Quantum Engineer Plugin

Overview

This is an enterprise-grade feature available exclusively to Classiq’s enterprise customers.

The Quantum Engineer plugin brings Classiq expertise directly into your agentic coding platform. It teaches coding agents like Claude Code, Cursor, Codex, and Copilot how to model, synthesize, execute, analyze, and post-process quantum programs with Classiq. Build, run, and analyze quantum programs in natural language: describe what you want and the agent handles every step, from writing the model to reading back results. Programs are expressed in Qmod, and draw on the open library of reusable quantum functions. Synthesis delegates to Classiq’s backend engine, which compiles high-level models into hardware-optimized circuits automatically. The plugin includes skills for each phase of quantum development:

Modeling: write quantum algorithms in Python using Classiq.
Synthesis: compile a model into a quantum program with synthesize(), constraints, and hardware targets.
Execution: run circuits with sample, observe, calculate_state_vector, and variational loops.
Post-processing: interpret results such as histograms, energy curves, and most-probable states.
Analyzer: inspect synthesized programs for gate counts, depth, and hardware fit.
Qiskit migration: translate existing Qiskit code into Classiq models.

A built-in reference layer keeps the agent’s knowledge current as it works.

Installation

First, download the plugin zip from the Classiq platform. The zip includes all skills discussed above and the MCP server configuration.

Classiq Platform
Quantum Engineering Interface

Open the Downloads menu in the top bar on the right, then click AI Plugin. The top bar is available from any page in the Classiq platform.

Downloading the plugin zip from the Downloads menu in the Classiq platform

After downloading and extracting the zip file, follow the steps below to install the plugin on your agentic coding platform of choice, or simply ask your coding agent to install it from the extracted directory.

Claude Code
Cursor
Codex
Copilot

The Quantum Engineer plugin is supported in the Claude Code CLI and in its VS Code and JetBrains IDE integrations. The desktop and web applications do not support loading local plugins.

For a persistent install, copy the extracted plugin directory into your Claude Code skills folder:

cp -r /path/to/classiq-quantum-engineer ~/.claude/skills/classiq-quantum-engineer

Any folder there with a .claude-plugin/plugin.json manifest loads automatically every session, with no flag or marketplace needed. The skills appear namespaced under the plugin, for example classiq-quantum-engineer:classiq-modeling.Alternatively, for a one-off session, point Claude Code at the extracted directory using the --plugin-dir flag (this loads the plugin for that session only, so you must pass it each launch):

claude --plugin-dir /path/to/classiq-quantum-engineer

On first use you will be prompted to approve the classiq-mcp server. After approving, the plugin’s skills and MCP configuration are active.To verify the installation, run /skills and confirm the Classiq skills are listed, then run /mcp and confirm the classiq-mcp server is connected. As a quick smoke test, you can also ask the agent to look up the Classiq docs for synthesize and watch that it calls the classiq-mcp server.To pick up a newer version of the plugin, download the latest zip from the platform, replace the directory (under ~/.claude/skills/ or wherever you loaded it from), and run /reload-plugins, or start a new session.

The Quantum Engineer plugin is supported on the Cursor desktop application and CLI.

First, create the local plugins folder if it doesn’t already exist, and copy the extracted plugin directory into it:

mkdir -p ~/.cursor/plugins/local
cp -r /path/to/classiq-quantum-engineer ~/.cursor/plugins/local/classiq-quantum-engineer

To load the plugin in the Cursor desktop application, reload Cursor by opening the Command Palette (Cmd+Shift+P) and running Developer: Reload Window, or simply restart the application. After the plugin is loaded, open Customize and confirm the plugin, its skills, and the classiq-mcp server all appear.If you’re using the Cursor CLI, point enabled_plugins in ~/.cursor/settings.json at the local plugin folder so it loads persistently in every session.Alternatively, for a one-off session, pass the --plugin-dir flag:

agent --plugin-dir ~/.cursor/plugins/local/classiq-quantum-engineer

Finally, run /mcp (or agent mcp list) to confirm the classiq-mcp server is connected, and type / to check the plugin’s skills appear in the menu. As a quick smoke test, you can also ask the agent to look up the Classiq docs for synthesize and watch that it calls the classiq-mcp server.To pick up a newer version of the plugin, download the latest zip from the platform, replace the extracted directory, and reload (desktop) or restart the CLI session.

