@repose-sh/sdk

TypeScript client for the Repose REST API. One file, no dependencies, global fetch.

Install

npm install @repose-sh/sdk

Node 20 or newer. ESM only. Published from packages/sdk in the Repose repo.

Getting a token

Settings → API tokens in the app. Tokens are workspace-scoped and look like rsk_…. A read token can list monitors and incidents. Creating, updating, deleting and acknowledging need a read-write token.

Usage

import { Repose, ReposeError } from "@repose-sh/sdk";

const repose = new Repose({ token: process.env.REPOSE_TOKEN! });

// what are we watching
const monitors = await repose.listMonitors();
for (const m of monitors) {
  console.log(`${m.name} — every ${m.intervalSeconds}s, ${m.consensusRequired} probes must agree`);
}

// add a check
const created = await repose.createMonitor({
  name: "checkout",
  kind: "http",
  url: "https://checkout.acme.com/health",
  intervalSeconds: 30,
  consensusRequired: 3,
});

// slow it down
await repose.updateMonitor(created.id, { intervalSeconds: 60 });

// who is on fire right now
const open = await repose.listIncidents({ status: "open" });
if (open[0]) await repose.acknowledgeIncident(open[0].id);

Errors that came from the API are ReposeError with the status and the API's own message — no wrapping, no re-phrasing:

try {
  await repose.deleteMonitor("mon_nope");
} catch (error) {
  if (error instanceof ReposeError) {
    console.error(error.status, error.message); // 404 monitor not found
  }
}

status is 0 when the request never reached the API at all (DNS, TLS, connection refused, wrong baseUrl).

Methods

| method | endpoint | token | | --- | --- | --- | | listMonitors() | GET /api/v1/monitors | read | | getMonitor(id) | — see below | read | | createMonitor(input) | POST /api/v1/monitors | write | | updateMonitor(id, patch) | PATCH /api/v1/monitors/:id | write | | deleteMonitor(id) | DELETE /api/v1/monitors/:id | write | | listIncidents(options?) | GET /api/v1/incidents | read | | acknowledgeIncident(id) | POST /api/v1/incidents/:id/acknowledge | write |

Two things the API can't do yet, and what we do instead

• There is no GET /api/v1/monitors/:id. getMonitor(id) lists the workspace and filters client-side. It returns the right answer; it just costs a full list. If you need several, call listMonitors() once and index it yourself.
• The incidents endpoint has no filters. It returns the 100 most recent incidents, newest first. listIncidents({ status: "open" }) filters those 100 after they arrive, so a workspace with more than 100 recent incidents can have older open ones this doesn't return.

Also worth knowing before you write a sync loop:

• The monitor list doesn't return timeoutMs or expectedStatus, so you can't read back what you wrote for those two.
• updateMonitor can't change a monitor's kind, url, host or port. Delete and recreate to move a target.
• deleteMonitor is a soft delete. Checks stop; history stays.
• Acknowledging an incident quiets the page. It does not resolve it — the probes decide when something is healthy again.

Options

new Repose({
  token: "rsk_…",
  baseUrl: "http://localhost:3000", // default https://repose.sh
  fetch: myFetch,                   // default globalThis.fetch
});