Conductor Quantum, a startup specializing in natural language interfaces for quantum computing, has introduced CODA MCP, a Model Context Protocol (MCP) server. The platform is designed to allow AI agents—such as Claude Desktop, VS Code, Cursor, and Zed—to utilize quantum computing resources as native tools. By providing a bridge between large language models and quantum backends, CODA MCP enables agents to automate the generation, simulation, and execution of quantum circuits through a standardized interface.
The server provides multi-provider access to physical hardware from IBM, AQT, IQM, IonQ, and Rigetti, encompassing over 1,000 qubits. To address the fragmentation of quantum software development, the platform supports cross-framework transpilation across Qiskit, NVIDIA CUDA-Q, Cirq, PennyLane, Amazon Braket, and PyQuil. This allows researchers and agents to prototype algorithms in one SDK and execute them on a different backend without manual code rewrites, utilizing a QPU leaderboard to assist in hardware selection.
For iterative development, CODA MCP supports high-performance simulations of up to 34 qubits utilizing NVIDIA cuQuantum libraries and the NVIDIA CUDA-Q platform. The server also includes a suite of workflow utilities, such as tools for resource estimation (tracking qubit count, depth, and gate counts), circuit splitting for distributed execution, and exporting to OpenQASM 3. These features are intended to allow AI agents to perform parameter sweeps and benchmark circuits before committing to physical QPU time.
The platform further integrates research-centric capabilities, including tools for searching and retrieving scientific papers. This functionality enables an “agentic loop” where an AI can ground a proposed experiment in existing literature, simulate the logic for verification, and then execute the job on a target QPU. By centralizing these tasks within the MCP framework, Conductor Quantum aims to provide a repeatable path for AI-driven quantum research, moving from natural language descriptions to analyzed results within a single environment.
CODA MCP is currently available via the coda-mcp Python package for users with a Coda account. Installation involves generating an API token and configuring the server within an MCP-compatible client. The release targets the growing ecosystem of AI-assisted coding tools, offering a natural language interface located at coda.conductorquantum.com for web-based interaction. The startup plans to expand its hardware coverage and backend capabilities as the adoption of agent-based quantum development increases.
For technical documentation on the MCP server integration and API details, consult the official announcement here and the PyPI project page here.
March 23, 2026
