Qiskit Gym MCP Server
by qiskitcommunityv0.3.0
28
14
An MCP server that exposes reinforcement-learning workflows for quantum circuit synthesis using qiskit-gym. It can create synthesis environments, run and monitor training, save/load models, and generate circuits for permutation, linear function, and Clifford synthesis on various hardware topologies.
From the registry: MCP server for qiskit-gym reinforcement learning quantum circuit synthesis
Install & configure
Prerequisites
- requiredInstall the Python packageInstall the MCP server from PyPI using pip.
$ pip install qiskit-gym-mcp-server
Install in your client
Paste into the AI you're chatting with — it will figure out the rest
Please install the `qiskit-gym-mcp-server` MCP server into my current AI client (that's you).
Required prerequisites (do these first if not already done):
- **Install the Python package** — Install the MCP server from PyPI using pip. Run: `pip install qiskit-gym-mcp-server`
Canonical MCP server config (stdio transport):
- command: `qiskit-gym-mcp-server`
- args: []
- optional environment variables:
- `QISKIT_GYM_MODEL_DIR`: Directory where trained models are stored. (example: `<your-model-directory>`)
- `QISKIT_GYM_TENSORBOARD_DIR`: Directory where TensorBoard logs are written. (example: `<your-tensorboard-directory>`)
- `QISKIT_GYM_MAX_ITERATIONS`: Maximum allowed training iterations; 0 means no limit. (example: `<your-max-iterations>`)
- `QISKIT_GYM_MAX_QUBITS`: Maximum number of qubits allowed for training or synthesis requests. (example: `<your-max-qubits>`)
- `QISKIT_GYM_MAX_SEARCHES`: Maximum number of searches allowed for applicable operations. (example: `<your-max-searches>`)
Note: README shows a stdio MCP server launched directly as the installed executable `qiskit-gym-mcp-server`. No required environment variables or API credentials are documented.
Add this MCP server to my current client's config in the correct format for you. If you need secrets or credentials I haven't provided, ASK me — do not invent values or leave raw placeholders. After adding it, tell me how to verify the server is connected.Note: README shows a stdio MCP server launched directly as the installed executable `qiskit-gym-mcp-server`. No required environment variables or API credentials are documented.
Tools (31)
create_permutation_env_tool
Create PermutationGym for SWAP routing
create_linear_function_env_tool
Create LinearFunctionGym for CNOT synthesis
create_clifford_env_tool
Create CliffordGym with custom gate sets
list_environments_tool
List active environments
get_environment_info_tool
Get environment details
delete_environment_tool
Remove an environment
start_training_tool
Start RL training (PPO or AlphaZero), supports background=true
wait_for_training_tool
Wait for background training to complete
batch_train_environments_tool
Train multiple environments
get_training_status_tool
Get training progress and metrics
stop_training_tool
Stop a training session
list_training_sessions_tool
List all training sessions
synthesize_permutation_tool
Generate optimal SWAP circuit
synthesize_linear_function_tool
Environment variables
QISKIT_GYM_MODEL_DIRDirectory for saved modelsQISKIT_GYM_TENSORBOARD_DIRDirectory for TensorBoard logsQISKIT_GYM_MAX_ITERATIONSOptional limit on training iterationsQISKIT_GYM_MAX_QUBITSMaximum number of qubits allowedQISKIT_GYM_MAX_SEARCHESMaximum number of searches allowedPackages
qiskit-gym-mcp-server
pypiv0.3.0transport: stdio
Details
- Language
- Python
- License
- Apache-2.0
- Released
- Mar 13, 2026
- Last commit
- Apr 22, 2026
- Registry name
- io.github.Qiskit/qiskit-gym-mcp-server