Prepare an environment for JITX and an AI agent#

Prepare the Python project, native runtime, network access, and agent tools before starting a JITX evaluation. An engineer should be able to build a design, open its editor, and run a read-only request against a live design session. IT staff provide package access, application permissions, and the organization’s AI data-handling policy.

1. Select a compatible version set#

JITX has separate Python packages, a native runtime, editor integration, and agent skills. Installing or updating one does not prove that the others match.

Item	Version	Purpose
`jitx`	`4.6.x`	Python API and `jitx` CLI
`jitxcore`	`4.6.x`	Runtime protocol support
`jitxlib-standard`	`4.6.x`	Standard design library
Native runtime	`4.6.x`	Design engine, layout, checks, and native editor
`jitx-skills`	`1.8.1`	Agent guidance; versioned independently
`jitxlib-parts`, when the project uses it	`2.x`	Parts-provider library; optional for the session example

Use the CLI to select a native runtime compatible with the Python packages. The interactive-session note contains setup commands and the connection procedure.

Optional HFSS workflows have additional packages, skills, licensing, and compute requirements. They are not prerequisites for connecting an agent to a live JITX session.

2. Prepare the host and software#

Engineer and IT: choose a Windows, macOS, or Ubuntu host. Confirm the supported OS version and architecture with JITX; treat other Linux distributions as a separate deployment check. Install Python 3.12 or newer and a package manager such as uv, and use one project virtual environment with a retained lockfile. Run this note’s commands from the project root.

Choose memory capacity from a representative design and concurrent workload. Pin count, layer count, routing, and copper geometry all affect resource use. The heap limit does not account for every process on the host. Use the memory application note when measuring and adjusting it; there is no universal “half-size” relationship between the two heap environment variables.

The CLI can build designs and open standalone board and schematic editors. VS Code with the JITX extension is an additional integrated workflow. Install a JITX extension version compatible with the selected runtime. If using a VS Code derivative, check its extension engine version and marketplace availability with that build.

For a fresh project, follow the JITX project setup guide, then add the dependencies below. For an existing project, review its dependency constraints and lockfile before updating them.

Install the packages from PyPI or your organization’s approved mirror:

uv add "jitx~=4.6.0" "jitxlib-standard~=4.6.0"
uv sync

These constraints allow patch updates. Retain the lockfile to reproduce the exact dependency set.

Use uv run jitx ... below, or activate that project environment in every terminal. In a packaged project, ensure uv sync installs the project itself so its design module is importable.

Install the compatible production runtime and inspect the resolved installation:

uv run jitx --version
uv run jitx runtime resolve-version --prod
uv run jitx runtime install --prod
uv run jitx runtime introspect

Confirm that runtime introspect reports the intended installation. Runtime installation and license authorization are separate steps. Complete authentication using the account and licensing instructions supplied by JITX.

3. Provide the required network access#

IT: permit outbound HTTPS to the destinations used by the selected installation route, or supply an approved mirror. Obtain the current redirect and storage-host list from JITX for your firewall rules. Validate actual artifact downloads as well as the initial API connection.

Destination	Why it is used
`pypi.org` and `files.pythonhosted.org`, or a package mirror	Public Python packages and dependencies
`app.jitx.com` and JITX-provided artifact download destinations	Production authentication, licensing, and runtime delivery
Editor marketplace, when installing the extension	Editor extension delivery, or use an approved VSIX
`github.com/JITx-Inc/jitx-skills`, or an internal copy	Agent skills
The selected AI service	Model access, if inference runs remotely

Configure proxy and certificate trust in Python, the package manager, editor, and agent environment as needed. Use the organization’s trusted CA chain rather than disabling certificate validation.

An offline Python wheel bundle must match the target OS, architecture, Python version, and pinned dependency set. It does not include the native runtime, editor extension, or license. Arrange those separately with JITX. A successful package install alone is not a complete offline deployment check.

4. Configure application permissions and data handling#

IT: allow the native runtime and editor to execute, and allow the local processes to communicate. Confirm that the agent can run the project’s interpreter and CLI and access only the intended project files. The session process executes submitted Python scripts with its own account permissions.

JITX’s design engine runs on the runtime host. The agent may send design source, prompts, script output, and diagnostics to its configured model service. Select the AI deployment and data controls according to the design’s requirements. Separately review any support-bundle upload, remote service, or optional simulation integration you enable; local runtime execution is not a blanket guarantee that every surrounding tool keeps all data on the host.

Confirm the required JITX license and its refresh path with your JITX contact. Network needs also depend on updates, package retrieval, the agent service, and optional integrations; do not assume licensing is the only possible ongoing connection.

5. Install the JITX skills and connect the agent#

Engineer: use an approved agent that can read project files, write Python scripts, execute terminal commands, and retain or connect to a long-running session. A chat interface without execution access can suggest code but cannot run this workflow by itself. Choose the model through your organization’s normal policy; no particular model name is required by these CLI commands.

Install the JITX skills bundle as a plugin. In Claude Code:

/plugin marketplace add JITx-Inc/jitx-skills
/plugin install jitx-skills@jitx

In Codex, which needs Codex CLI 0.142.0 or newer:

codex plugin marketplace add JITx-Inc/jitx-skills
codex plugin add jitx-skills@jitx

For other agent hosts, follow the repository’s installation instructions. Record the installed bundle version or commit. Follow the host’s reload instructions and ask the agent to identify the available JITX skills before relying on them.

Give the agent the project root, the exact design name, this note’s version baseline, and the interactive-session application note. Start a session in a persistent terminal, then have the agent run the read-only readiness check through the same project environment. The note supplies a bounded placement task, executable scripts, result handling, and stop/restart instructions.

For remote compute, run the agent’s commands on the runtime host. The remote-access note covers forwarding the editor connection to another machine.