# Hybriqx Workout Adaptation

> Paid workout adaptation via REST + x402 and MCP. No A2A endpoint or A2A conformance.

- Agent landing: https://www.hybriqx.com/agents
- Human browser experience (separate from the paid agent API): https://www.hybriqx.com/app
- Buyer guide: https://www.hybriqx.com/buyers.txt
- API schemas: https://www.hybriqx.com/openapi.json and https://www.hybriqx.com/docs
- Synthetic examples: https://www.hybriqx.com/examples.json
- Priced resources (local discovery, not a facilitator index): https://www.hybriqx.com/.well-known/x402
- MCP server metadata: https://www.hybriqx.com/.well-known/mcp/server-card.json
- MCP endpoint: https://www.hybriqx.com/mcp/ (stateless Streamable HTTP)
- Legacy custom descriptor (not A2A): https://www.hybriqx.com/.well-known/agent-card.json

POST https://www.hybriqx.com/api/v1/scale-wod or https://www.hybriqx.com/v1/wod/scale.
MCP service_info is free metadata; scale_workout is paid.
Configured price: $0.05 USDC per request/tool call,
network eip155:8453. An x402-capable wallet client and explicit
spending approval are required; do not automatically pay or retry.
Availability: {"paid_scaling_available": true, "payments_enabled": true, "wod_scaler_enabled": true}.
Disabled or unconfigured scaling is unavailable, not a free fallback.
The browser app's internal APIs and generated workouts are not this paid surface.

Public buyer client (MIT): https://github.com/LKennedyShare/hybriqx-workout-assistant
Install from Git (no PyPI publication claim):
python -m pip install "git+https://github.com/LKennedyShare/hybriqx-workout-assistant.git"
Offline demo (synthetic fixture; no network, model call or payment):
python -m hybriqx_workout_assistant --demo
The client cap is at most 0.05 USDC per paid attempt and per in-memory session; a runtime price above the cap is refused, not automatically approved. LangChain and OpenAI Agents integrations require optional langchain/openai extras. Paid framework tools are disabled unless a trusted operator approval_policy is supplied; model arguments cannot authorize spending. Reuse the tool factory for session limits; production needs a durable external budget/approval ledger. No automatic paid retries. This is a buyer client, not an MCP server or A2A agent.

athlete_notes is optional. Submitted workout text, equipment and notes are sent to the configured Google Gemini model. Do not send personal or health information without the athlete's informed consent and appropriate data-handling authorization.
Model-generated adaptations are not medical advice or medically approved workouts. Equipment checks are bounded and cannot verify all movement, medical, or duration constraints. Human review is required, especially for injuries or restrictions. Do not exercise through pain; seek appropriate professional guidance.
SYNTHETIC_ILLUSTRATION: static request/response contract illustration only. Not executed, not a paid result, not a verified duration, and not medically approved. execution_time_ms=0.0 is a placeholder, not a latency measurement.