QECTOR for OpenAI & Codex
First-class quantum error correction for OpenAI Codex and every MCP-compatible client. A local Model Context Protocol server over stdio JSON-RPC 2.0 (protocol 2024-11-05) connects your agent directly to the Rust-core decoder: circuits, parity matrices, and syndromes never leave your machine.
Capability matrix (per product, per version)
| Surface | Transport | MCP tools | Scope |
|---|---|---|---|
| OpenAI / Codex adapter | local stdio | 8 | 8 verified library tools over the qector_decoder_v3 1.0.0 wheel |
| Claude Plugin v1.0.2 | local stdio | 37 | 8 library + 29 benchmark tools, 28 skills, 5 agents |
| Workbench v1.0.2 | local stdio (--mcp) | 85 | desktop GUI + full MCP surface, Windows / Linux / macOS |
First-boot verification
pip install qector-decoder-v3==1.0.0 mcp qector-doctor codex mcp add qector -- python path/to/mcp_server_library.py
Every decode is verified against the H·c = s contract by the self-debug layer. The server script ships with the qector-claude-plugin repository.
Evidence model
Claims surfaced through this adapter are grounded in the six-record Zenodo evidence corpus (qector.store/evidence) and the normative reference manual (DOI 10.5281/zenodo.21941046). The adapter runs entirely on your machine over local stdio — no syndrome data is transmitted to QECTOR systems or any third party.