Setting Up Claude Code & Serena MCP on Windows
Lately I’ve been spamming quite a few AI-generated newsletters, which gets a bit repetitive. Since I recently tinkered with the Claude Code & Serena combo and found it pretty solid, I decided to write a post myself for a change of pace.
Introduction
Recently I paid for a Claude Pro subscription, which includes Claude Code. Playing around with it has been fun, and I rate it higher than other agents because it works via the CLI without forcing me to install a new IDE (I’m much more used to JetBrains IDEs than VS Code. Plus, Cursor and Windsurf are annoying because they don’t sync settings across different machines, whereas Claude Code lets you share project configs). Besides, since Claude models are first-party from Anthropic, using them inside Claude Code is probably much better optimized :v
In this post, I’ll walk through setting up Claude Code & Serena on Windows. Although Claude Code already supports Windows natively, when I tried it recently I still frequently ran into errors when creating new files. My guess is that it’s probably due to / and \ in file paths, so for now I still use Claude Code through WSL. In addition, I’ll also show how to set up Serena for a Java project in this post.
Alright, let’s get right to it!!
Installing WSL
This part is easy — just follow the official guide here: https://
TLDR: Run the following command:
wsl --install
Wait for the command to finish, restart your machine, and after rebooting it will install Ubuntu. Follow the prompts to set up a username and password for the distro.
Installing Claude Code
Installing Node.js
Even though Claude Code currently only requires Node.js 18+ and a fresh WSL install gives you Ubuntu 24.04 (whose default nodejs package is already version 18.x), my view is that you should always use the latest Node.js LTS release (currently 22.x) — who knows if Claude Code will bump its requirements later, and this way we’ll still be good to go :v So let’s follow the instructions here: https://
TLDR:
sudo apt-get install -y curl
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt-get install -y nodejs
node -v
Next, we need to set up a global npm prefix so packages are installed into a subdirectory in your home folder, avoiding permission errors when running without root privileges, following this guide: https://
TLDR:
mkdir -p ~/.npm-global
npm config set prefix ~/.npm-global
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc
source ~/.bashrc
Installing Claude Code
Follow the official setup guide here: https://
TLDR:
npm install -g @anthropic-ai/claude-code
Then run claude and log in with your Claude Pro account or Anthropic API_KEY.
Installing Serena
Installing uv
Follow the installation guide here: https://
TLDR:
curl -LsSf https://astral.sh/uv/install.sh | sh
Setting Up Serena in Your Project
Use cd to navigate to your project directory and run the following command:
claude mcp add -s project serena -- uvx --from git+https://github.com/oraios/serena serena-mcp-server --context ide-assistant --project $(pwd)
And that’s it! Every time you run claude inside the project, the Serena MCP server will start automatically (remember to check Claude’s prompt when you run claude, since Claude Code may ask you to confirm which MCP servers to launch).
Setting Up Serena for a Java Project
Because Serena currently has issues with Java projects when running on Linux and macOS — I tested it on a Gradle project and kept running into permission errors where it couldn’t execute the java command, even though calling it directly in the terminal worked fine with no path or chmod issues — there’s a slightly janky workaround for now: run Serena on Windows, then connect the Serena MCP client from WSL over SSE. Here are the details:
On Windows
Install Python
This part is pretty straightforward. Head over to the official site to download and install it: https://
Install uv
TLDR:
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Or:
pip install uv
Run the Serena Server
Navigate to your project folder, open CMD or PowerShell, and run the following command:
uvx --from git+https://github.com/oraios/serena serena-mcp-server --context ide-assistant --project . --transport sse --port 9121
In WSL
Find the Host IP
Find the Windows host IP to access it from WSL following this guide: https://
TLDR:
ip route show | grep -i default | awk '{ print $3}'
Note down the IP in the output. For example, mine is 172.21.0.1.
Note: On WSL1, the IP is simply 127.0.0.1.
Set Up the Serena MCP Client
Use cd to go to your project folder and run:
claude mcp add -s project --transport sse serena http://172.21.0.1:9121/sse
Replace 172.21.0.1 with the IP from the output in the previous step.
After that, running claude will connect to the MCP server running on the Windows host.
Note: For Java projects, Serena takes a while to initialize (sometimes up to a few minutes), so keep an eye on the dashboard (by default at http://127.0.0.1:24282/dashboard/index.html) to monitor its progress.
Conclusion
That wraps up my guide on setting up Claude Code & Serena MCP on Windows. Thanks for reading all the way to the end! If you run into any issues, feel free to leave a comment below and I’ll get back to you. See you in future posts!
P.S.: Right now I’m experimenting with Claude Code on a refactoring task — if it goes well, proves effective, and I’m not too lazy, I’ll write a follow-up post to share how it went hehe :v
Update 29/07/2025:
For large projects, you should run index-project before starting serena-mcp-server, as documented here: uvx –from git+https://github.com/oraios/serena index-project
TLDR:
uvx --from git+https://github.com/oraios/serena serena project index
Comments