The Quantum Engineer plugin is supported in local Codex environments, including the Codex CLI, desktop application, and IDE extensions. The web application does not support local plugins at the moment.

Installing the plugin on Codex requires creating a personal marketplace (if you don’t already have one) and registering the plugin within it.First, place the extracted Quantum Engineer plugin directory under ~/plugins/:

mkdir -p ~/plugins
cp -r /path/to/classiq-quantum-engineer ~/plugins/classiq-quantum-engineer

Then, create a marketplace file at ~/.agents/plugins/marketplace.json with the following content:

{
  "name": "personal",
  "interface": {
    "displayName": "Personal"
  },
  "plugins": [
    {
      "name": "classiq-quantum-engineer",
      "source": {
        "source": "local",
        "path": "./plugins/classiq-quantum-engineer"
      },
      "policy": {
        "installation": "AVAILABLE",
        "authentication": "ON_INSTALL"
      },
      "category": "Productivity"
    }
  ]
}

Or, if you already have a personal marketplace, simply add the following entry to its plugins array:

{
  "name": "classiq-quantum-engineer",
  "source": {
    "source": "local",
    "path": "./plugins/classiq-quantum-engineer"
  },
  "policy": {
    "installation": "AVAILABLE",
    "authentication": "ON_INSTALL"
  },
  "category": "Productivity"
}

Finally, install the plugin from the personal marketplace:

codex plugin add classiq-quantum-engineer@personal

Start a new Codex task or session so the skills and MCP configuration are loaded successfully, as already-running sessions will not pick up newly installed plugins.To verify the installation, run /skills and confirm the Classiq skills are listed, then run /mcp and confirm the classiq-mcp server is connected. As a quick smoke test, you can also ask the agent to look up the Classiq docs for synthesize and watch that it calls the classiq-mcp server.To update to a newer version, download the latest zip from the platform, replace the directory under ~/plugins/, and re-run codex plugin add classiq-quantum-engineer@personal. Then start a new session. If the plugin version in .codex-plugin/plugin.json has changed between releases, Codex installs the updated package instead of reusing a cached copy.

The Quantum Engineer plugin is supported on GitHub Copilot in VS Code. The desktop application does not support local plugins.

In VS Code, open the Copilot agent window. In the left sidebar, under Customizations at the bottom, click Plugins. Then, to the right of the Available section, you will find the Install from Source button. Click it and provide the path to the extracted plugin directory.The plugin then appears under Installed as classiq-quantum-engineer with a Local badge, listing the skills and MCP server it provides. Use the toggle beside it to enable or disable the plugin.

Installing the Quantum Engineer plugin from source in Copilot

To verify the installation, head to Skills and MCP Servers under Customizations, and confirm the Classiq skills and the classiq-mcp server are listed.To pick up a newer version of the plugin, download the latest zip from the platform, replace the extracted directory, and install it from source again.

The plugin requires the classiq Python package. If you don’t have it yet, ask the agent to install it and the bundled classiq-setup skill will handle installation and authentication. See the Install Classiq example in Examples.

Project context setup (optional)

Each platform supports a project-level file that the agent reads at the start of every conversation. Use it to record your Python environment and any project-specific conventions, so the agent never has to ask and never guesses wrong.

Claude Code
Cursor
Codex
Copilot

Create CLAUDE.md at the root of your project and fill in the bracketed fields:

# [Project Name]

## Environment

- Python: [path/to/venv/bin/python]  <!-- or just `python3` if using the system default -->
- Classiq SDK: installed in the environment above

## Backend

Default: Classiq cloud simulator.
[Optional: list hardware targets or execution preferences for this project.]

## Project overview

[One or two sentences describing the quantum algorithms or application this project implements.]

## Conventions

[Any project-specific patterns: naming, file layout, preferred synthesis constraints, etc.]

Claude Code picks up CLAUDE.md automatically, no additional configuration required.

Create .cursor/rules/classiq.mdc at the root of your project. This file gives Cursor persistent Classiq-specific context for the project.

---
description: Classiq project context
alwaysApply: true
---

# [Project Name]

## Environment

- Python: [path/to/venv/bin/python]  <!-- or just `python3` if using the system default -->
- Classiq SDK: installed in the environment above

## Backend

Default: Classiq cloud simulator.
[Optional: list hardware targets or execution preferences for this project.]

