RsNLME Installation and Configuration Troubleshooting Guide • Certara.RsNLME

Installation Errors/Warnings in R

ERROR: compilation failed for package ‘…’

You may also see the additional error:

c:/rtools40/mingw32/bin/g++: No such file or directory

This error can occur when installing a package dependency from source. R will always attempt to install the Windows binaries first (for Windows users), but in some cases, the binaries on CRAN may have version older than the source files.

In such cases, R will prompt you with a message -

“Do you want to install from source the packages which need compilation?”

The correct option is to click No. Clicking Yes may require the additional installation of RTools to install the package successfully, if that package dependency must be compiled from source.

Note: For non-recent R versions CRAN removes binaries from the server and the only way to install the dependent package is to install it from source.

Warning in install.packages:

Warning in install.packages :
  package ‘Certara.RSNLME’ is not available for this version of R

If you have confirmed that R >= 4.0 is installed (use the command R.Version() to check) and still receive this error, ensure that the name of the package is specified correctly.

Note: Package names are case-sensitive.

Warning in install.packages :
  unable to access index for repository https://certara.jfrog.io/artifactory/certara-cran-release-public/bin/windows/contrib/4.4

This warning implies that the R package binaries for the given OS are unavailable and can be ignored. R will subsequently install the packages from source successfully.

WARNING: RTools is required …

RTools is required to build R packages and/or compile libraries from source on Windows, however, the installation of RTools is NOT required to use RsNLME.

If this warning is received, and the packages were not successfully installed:

Option 1: Check R Library Path

Execute the following command to check the location of R libraries:

.libPaths()

Note: If your R Library path is located on a network drive e.g., OneDrive, you may experience issues installing R packages. The solution is to create a local folder for your R libraries. Click here to learn more.

Option 2: Install missing package separately

If you’re still unable to install one or more Certara_Packages, even after confirming that your R library path isn’t on a network drive, the problem is likely indicated by an error message in your R console.

Example Error Messages:

namespace ‘rlang’ 0.4.2 is already loaded, but >= 0.4.5 is required

ERROR: lazy loading failed for package ‘rlang’

Troubleshooting Steps:

Restart R: To ensure a clean slate, restart your R session. This will unload all currently loaded packages.
Install the Package Directly: Try installing the problematic package (e.g., “rlang”) on its own using the following command in R:

install.packages("rlang")

Reboot and Manually Remove: If the direct installation still fails:
- Reboot your system.
- Find the package’s folder within your R library directory (you can find this directory using the .libPaths() command in R).
- Delete the package’s folder manually.
- Attempt to reinstall the package.

Note: These troubleshooting steps focus on resolving issues with individual package installations. If you continue to encounter difficulties, please contact Certara support for further assistance.

Then proceed with the installation of Certara_Packages.

Installation Help for NLME Engine

The NLME Engine installer for Windows will automatically install and configure the following components:

MS MPI 10.0
GCC 8.4
NLME Engine 26.9.1

What if I have some components previously installed?

If you have an existing Phoenix installation, you may receive a message stating that GCC and/or MPI is already installed and configured on your system.

If the installer detects that these components are already installed, you will be prompted to skip the installation of these components.

vcruntime140_1.dll not found

DLLs (Dynamic-Link Libraries) are shared libraries in Microsoft Windows, implemented by Microsoft Corporation, and used by the NLME Engine for model execution. If you receive an error such as “vcruntime140_1.dll Not Found”, the file may be corrupted or missing from your system.

The solution is to download the Microsoft Visual C++ Redistributable packages for Visual Studio 2015, 2017, 2019, and 2022 here, installing both X86 and X64 Architecture.

Note: You will be required to restart your computer after installing Microsoft C++ Redistributable Packages

Modify/Uninstall

If you have run the NLME Engine installer previously, you will be prompted with the following options:

Modify Available Options
Repair Application
Uninstall

Select Modify Available Options to re-install components, or, Uninstall to remove components from system.

Check environment variables

To ensure that all components have been installed successfully, open a new session of R/RStudio and execute the following:

Sys.getenv("INSTALLDIR") # NLME Engine path
Sys.getenv("NLMEGCCDir64") # GCC path
Sys.getenv("PhoenixMSMPIDir") # MSMPI path if MSMPI was installed

If the NLME Engine installation has been successful, the above system paths should be returned.

Note: If GCC and/or MSMPI was installed by Phoenix and not using the NLME Engine installer, Sys.getenv("NLMEGCCDir64") and Sys.getenv("PhoenixMSMPIDir") will return a different folder - e.g., ‘“C:\PHSTMinGW64”’ and "C:\\Program Files (x86)\\Certara\\MPI\\". They could be used as well.

To ensure that all components have been installed successfully, open a new session of R/RStudio and execute the following:

Sys.getenv("INSTALLDIR") # NLME Engine path
Sys.getenv("PhoenixMPIDir64") # OpenMPI path if OpenMPI was installed

If the NLME Engine installation has been successful, the above system paths should be returned.

