Use MCP

Endpoint

https://www.quantum-expectations.com/api/mcp

Streamable HTTP · Stateless sessions · CORS open · Rate-limited per IP

One-click install

Each opens the client's add-connector dialog pre-filled, you confirm before anything is added.

Claude Desktop / Cursor / Windsurf

Add to your MCP config file:

{
  "mcpServers": {
    "quantum-expectations": {
      "type": "http",
      "url": "https://www.quantum-expectations.com/api/mcp"
    }
  }
}

Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json on macOS. Restart the client after saving.

Claude Code CLI

One command:

claude mcp add --transport http quantum-expectations https://www.quantum-expectations.com/api/mcp

ChatGPT

OpenAI offers no pre-filled link, so add it by hand:

1. Turn on Settings → Security and login → Developer mode. Availability depends on your account and workspace policy.
2. Open ChatGPT Plugins and press the + button.
3. Enter a user-facing name and description, for example “Quantum Expectations”.
4. Under Connection, choose Public endpoint and paste the full endpoint URL shown above, including its /api/mcp path. The bare domain will not work.

OpenAI renamed “Connectors” to “Plugins”, so older guides use the earlier name.

Grok

Custom connectors are available on paid Grok tiers:

1. Open grok.com/connectors and choose New Connector, then Custom.
2. Paste the full endpoint URL shown above, including its /api/mcp path.

Stdio-only clients

Bridge through mcp-remote:

{
  "command": "npx",
  "args": ["-y", "mcp-remote", "https://www.quantum-expectations.com/api/mcp"]
}

In-browser agents (no setup)

Browsing this site with an agent-capable browser (ChatGPT's browser, Gemini in Chrome)? The calculator pages register the core compute tools directly with the page via WebMCP, so the agent can calculate without any connector setup. Machine-readable manifest: /.well-known/webmcp.json

What you get

• compute_expectation, effective error rate, success probability, and surface-code / qLDPC overhead for a given circuit (error rate, qubit count, depth).
• compute_required_error_rate, inverse: given a circuit and an acceptable error budget, what physical error rate does the hardware need to achieve?
• compute_fault_tolerant_resources, full fault-tolerant footprint for an algorithm (logical qubits + T count): distillation factory, code distance, physical qubits, and wall-clock time from the Litinski cost model, under two published logical-error fits.
• compare_hardware_scenarios, run the same circuit against all current hardware platforms in one call and rank them.
• fit_historic_series, fit a log-linear trend to the error-rate or qubit-count history and extrapolate when hardware might reach a target value.
• compute_quantum_volume_rate, first-order QV/second estimate for a hardware configuration.
• list_current_quantum_computers, list_hardware_timings, list_example_algorithms, list_qldpc_codes, reference tables (same data as on this site, plus gate-cycle timings not shown on the homepage).
• get_historic_series, full time series with source URLs for trend extrapolation. Pass seriesType as "fidelity" or "qubit-count".
• get_agent_brief, scope, assumptions, and honesty clause.
• Prompts (slash commands in supporting clients): assess_feasibility, assess_fault_tolerant_feasibility, compare_platforms, guided analyses your assistant performs itself with the tools above, assumptions stated.

More tools coming soon.

Try it

Once connected, ask your assistant to do analyses, e.g.: