Real quantum + qBraid Lab infrastructure capabilities for Kannaka via qBraid (free simulator) and OpenQuantum (real QPUs) — a CLI bridge for the coding agent and an MCP server for any agent.
Four tools with complete schemas, strong descriptions (194 - 392 chars), and clear parameter types. Naming is verb-forward (quantum_devices, run_circuit, quantum_random, resonance_recall). Descriptions explain WHAT, WHEN, and prerequisites (e.g., 'free simulator vs. paid QPU'). Parameters are well-typed with defaults. However, no output schemas are documented, LLMs cannot predict return structure. Error handling is absent; no guidance on retryability or recovery. No tool annotations (readOnlyHint/destructiveHint). Missing parameter constraints (e.g., n_bits range, max_credits bounds). Descriptions are strong but could be more concise (some exceed 300 chars).
List quantum devices across providers (qBraid + OpenQuantum) with status, qubit counts, and cost. Use this to discover what's available before running a circuit. The free ``qbraid:qbraid:sim:qir-sv`` simulator (30 qubits) needs no credits. ``openquantum:*`` entries are real QPUs (IonQ/Rigetti/IQM/AQT) that spend Spark credits — there is no free OpenQuantum simulator.
Generate true quantum random bits from measurement collapse (not a PRNG). Defaults to the free qBraid simulator. Pass ``device="openquantum:<backend>"`` + ``allow_spend=True`` to draw entropy from a real QPU (spends Spark credits). Returns the bitstring, its integer value, and a float in [0,1).
Run Kannaka's resonance recall *as a quantum circuit*. Amplitude-encodes the candidate memory resonances into a quantum state and (with `amplify`) runs amplitude amplification toward the strongest — the quantum analogue of "attention as gravity." Returns the measured distribution over candidates plus the quantum vs classical top pick. Defaults to the free qBraid simulator. For the real-hardware artifact pass ``device="openquantum:iqm:garnet"`` + ``allow_spend=True`` — note recall defaults to 1024 shots, so keep shots low on paid QPUs (the ``max_credits`` cap, default 1.0, guards against an accidental budget-draining run).
Execute an OpenQASM 3 program on a backend and return measurement counts. `qasm3` is OpenQASM 3.0 source (include "stdgates.inc"; declare qubit[]/bit[], apply gates, measure). Defaults to the free qBraid simulator. To run on a real OpenQuantum QPU pass e.g. ``device="openquantum:iqm:garnet"`` AND ``allow_spend=True`` (it spends real Spark credits; a per-shot cost cap of ``max_credits`` credits applies, default 1.0 ≈ $2).
No output schemas documented. LLMs cannot predict return structure (e.g., what fields are in the dict returned by run_circuit). This forces agents to guess and risks parsing errors.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). run_circuit and resonance_recall modify state (execute on QPU, spend credits) but lack destructiveHint. quantum_devices and quantum_random are read-only but lack readOnlyHint.
Missing parameter constraints. n_bits has no min/max (should be 1 - 1024 or similar). max_credits has no bounds. shots has no upper limit. Unbounded numeric params let LLMs pass absurd values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
No error handling guidance. If a device is offline, credentials fail, or QPU quota is exceeded, the tool returns an error but does not tell the LLM what to do next (retry, try a different device, ask user). Agents cannot recover.
Descriptions are verbose (some 300+ chars). Baseline for A+ tools is 50 - 200 chars. Excess length wastes tokens and buries key details. E.g., resonance_recall description could trim the amplitude-amplification explanation.