# KG-NINJA Agent Error Fix Receipt

KG-NINJA provides a small paid x402 endpoint for AI agents and developers.

Paid endpoints:
POST /fix-error
POST /summarize-url
POST /shell-risk-check
POST /agent-visibility-report

Free preview:
GET /fix-error/preview

Pricing:
GET /pricing.json

x402 config:
GET /x402/config

Human explanation:
GET /buy

Purpose:
Submit a failing command, error message, or log snippet. Receive a concise JSON repair receipt containing root cause, one next command, retry plan, risk note, and prevention note.
You can also summarize a public URL or score the risk of a shell command before execution.

Fast start:
- GET /catalog.json for existing offers, limits, input examples and free alternatives
- POST /match with either need or intent, max_price_usdc (decimal string) and network (eip155 identifier) for free goal matching; no purchase or authorization occurs
- GET /offers.json or /offers.xml to opt in to offer revision updates; filter by intent, max_price_usdc and network, poll hourly, and deduplicate by item.id
- GET /find for human goal matching and subscription links
- GET /goal for the primary buyer, pain, result, CTA, and paid endpoint
- GET /agent.json for machine-readable metadata
- GET /payment-options.json for x402 payment details
- GET /openapi.json for request/response contracts
- GET /proof.json for machine-readable x402 mainnet paid flow proof
- GET /proof.md for human-readable proof and cautious interpretation
- GET /share-card.json for concise social/share metadata
- GET /capabilities for the paid routes and payment protocol
- GET /buy for a human-readable landing page

Autonomous purchase decision:
- Use the autonomous_purchase contract in /agent.json, including its buy/skip criteria.
- Preview is a fixed sample; paid output is deterministic triage, not an executed or guaranteed fix.
- Buy only within an existing authorized wallet budget; validate price, network, asset and recipient.
- Do not authorize a second payment while the first payment or delivery is unresolved.
- Verify the result schema and settlement evidence; 402 alone is not revenue or delivery.

Agent payment flow:
1. Read /agent.json
2. Read /goal
3. Read /payment-options.json
4. Read /.well-known/x402/discovery/resources
5. POST /fix-error without payment to receive 402
5b. POST /hyperxosist-query without payment to receive 402
6. Retry POST /fix-error with x402 payment
7. Expect 200 paid response

First-time agent path:
1. Read /.well-known/x402/discovery/resources
2. Read /agent.json
3. Read /payment-options.json
4. Read /goal
5. POST /fix-error, /hyperxosist-query, or /agent-visibility-report without payment to receive 402
6. Retry with x402 payment proof
7. Expect 200 paid response

When to use /fix-error:
- A command, build, deployment, API call, or service check failed and the next action is unclear.
- A retry produced a new error or failed at a different stage.
- The runtime, dependency version, environment, or deployment target changed.
- A machine-readable diagnosis is needed before automation continues.

Complete request examples:
{"command":"npm run build","error":"Error: Cannot find module 'hono'","environment":"Node.js Cloudflare Worker build"}
{"command":"npm run deploy","error":"401 Unauthorized: API token invalid","environment":"CI deployment"}
{"command":"curl http://127.0.0.1:3000/health","error":"Connection refused","environment":"Linux service"}

What you receive after payment:
- HTTP 200 with one JSON repair receipt.
- Fields: root_cause, next_command, retry_plan, risk_note, prevention_note, generated_at.
- Apply the receipt, then submit only a new or materially changed error as another request.

Expected output schema:
{"type":"object","required":["root_cause","next_command","retry_plan","risk_note","prevention_note","generated_at"],"properties":{"root_cause":{"type":"string"},"next_command":{"type":"string"},"retry_plan":{"type":"array","items":{"type":"string"}},"risk_note":{"type":"string"},"prevention_note":{"type":"string"},"generated_at":{"type":"string","format":"date-time"}}}

Complete response example:
{"root_cause":"The build cannot resolve the hono package from the current project dependencies.","next_command":"npm install hono","retry_plan":["Install the missing dependency.","Run npm run build again.","If the error remains, verify package.json and the active working directory.","Submit the new error output as a separate /fix-error request."],"risk_note":"Review package changes before deployment. Do not include tokens or secrets in error logs.","prevention_note":"Commit the lockfile and run dependency installation before the build step.","generated_at":"2026-06-14T00:00:00.000Z"}

Repeat-use scenarios:
- Submit the new error after applying the previous receipt's next command.
- Submit the failure again after changing dependencies, runtime, or environment.
- Submit each distinct build or deployment failure stage separately.
- Use a new request when fresh logs add evidence.
- Do not repeat an unchanged request unless the execution context changed.

Agent Visibility Report:
- Endpoint: POST /agent-visibility-report
- Purpose: returns a privacy-safe report about machine discoverability and payment readiness.
- Unpaid request: returns 402 Payment Required.
- Paid request: returns structured JSON with scores, gaps, matched discovery patterns, and next fix.
- Paid value: observed machine-discovery patterns, payment-readiness scoring, metadata gaps, and one actionable next fix.
- Privacy: aggregate-only; no raw IP, raw User-Agent, wallet address, receipt, tx hash, or event-level log is exposed.

Payment:
x402 required for POST /fix-error.
x402 required for POST /summarize-url.
x402 required for POST /shell-risk-check.
x402 required for POST /agent-visibility-report.
Current mode: mainnet.
Current network: eip155:8453.
Current price: $0.01.
Real revenue: true only after settled mainnet payment.

Trust signal:
mainnet_paid_flow_verified: true
Proof JSON: /proof.json
Proof page: /proof.md
Share card: /share-card.json
Safe claim: This service can complete a real x402 mainnet paid API flow.
Not claiming: a fully autonomous AI independently chose to buy, attribution to any specific company, or recurring external revenue yet.

Safety:
This service does not guarantee a fix. It provides diagnostic guidance only. It avoids destructive commands by default.