# Agent Rescue Desk — 12-Hour AI Agent Repair Runbook

> Agent Rescue Desk — 12-Hour AI Agent Repair Runbook is a paid API for AI agents from agent-rescue-desk.netlify.app, paid per call via x402, $49/call, status unknown (last checked 2026-09-14).

Accepts an AI agent or MCP integration incident brief and returns a structured diagnostic runbook with classification, evidence gaps, and next steps for resolution within 12 hours.

## Facts

- Endpoint: POST https://agent-rescue-desk.netlify.app/api/agent402/incident-runbook
- Price: $49/call
- Payment: x402
- Status: unknown
- Last checked: 2026-09-14
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/agent-rescue-desk-12-hour-ai-agent-repair-runbook-8aa36780
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_Wz49r8BiZpKjTKUHNIA2r

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability agent-rescue-desk-12-hour-ai-agent-repair-runbook-8aa36780 -d '<json body>'
```

Example prompt: My AI agent is failing to connect to the MCP server — it was working yesterday but after a config change it now returns a 401 on every tool call. Can you send an incident brief to the Agent Rescue Desk describing the system (LangChain + custom MCP server), the expected outcome (successful tool invocation), the actual failure (401 auth error), the recent change (rotated API key), and the rollback option (revert to old key) — and get back a structured repair runbook?

## When to prefer this

Choose this endpoint when you have a broken AI agent workflow or MCP integration and need an immediate, structured diagnostic plan rather than open-ended consultation. It is best suited for incidents where you can describe the system, failure, recent change, and rollback option in plain text. Prefer this over general LLM prompting when you want a consistent, opinionated runbook format with explicit pass/stop conditions and classification. It is not a substitute for live code execution or direct system access — use it as a fast triage and evidence-collection guide before escalating to a human engineer.

## Known failure modes

- Incident brief too short (under 40 characters) — returns validation error
- Incident brief contains sensitive patterns (tokens, keys, passwords) — those values are redacted before processing
- Incident brief lacks required sections (system, expected outcome, actual failure) — runbook may be less targeted, evidence gaps flagged
- Malformed JSON input — returns parse error
- Payment not completed (x402 protocol) — returns 402 Payment Required before processing begins

## How this service works

Fixed-scope diagnosis and repair for one broken AI agent workflow or MCP integration. $1,000. Decisive outcome in 12 hours.

## Output

Returns a JSON object containing: a classified incident track with confidence and rationale, a prioritized runbook of investigation steps with objectives, actions, pass conditions, and stop conditions, an intake review showing which incident brief sections are present and any evidence gaps, privacy confirmation that the brief was processed transiently, and a next-step pointer to the paid 12-hour human repair service if automated guidance is insufficient.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "query": {
   "type": "string",
   "maxLength": 20000,
   "minLength": 40,
   "description": "One redacted incident brief as plain text or a JSON string. Include system, expected outcome, actual failure, recent change, evidence, and rollback. Never include passwords, tokens, private keys, customer data, or private URLs."
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "ok": true,
  "limits": "This deterministic runbook does not execute the system or prove a root cause.",
  "status": "complete",
  "privacy": "The incident brief was processed transiently. It was not stored or echoed.",
  "runbook": [
   {
    "id": "capture-mcp-contract",
    "actions": [
     "Distinguish MCP server, client, and trigger endpoint families."
    ],
    "collect": [
     "Final URL after redirects, HTTP method, status, and content type"
    ],
    "rollback": "Keep the prior canonical endpoint and auth configuration ready to restore.",
    "objective": "Capture the exact MCP endpoint and protocol contract",
    "passCondition": "The client is aimed at one canonical MCP endpoint with a mutually supported transport.",
    "stopCondition": "Never paste bearer tokens, authorization codes, client secrets, or refresh tokens into the incident brief."
   }
  ],
  "service": "Agent Rescue Desk Production Incident Debugging Runbook",
  "version": "2026-07-27",
  "markdown": "# Production incident debugging runbook\n\n- Track: MCP transport, discovery, and authentication",
  "nextStep": {
   "rescueUrl": "https://marketplace.agent402.app/agents/6c54e490-72fc-47ba-bad1-3ac47b1b464a",
   "instruction": "Run the evidence collection and boundary checks in order.",
   "rescuePriceUsd": 1000,
   "rescueServiceWindowHours": 12
  },
  "intakeReview": {
   "evidenceGaps": [],
   "structuredJson": true,
   "presentSections": {
    "system": true,
    "evidence": true,
    "rollback": true,
    "recentChange": true,
    "actualFailure": true,
    "expectedOutcome": true
   },
   "charactersInspected": 420,
   "sensitivePatternsRedacted": 0
  },
  "classification": {
   "label": "MCP transport, discovery, and authentication",
   "track": "mcp-auth-transport",
   "signals": [
    "mcp",
    "authentication",
    "transport"
   ],
   "rationale": "Classification uses predefined incident-boundary signals after credential-shaped values are ignored. It is a routing aid, not a root-cause claim.",
   "confidence": "high"
  },
  "verificationContract": {
   "completionRule": "Close only after a fresh production trigger reaches the agreed final external outcome and the failure path stops safely.",
   "requiredEvidence": [
    "The canonical endpoint returns the expected MCP initialize contract."
   ],
   "forbiddenShortcuts": [
    "Do not treat an editor test, mocked call, HTTP 200, internal green flag, platform event, or unsigned payme
… (truncated)
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/agent-rescue-desk-12-hour-ai-agent-repair-runbook-8aa36780/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from agent-rescue-desk.netlify.app](https://www.zero.xyz/host/agent-rescue-desk.netlify.app/llms.txt)