## Project overview

[One or two sentences describing the quantum algorithms or application this project implements.]

## Conventions

[Any project-specific patterns: naming, file layout, preferred synthesis constraints, etc.]

The alwaysApply: true frontmatter tells Cursor to load this rule for every conversation in the project.To verify the rule is active, open Settings → Rules and confirm the project rule appears. You can also ask the agent: “What Classiq project context do you see?” The response should reflect the environment and conventions from .cursor/rules/classiq.mdc.

Create AGENTS.md at the root of your project and fill in the bracketed fields:

# [Project Name]

## Environment

- Python: [path/to/venv/bin/python]
- Classiq SDK: installed in the environment above

## Backend

Default: Classiq cloud simulator.
[Optional: list hardware targets or execution preferences for this project.]

## Project overview

[One or two sentences describing the quantum algorithms or application this project implements.]

## Conventions

[Any project-specific patterns: naming, file layout, preferred synthesis constraints, etc.]

Codex reads applicable AGENTS.md files when working in the project. In the Codex CLI and desktop application, you can also run /init to generate a starter AGENTS.md.

Copilot reads AGENTS.md and CLAUDE.md from your project root, so if you already created one of those for Codex or Claude Code, Copilot picks it up as is with no extra file.To use a Copilot-specific file instead, create .github/copilot-instructions.md with the same content:

# [Project Name]

## Environment

- Python: [path/to/venv/bin/python]
- Classiq SDK: installed in the environment above

## Backend

Default: Classiq cloud simulator.
[Optional: list hardware targets or execution preferences for this project.]

## Project overview

[One or two sentences describing the quantum algorithms or application this project implements.]

## Conventions

[Any project-specific patterns: naming, file layout, preferred synthesis constraints, etc.]

Usage

Once the plugin is installed, just describe what you want in natural language, no special syntax required. The agent picks the right skill for each phase of the work:

You ask for	Skill that responds
Writing, editing, or debugging a quantum algorithm in Python (`@qfunc`, `QBit`, `QNum`, `QArray`)	Modeling
Writing or debugging native Qmod (`.qmod` files)	Native Qmod
Compiling a model: `synthesize()`, constraints, hardware targets	Synthesis
Running a circuit: `sample`, `observe`, `calculate_state_vector`, variational loops	Execution
Interpreting results: histograms, energy curves, most-probable states	Post-processing
Inspecting a synthesized program: gate counts, depth, hardware fit	Analyzer
Translating Qiskit code into Classiq models	Qiskit migration

For requests that span multiple phases (for example, “model this and run it”), the agent organizes the full pipeline end-to-end.

Examples

Hello World
Quantum Arithmetic
Grover's Search
Install Classiq

Prompt:

Create a Bell state, synthesize it, run it on the simulator, and show the measurement counts.

Generated code:

from classiq import *

@qfunc
def main(res: Output[QArray[QBit, 2]]):
    allocate(2, res)
    H(res[0])
    CX(res[0], res[1])

qprog = synthesize(main)
result = sample(qprog)
print(result)

Output:

      res  counts  probability bitstring
0  [1, 1]    1036     0.505859        11
1  [0, 0]    1012     0.494141        00

Prompt:

Write a quantum program that puts a 1-qubit integer in superposition, assigns 2 to a second integer, computes their sum, and runs it.

Generated code:

from classiq import *

@qfunc
def main(x: Output[QNum], y: Output[QNum], z: Output[QNum]):
    allocate(1, x)
    H(x)
    y |= 2
    z |= x + y

qprog = synthesize(main)
res = sample(qprog)
print(res)

Output:

x	y	z	counts	probability
0	2	2	1049	0.512
1	2	3	999	0.488

Prompt:

Use Grover's algorithm to find all pairs (x, y) of 2-bit integers where x + y < 4 and (x * y) % 4 == 2. Run it and show which states get amplified.

Generated code:

from classiq import *

VAR_SIZE = 2

class GroverVars(QStruct):
    x: QNum[VAR_SIZE]
    y: QNum[VAR_SIZE]

@qperm
def my_predicate(vars: Const[GroverVars], res: QBit) -> None:
    res ^= (vars.x + vars.y < 4) & ((vars.x * vars.y) % 4 == 2)

@qfunc
def main(vars: Output[GroverVars]):
    allocate(vars)
    hadamard_transform(vars)
    power(
        2,
        lambda: grover_operator(
            lambda vars: phase_oracle(predicate=my_predicate, target=vars),
            hadamard_transform,
            vars,
        ),
    )

qprog = synthesize(main)
result = sample(qprog)
result

Output:

vars.x	vars.y	probability
1	2	0.481
2	1	0.463
2	2	0.006

Prompt:

Install classiq and set it up for me.

