index, `codeprism index` and the `index_project` MCP tool read an optional
+`.codeprism.toml` in the project root. Explicit indexing flags or tool arguments override the file;
+omitted options inherit it. For example, `index --languages python` overrides the configured language
+list, and `--embeddings` / `--no-embeddings` override `enable_embeddings`. The MCP tool likewise accepts
+`embeddings=true` or `embeddings=false`; omitting it preserves the file's setting.
+
+Parser workers keep their entry-point defaults: CLI and MCP indexing use one process per CPU core
+(`parse_workers = 0`), while library callers default to in-process parsing (`parse_workers = 1`).
+Use `index --workers N` to override the CLI worker count.
 
 ```toml
 [codeprism]

codeprism/cli.py

@@ -105,13 +105,13 @@ def index(
         None,
         "--languages",
         "-l",
-        help="Comma-separated language list (default: python,javascript,typescript,go,rust,java,c,cpp,ruby,php)",
+        help="Comma-separated language list (default: project config, or all supported languages)",
     ),
-    embeddings: bool = typer.Option(
-        False,
-        "--embeddings",
+    embeddings: bool | None = typer.Option(
+        None,
+        "--embeddings/--no-embeddings",
         "-e",
-        help="Also build semantic vector index (requires codeprism[embeddings])",
+        help="Override semantic vector indexing (default: project config; requires codeprism[embeddings])",
     ),
     force: bool = typer.Option(
         False, "--force", "-f", help="Re-parse all files even if unchanged (skip incremental check)"
@@ -124,6 +124,7 @@ def index(
 
     Re-runs are incremental by default: only changed or new files are parsed.
     Use --force to re-parse everything from scratch.
+    Project .codeprism.toml settings apply unless explicitly overridden by flags.
     """
     if languages:
         requested_languages = [language.strip() for language in languages.split(",")]
@@ -148,22 +149,23 @@ def index(
 async def _index(
     path: str,
     languages: str | None,
-    embeddings: bool = False,
+    embeddings: bool | None = None,
     force: bool = False,
     workers: int = 0,
 ) -> None:
     from .core.config import CodePrismConfig
     from .core.graph import GraphEngine
-    from .core.paths import get_db_path
+    from .core.paths import get_db_path, get_project_config_path
     from .core.storage import StorageManager
     from .indexer.project_indexer import ProjectIndexer
 
     langs = [lang.strip() for lang in languages.split(",")] if languages else None
-    config = (
-        CodePrismConfig(languages=langs, enable_embeddings=embeddings, parse_workers=workers)
-        if langs
-        else CodePrismConfig(enable_embeddings=embeddings, parse_workers=workers)
-    )
+    config = CodePrismConfig.load(get_project_config_path(path))
+    if langs is not None:
+        config.languages = langs
+    if embeddings is not None:
+        config.enable_embeddings = embeddings
+    config.parse_workers = workers
 
     db_path = get_db_path(path)
     storage = StorageManager(db_path)
@@ -172,7 +174,7 @@ async def _index(
 
     mode = "[dim](full re-index)[/dim]" if force else "[dim](incremental)[/dim]"
     console.print(f"Indexing [bold]{path}[/bold] {mode}")
-    if embeddings:
+    if config.enable_embeddings:
         console.print("[dim]Embeddings enabled — will build vector index after parsing...[/dim]")
     indexer = ProjectIndexer(graph, storage, config)
     result = await indexer.index(path, force=force)
@@ -193,7 +195,7 @@ async def _index(
             f"{result.edge_count} edges · {result.duration_seconds:.2f}s"
             f"{skipped_note}"
         )
-        if embeddings:
+        if config.enable_embeddings:
             console.print("[green]Semantic index built.[/green] search_symbol now uses embeddings.")
     else:
         console.print(f"[yellow]Completed with {len(result.errors)} error(s).[/yellow]")

codeprism/mcp/server.py

@@ -218,25 +218,28 @@ async def _lifespan(server: FastMCP) -> AsyncIterator[None]:
 async def index_project(
     path: str,
     languages: list[str] | None = None,
-    embeddings: bool = False,
+    embeddings: bool | None = None,
 ) -> dict[str, Any]:
     """Build or rebuild the knowledge graph for a project directory.
 
-    embeddings: also build the vector index for semantic search (requires codeprism[embeddings]).
+    Load the project's .codeprism.toml, then override explicitly supplied languages
+    or embeddings. embeddings: build the vector index for semantic search
+    (requires codeprism[embeddings]); omit to inherit the project setting.
     """
     global _engine
     from ..core.config import CodePrismConfig
     from ..core.graph import GraphEngine
-    from ..core.paths import get_db_path
+    from ..core.paths import get_db_path, get_project_config_path
     from ..core.storage import StorageManager
     from ..indexer.project_indexer import ProjectIndexer
 
     # parse_workers=0: one parser process per core (the server entry point is guarded)
-    cfg = (
-        CodePrismConfig(languages=languages, enable_embeddings=embeddings, parse_workers=0)
-        if languages
-        else CodePrismConfig(enable_embeddings=embeddings, parse_workers=0)
-    )
+    cfg = CodePrismConfig.load(get_project_config_path(path))
+    if languages is not None:
+        cfg.languages = languages
+    if embeddings is not None:
+        cfg.enable_embeddings = embeddings
+    cfg.parse_workers = 0
     db_path = get_db_path(path)
     storage = StorageManager(db_path)
     await storage.initialize()

docs/changelog.md

@@ -8,6 +8,10 @@ Versioning follows [Semantic Versioning](https://semver.org/).
 ## [Unreleased]
 
 ### Fixed
+- CLI `index` and the `index_project` MCP tool now load `.codeprism.toml`, matching the server's
+  startup index. Explicit language and embedding options override the file, while omitted options
+  preserve it; `index --no-embeddings` can disable configured embeddings. Automatic CLI/MCP parser
+  workers and the library's in-process default remain unchanged. (#48)
 - Substring symbol search ranks exact names and prefixes before limiting results. MCP
   `search_symbol` now accepts `limit` (1-500, default 50) and reports filtered `total` and
   `truncated`; kind/project filters apply before pagination. Semantic search keeps its