Run, debug, test

Inside a capsule, your editor, build tool, test runner, and interactive debugger all execute inside the same Linux container against /workspace/<slug>. You do not configure remote SSH interpreters or Docker run targets inside the IDE: to the IDE, the capsule is the local machine.

Running and debugging by stack

Start with the IDE's normal Run or Debug configuration. The capsule drops Linux capabilities and enables no-new-privileges by default. Debugger requirements depend on the language, extension, and whether you start a program or attach to an existing process.

• Python in PyCharm (python-ide): Create or reuse a project virtual environment in the IDE terminal (python3 -m venv .venv). Install the dependencies specified by your project, then select .venv/bin/python in PyCharm (Settings → Project → Python Interpreter), and click Run or Debug in the editor gutter on any module or pytest test.
• Node.js and TypeScript in VSCodium or VS Code (frontend-ide, vscode-ide): Run npm install in the integrated terminal, open Run and Debug (Ctrl+Shift+D), and choose JavaScript Debug Terminal or launch an npm script directly with breakpoints set in your .ts/.js files. For Python, Java, or C# in VSCodium/VS Code, install the language extension once; it persists in your checkout's extensions slot.
• Java and Maven in IntelliJ IDEA (java-ide) or Eclipse (eclipse-ide): Temurin Java (JAVA_HOME) and Maven (MAVEN_HOME) are on PATH from the base image. Open pom.xml as a project in IntelliJ—or import /workspace/<slug> in Eclipse via File → Import → Maven → Existing Maven Projects—set breakpoints, and debug JUnit tests or main classes directly. Dependency downloads under /home/devcapsule/.m2 persist across sessions.
• .NET in JetBrains Rider (dotnet-ide): Open your .sln or .csproj, build with dotnet build, and use Rider's application or test Debug configuration. Use dotnet test or dotnet run for terminal runs. NuGet packages under /home/devcapsule/.nuget persist per checkout.

The IDE terminal and coding agents use the same installed tools. The base also includes PostgreSQL client commands such as psql; select postgresql-client when declaring that project requirement.

Browser and end-to-end tests

When your test suite drives a browser, run one of these commands on the host:

# Shared in the project manifest:
devcapsule project config offer browser-automation --because "Playwright E2E tests"
# Or added only on your checkout:
devcapsule project config add browser-automation

An offered capability is included unless you omitted it. Use config restore browser-automation to restore an omitted offer; config add is for a capability the project does not declare.

On the next launch, DevCapsule installs Python Playwright and pinned Chromium. Inside the capsule, $DEVCAPSULE_PLAYWRIGHT_PYTHON names its Python interpreter. $PLAYWRIGHT_BROWSERS_PATH names the browser directory, and $DEVCAPSULE_PLAYWRIGHT_WHEELS names the bundled Python wheels. Your project's separate virtual environment still needs its test dependencies. For a browser window on the capsule desktop, launch Playwright with headless=False. It can reach capsule servers on 127.0.0.1 without a published port (Previews and ports).

If your integration tests start companion containers through docker compose, declare docker-cli and authorize docker-daemon host-socket on your checkout; see Services and ports for host-bridge networking caveats.

Native debuggers (gdb, lldb) and resource limits

• Native C/C++/Rust binaries: gcc, g++, cmake, gdb, lldb, and strace are preinstalled in the base image. Starting a program as a child of the debugger (gdb ./build/my-app) avoids some attach restrictions. Attaching to another process can require SYS_PTRACE; the launcher accepts devcapsule project run -- --cap-add=SYS_PTRACE from your host terminal. Host kernel policy and Docker security settings still determine whether debugging is permitted.
• Memory or CPU limits: Exit status 137 can indicate a SIGKILL; it does not prove an out-of-memory kill. Check logs and resource use before changing limits. Increase memory for memory pressure and CPUs for CPU-bound work. For one launch, use devcapsule project run --memory 8G --cpus 4. Save limits with devcapsule project config set memory 8G and devcapsule project config set cpus 4 on the host; see Resource limits.