The agent invokes the classiq-setup skill, which walks through the full installation in your environment: it runs an environment diagnostic, installs the classiq package via pip, opens a browser window for authentication, and confirms everything is ready with a final status summary.

claude code completing the classiq installation using quantum engineer plugin

Troubleshooting

MCP server does not connect

Claude Code
Cursor
Codex
Copilot

The bundled classiq-mcp server connects automatically when the plugin is enabled.

Run /plugin list and confirm the plugin appears.
Run /mcp to open the MCP panel and check the classiq-mcp status. If it shows as failed or cached, select Reconnect; if it is disabled for the project, toggle it back on.
If it still does not appear, run /reload-plugins. Plugins loaded via --plugin-dir are session-scoped, so if you restart Claude Code you must pass the flag again.

In the desktop application, open Customize in the sidebar, confirm the classiq-mcp server is enabled (use the toggle if it is off), then reload the window (Command Palette → Developer: Reload Window).In the CLI, the server appears under /mcp; you can also run agent mcp list to check its status or agent mcp enable to enable it, and pass --approve-mcps to auto-approve servers on launch.If the server still does not appear, re-copy the extracted plugin directory into ~/.cursor/plugins/local/ (desktop) or re-run with --plugin-dir (CLI), then reload.

Run codex plugin list to confirm the plugin is installed.
Start a new Codex task or session. Already-running sessions do not pick up newly installed plugins.
In the new session, run /mcp to confirm the classiq-mcp server is connected.

Open Agent Customizations and go to MCP Servers, then confirm the classiq-mcp server is listed and enabled.If it is missing, check that the plugin itself is enabled under Plugins, and that MCP access is not turned off in your settings (chat.mcp.access).

Skills do not load

Claude Code
Cursor
Codex
Copilot

Run /skills to confirm the plugin’s skills appear. If they are absent:

Run /reload-plugins to reload the plugin.
Check the /plugin Errors tab for load errors.
Plugins loaded via --plugin-dir are session-scoped, so if you relaunched Claude Code, pass the flag again.

In the desktop application, skills appear in Customize → Skills, and in the CLI they appear in the / menu. If they are absent:

Reload the window (Command Palette → Developer: Reload Window) in the desktop application, or start a new CLI session.
Confirm the plugin appears in Customize (desktop), or that enabled_plugins/--plugin-dir points at the plugin folder (CLI).
Make sure no marketplace plugin of the same name is installed, since it takes precedence over the local copy.

Run /skills to confirm the plugin’s skills appear. If they are absent:

Start a new Codex task or session. Already-running sessions do not pick up newly installed plugins.
Run codex plugin list to confirm the plugin itself is installed.

Open Agent Customizations and go to Skills, then confirm the Classiq skills are listed. The plugin’s row under Plugins also shows how many skills it provides.If they are absent, confirm the plugin is enabled under Plugins, and that skills are not turned off in your settings (chat.useAgentSkills).

Local plugins do not load on enterprise plans

Enterprise and team plans may restrict local plugin loading by default. If the steps above haven’t resolved the issue, check with your organization admin whether any of the following restrictions are in effect.

Claude Code
Cursor
Codex
Copilot

The disableSideloadFlags managed setting, when enabled by an admin, rejects the sideload flags including --plugin-dir.

An admin-managed requirements.toml policy file can restrict local plugin sources (for example, restrict_to_allowed_sources = true).

An enterprise-managed managed-settings.json file can restrict which plugins are allowed, through enabledPlugins (plugins installed or blocked for all users).

AI

Authentication

Modeling

Synthesis

Execution

Analysis

Error Correction

Studio

Applications

Quantum Engineer Plugin

Overview

Installation

Project context setup (optional)

Usage

Examples

Troubleshooting

See also

​Overview

​Installation

​Project context setup (optional)

​Usage

​Examples

​Troubleshooting

​See also

Overview

Installation

Project context setup (optional)

Usage

Examples

Troubleshooting

See also