Preserve user MCP env block; unify home resolution across the bridge

ClaudeR 0.13.1 / clauder-mcp 0.14.5. Two pre-existing bugs, both found
while auditing an unrelated plan.

configure() replaced the whole r-studio entry with a fresh command/args
list, so re-running install_clauder() deleted any env block on it. That
silently dropped CLAUDER_AGENT_ID and collapsed multi-agent identity
back to a random id, the failure that 0.11.0 set out to fix. The entry
is now updated in place and unknown fields survive.

The bridge resolved the session discovery directory through USERPROFILE
on Windows to match the R side, but _coord_dir still used a plain
expanduser("~"). With HOME pointed at a OneDrive-redirected Documents
folder, R and the bridge wrote coordination logs to different places:
send_message reported success, check_messages stayed empty, and
wait_for_message blocked until timeout with nothing to report. Both
paths now go through one _home_dir() helper.

R CMD check: Status OK. checks.R and pytest 37/37 pass.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: IMNMV

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: ClaudeR
 Title: R Integration for Claude AI
-Version: 0.13.0
+Version: 0.13.1
 Authors@R: person("Nykko", "Vitali", email = "nykvt@icloud.com", role = c("aut", "cre"))
 Description: Connects RStudio with Claude AI to enable interactive coding sessions.
 License: MIT + file LICENSE

R/setup.R

@@ -105,10 +105,14 @@ configure <- function(for_cursor = FALSE, use_uvx = TRUE, python_path = NULL) {
 
 
   # --- Add or update the r-studio server entry ---
-  config$mcpServers$`r-studio` <- list(
-    command = mcp_command,
-    args = mcp_args
-  )
+  # Update command and args in place. Anything else the user put on this entry
+  # (most importantly an env block carrying CLAUDER_AGENT_ID) must survive, so
+  # do not replace the whole entry.
+  existing <- config$mcpServers$`r-studio`
+  if (!is.list(existing)) existing <- list()
+  existing$command <- mcp_command
+  existing$args <- mcp_args
+  config$mcpServers$`r-studio` <- existing
 
 
   # --- Write the updated config back to the file ---

clauder-mcp/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "clauder-mcp"
-version = "0.14.4"
+version = "0.14.5"
 description = "MCP server connecting AI assistants to RStudio for interactive R coding and data analysis"
 readme = "README.md"
 requires-python = ">=3.10"

clauder-mcp/server.json

@@ -7,12 +7,12 @@
     "url": "https://github.com/IMNMV/ClaudeR",
     "source": "github"
   },
-  "version": "0.14.4",
+  "version": "0.14.5",
   "packages": [
     {
       "registryType": "pypi",
       "identifier": "clauder-mcp",
-      "version": "0.14.4",
+      "version": "0.14.5",
       "transport": {
         "type": "stdio"
       }

clauder-mcp/src/clauder_mcp/server.py

@@ -30,13 +30,18 @@
 # On Windows, prefer USERPROFILE explicitly. Python's expanduser already does
 # this internally, but being explicit guards against odd HOME settings (e.g.
 # OneDrive-redirected Documents) and matches the R-side discovery_dir().
-if sys.platform == "win32":
-    SESSIONS_DIR = os.path.join(
-        os.environ.get("USERPROFILE") or os.path.expanduser("~"),
-        ".claude_r_sessions",
-    )
-else:
-    SESSIONS_DIR = os.path.expanduser("~/.claude_r_sessions")
+def _home_dir() -> str:
+    """Home directory, resolved the same way the R side resolves it.
+
+    Every path shared with R must go through this. R uses path.expand("~"),
+    which on Windows follows USERPROFILE, so a HOME pointing at OneDrive would
+    otherwise put the two halves in different folders."""
+    if sys.platform == "win32":
+        return os.environ.get("USERPROFILE") or os.path.expanduser("~")
+    return os.path.expanduser("~")
+
+
+SESSIONS_DIR = os.path.join(_home_dir(), ".claude_r_sessions")
 _agent_id: Optional[str] = None       # Set in main()
 _agent_id_source: str = "unset"       # Where the identity came from (for the intro)
 _target_session: Optional[str] = None  # Set by connect_session tool
@@ -329,7 +334,7 @@ def _coord_dir() -> str:
     get_r_addin_url()  # latch a session if not bound yet
     session = _target_session or "default"
     safe = re.sub(r"[^a-zA-Z0-9_-]", "_", session)
-    d = os.path.join(os.path.expanduser("~"), ".clauder_coord", safe)
+    d = os.path.join(_home_dir(), ".clauder_coord", safe)
     os.makedirs(d, mode=0o700, exist_ok=True)
     return d
 

clauder-mcp/uv.lock

@@ -30,13 +30,18 @@
 # On Windows, prefer USERPROFILE explicitly. Python's expanduser already does
 # this internally, but being explicit guards against odd HOME settings (e.g.
 # OneDrive-redirected Documents) and matches the R-side discovery_dir().
-if sys.platform == "win32":
-    SESSIONS_DIR = os.path.join(
-        os.environ.get("USERPROFILE") or os.path.expanduser("~"),
-        ".claude_r_sessions",
-    )
-else:
-    SESSIONS_DIR = os.path.expanduser("~/.claude_r_sessions")
+def _home_dir() -> str:
+    """Home directory, resolved the same way the R side resolves it.
+
+    Every path shared with R must go through this. R uses path.expand("~"),
+    which on Windows follows USERPROFILE, so a HOME pointing at OneDrive would
+    otherwise put the two halves in different folders."""
+    if sys.platform == "win32":
+        return os.environ.get("USERPROFILE") or os.path.expanduser("~")
+    return os.path.expanduser("~")
+
+
+SESSIONS_DIR = os.path.join(_home_dir(), ".claude_r_sessions")
 _agent_id: Optional[str] = None       # Set in main()
 _agent_id_source: str = "unset"       # Where the identity came from (for the intro)
 _target_session: Optional[str] = None  # Set by connect_session tool
@@ -329,7 +334,7 @@ def _coord_dir() -> str:
     get_r_addin_url()  # latch a session if not bound yet
     session = _target_session or "default"
     safe = re.sub(r"[^a-zA-Z0-9_-]", "_", session)
-    d = os.path.join(os.path.expanduser("~"), ".clauder_coord", safe)
+    d = os.path.join(_home_dir(), ".clauder_coord", safe)
     os.makedirs(d, mode=0o700, exist_ok=True)
     return d