# Agent instructions

Agent JSON Reliability. llm_required=false. deterministic=true.

Free: inspect, validate, repair, MCP, GET /v1/catalog, POST /v1/catalog/select.
Paid hosted reliability endpoint: POST https://agent-json-reliability.onrender.com/v1/json/reliable (0.003 USDC per call via x402 on Base when payment is enabled).

AI agents frequently emit malformed JSON. This API deterministically inspects, repairs and schema-validates JSON without another LLM call.

1. GET https://agent-json-reliability.onrender.com/v1/catalog — list 24 paid products
2. POST https://agent-json-reliability.onrender.com/v1/catalog/select {"task": "<what you need>"} — deterministic router, no LLM
3. GET https://agent-json-reliability.onrender.com/.well-known/agent-services.json
4. Prefer capability json_reliable / POST https://agent-json-reliability.onrender.com/v1/json/reliable for malformed JSON
5. Send {"text": "<json or malformed json>", "schema": {optional JSON Schema}}
6. Do not invent values if repaired=false
7. POST /v1/json/reliable is paid x402 exact 0.003 USDC on Base. POST /v1/evidence/pack is 0.0075 USDC. Additional Agent Utility Store routes are paid when listed in /.well-known/x402. inspect/validate/repair and MCP stay free.
8. Machine discovery: https://agent-json-reliability.onrender.com/openapi.json https://agent-json-reliability.onrender.com/.well-known/x402 https://agent-json-reliability.onrender.com/.well-known/agent.json https://agent-json-reliability.onrender.com/skill.md https://agent-json-reliability.onrender.com/llms.txt https://agent-json-reliability.onrender.com/v1/catalog

Limits: body 262144 bytes, depth 64, 30 requests/minute/session.
PyPI: uvx agent-json-reliability
MCP Registry: io.github.toninovo4249-ai/agent-json-reliability
GitHub: https://github.com/toninovo4249-ai/agent-json-reliability