To ensure that all components have been installed successfully, open a new session of R/RStudio and execute the following:

Sys.getenv("INSTALLDIR") # NLME Engine path
Sys.getenv("PML_BIN_DIR") # Enables access to libraries compiled specifically for Ubuntu 22.04
Sys.getenv("PhoenixMPIDir64") # OpenMPI path if OpenMPI was installed

If the NLME Engine installation has been successful, the above system paths should be returned.

MCP server setup troubleshooting

The optional Certara MCP server exposes RsNLME to AI coding agents (Cursor, Claude Code, Codex, and Claude Desktop). Setup friction here is largely an environment problem – the table below collects the common failures and the fix.

For the full setup and usage guide (supported clients, tool profiles, live-session tools, and Claude Desktop on Windows), see the Certara MCP Server: Setup and Usage vignette. The same checklist is available to agents through the knowledge base (Certara.R.mcp.setup_troubleshooting).

Symptom	Likely cause	What to try
MCP tools missing / server not listed	`write_mcp_config()` not run, or the client was not reloaded	Run `Certara.R::write_mcp_config(client = "cursor", scope = "project")`; reload the client; in Cursor, Settings -> Tools & MCPs and enable the Certara R server
`ellmer` / `mcptools` / `btw` not found	A development install with `repos = NULL` skips `Suggests`	Install from the release repository, or install the MCP stack directly: `install.packages(c("btw", "mcptools", "ellmer"))`. Version floors: `btw >= 1.2.0`, `mcptools >= 0.2.0`, `ellmer >= 0.4.0`
Cursor “tool not found” when calling the server	Using the `mcp.json` key instead of the `CallMcpTool` id	Cursor prefixes user servers with `user-`, e.g. `user-certara-r`; check `client_routing` in `certara_mcp_capabilities()`
Fit/VPC “can’t find the model”	Confusing the MCP server’s R process with an interactive session, or acting on the wrong model	Call `certara_session_status()` to see which R process holds state; confirm which fit to use and enumerate runs with `list_mcp_runs()`
Engine / compile failures (RStudio / terminal)	`INSTALLDIR` / NLME not resolvable	Confirm the NLME Engine install and that `Sys.getenv("INSTALLDIR")` is set (see the environment-variable checks above)
MCP-launched fit fails with `Compile/Link Failed` on Claude Desktop (Windows), but the same model fits from RStudio	Claude Desktop sanitizes the MCP server’s environment, dropping `ComSpec`, `INSTALLDIR`, and `NLMEGCCDir64`	Add those variables to the `env` block of `claude_desktop_config.json` (see below), then fully quit and relaunch Claude Desktop
Agent behavior feels out of date	The installed package predates the current MCP knowledge base/rules	Install the current release and call `certara_mcp_capabilities()` to confirm the advertised rules
Too many tools listed	Normal when `btw` tool groups are enabled	Use the workflow phases in `certara_mcp_capabilities()$workflows`; for the leanest server set `btw_groups = "docs"`

# Re-write the client configuration and confirm the live wiring
Certara.R::write_mcp_config(client = "cursor", scope = "project")
Certara.R::certara_session_status()

Claude Desktop on Windows: extra environment variables

Claude Desktop launches MCP servers with a sanitized, minimal environment. On Windows that snapshot is inherited by every MCP-launched NLME fit, so variables that are set in RStudio or a terminal never reach TDL5/gcc. The usual symptom is Compile/Link Failed / ERROR in generating Model.cpp even though the same model fits from RStudio.

Add an env block to the certara-r server in claude_desktop_config.json with at least:

ComSpec – C:\WINDOWS\system32\cmd.exe (this is the one that unblocks Compile/Link Failed)
INSTALLDIR – NLME Engine install directory
NLMEGCCDir64 – Certara-shipped GCC 8.4
TMP, NUMBER_OF_PROCESSORS, and PATHEXT (commonly dropped as well)

Then fully quit Claude Desktop (system tray -> Quit, not just closing the window) and relaunch. A sample env block, the MSIX config-path gotcha, and a diagnostic job are in Claude Desktop on Windows: required environment variables.

Claude Code (CLI) generally does not need this – a terminal-launched server inherits the shell environment.

There is no automated “doctor” command - setup spans the client, the R environment, and the engine install, so the checklist above (and certara_session_status() for the live wiring) is the supported path.

FAQ

Do existing Phoenix users need to install the NLME Engine?

Yes. Certara.RsNLME version 3.2.0 requires NLME-Engine-26.9.1. Existing Phoenix users must install the NLME-Engine executable, separate from NLME-Engine that is embedded in Phoenix application.

Note: Phoenix NLME-Engine is kept in a separate location and installing NLME-Engine-26.9.1 will not affect your existing Phoenix installation.

Is the Certara.NLME8 R package distributed in a Phoenix release compatible with RsNLME?

It is recommended to maintain a distinct R version (or separate R library) that ensures the version of the Certara.NLME8 package that you installed (given the installation instructions provided in your version of Phoenix) is not automatically updated when installing Certara.RsNLME.