Developer API

Wrench is a web & mobile assistant first — but the same troubleshooting engine is available over an API so you can wire it into your own tools. One endpoint: send context, read streamed guidance.

POST /api/v1/troubleshoot/stream

Send a reported problem plus whatever context you have and receive the same streamed diagnostic guidance the Wrench apps produce — safety warnings, ranked likely causes, an ordered diagnostic sequence and a parts/tools list. Guidance streams back token by token as Server-Sent Events so you can render it live.

Authentication

Every request needs a Bearer API key in the Authorization header:

Authorization: Bearer wr_live_xxxxxxxxxxxxxxxxxxxxxxxx

Keys are created and revoked under Settings → Developer. Keys are hashed at rest — copy a key when it's shown; it can't be retrieved later. A missing or invalid key returns 401.

Request body

Send a JSON body. Provide context, messages, or both — at least one is required (an empty body returns 400).

Field	Type	Notes
context	object	Freeform. All fields optional — send what you have.
context.title	string	Short summary of the problem.
context.description	string	The reported problem in the technician's words.
context.equipmentType	string	Equipment class / type, e.g. "commercial refrigeration".
context.priorNotes	string	Prior history, recent fixes, or anything else relevant.
messages	array	Conversation history for follow-ups. Each item is { "role": "user" | "assistant", "content": string }. The server prepends your context as the first turn.

There is no fixed equipment schema — context is intentionally freeform. Send a single problem for a first answer, or replay the visible conversation in messages to ask a follow-up.

Response — Server-Sent Events

On success the response is text/event-stream. Each frame is a data: line carrying a small JSON object:

Frame	Meaning
data: {"text":"…"}	A chunk of guidance. Concatenate text across frames to build the full answer.
data: {"done":true,"requestId":N}	Terminal frame. The stream is complete; requestId is the logged request's id.
data: {"error":"…"}	The generation failed mid-stream. No further frames follow.

Errors & limits

Errors that occur before streaming starts return a JSON error envelope with the matching HTTP status:

Status	When
401	Missing or invalid API key.
400	Bad body — neither context nor messages provided.
402	Your plan doesn't include Wrench troubleshooting, or your monthly call quota is exhausted.
429	Rate limited — you hit the per-hour cap or a short-window burst guard. Slow down and retry.

Monthly and hourly caps are per account and scale with your plan — see pricing. A failure after the stream has started arrives as an error SSE frame (above) rather than an HTTP status, because headers are already sent.

Example — curl

curl -N https://wrench.example.com/api/v1/troubleshoot/stream \
  -H "Authorization: Bearer $WRENCH_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "context": {
      "title": "Walk-in freezer warm",
      "equipmentType": "commercial refrigeration",
      "description": "Not holding temp; compressor short-cycles.",
      "priorNotes": "Condenser cleaned last month."
    }
  }'

# → text/event-stream
data: {"text":"SAFETY: LOTO before opening the electrical panel.\n"}
data: {"text":"LIKELY CAUSES: 1) Low refrigerant charge..."}
data: {"done":true,"requestId":8412}

Example — Node (fetch)

const res = await fetch(
  "https://wrench.example.com/api/v1/troubleshoot/stream", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.WRENCH_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    context: { description: "Pump losing prime on startup" },
  }),
});

let answer = "";
for await (const chunk of res.body) {
  for (const line of chunk.toString().split("\n")) {
    if (!line.startsWith("data:")) continue;
    const evt = JSON.parse(line.slice(5).trim());
    if (evt.text) answer += evt.text;
    if (evt.error) throw new Error(evt.error);
    if (evt.done) console.log("done", evt.requestId);
  }
}
console.log(answer);

Follow-up example

To ask a follow-up, replay the conversation in messages. Keep the original context so the guidance stays grounded:

{
  "context": { "description": "Walk-in freezer warm; compressor short-cycles." },
  "messages": [
    { "role": "assistant", "content": "LIKELY CAUSES: 1) Low refrigerant charge..." },
    { "role": "user", "content": "Charge is good. What next?" }
  ]
}

Start your free trial See pricing