/ SONDA · AUTOMATE

Your APIs, checked every minute. From inside your network.

A schedule runs a request, a folder or a whole project on a cadence — every request with its scripts and its checks, exactly as Send does on your desktop. Sonda Server runs it on a box of yours, so it reaches the APIs only your network reaches, and nothing of a run ever leaves it. When it breaks, the people who need to know hear it. When it recovers, they hear that too.

EVERY MINUTE TO WEEKLY MAIL AND WEBHOOK ALERTS A DAILY OR WEEKLY SUMMARY RUNS SEALED UNDER THE PROJECT'S KEY

A schedule

WhatA request, a folder or the whole project, with the shared environment it runs with. Right-click any of them and choose Schedule.
WhenEvery N minutes — one at least — hourly, daily at a time, or weekly on the days you pick, in the server's local time.
HowA timeout for the whole run, five minutes by default, up to an hour. Retry once when it fails. Stop at the first failure. From 2 to 500 runs kept, 30 by default.
WhoWhoever holds Schedule runs on the project: create, change, pause, resume, remove, run now. A schedule runs with what its saver may send, and no more.
WhereSonda Server, paired by the owner after comparing its fingerprint. A team on a store with no server sees the Schedules page say so.
/ HOW A RUN GOES

The server runs it. The store keeps it. The team reads it.

On the minute, Sonda Server opens the copy of the project sealed for its key — written from the store at the last publish, never from anyone's computer — checks every file against the project's signed head, and sends each request through the same engine your desktop uses: the folder's auth, the scripts, the checks, the shared environment as the saver resolved it.

The run becomes a record in the team store, signed by the server and sealed under the project's key: whoever opens the project reads its runs, and nobody else does. Not LockFlare, not the box the store is on.

Sonda Server, and pairing it
/ THE RUNS

Every request, every check. Kept where the team reads it.

What a run record holds, what the board shows, and what a record never keeps.

Request by request, as the runner shows them.

Status, time, size and the checks' verdicts for every request. A request that failed keeps its response for diagnosis — 64 KB a body, 8 MB a run.

Run now, when you cannot wait for the cadence.

One run out of its cadence, after a confirmation, within a minute. A paused schedule runs once and stays paused. The run lands in the list as the server writes it.

One board for the workspace, failing first.

Every schedule the person may see, with how many runs in a row it has failed. A project whose last run failed is marked on the Projects page. A bar per run — its height the duration, its color the verdict — with the success rate, the average and the slowest run from the newest 30.

A server started again picks up where it was.

It knows where each schedule was from the runs in the store. One run at a time per schedule; slots missed while the server was down are not made up — one run, then on from there.

A record never keeps what it should not.

No address as it went out — the URL resolved, a key in its query. Every secret value masked in every text of the record and of the alerts: a check's reason, a console line, a response's headers and body.

/ ALERTS AND THE SUMMARY

Told when it breaks. Told when it's fixed. Not every minute in between.

The server sends on the changes, read from the runs themselves, so a restart never loses count. By mail, through the organization's own mail server handed to Sonda Server sealed for its key — plain text, no body, no header, no value, no link, up to 20 addresses. By webhook, a POST of JSON signed with HMAC-SHA256 when a secret is set, set by administrators, never following a redirect.

The mail goes to your mail server, the webhook to the address you give, the runs to your store. Nothing of it reaches LockFlare.

From: sonda@mail.example.com To: api-team@example.com Subject: [Sonda] Payments API · hourly — failing Project Payments API Schedule hourly · every 60 minutes Run 2026-10-21 08:00 · 11 of 12 passed Failures 3 in a row — the threshold ✗ POST /orders/{id}/capture check status is 200 got 503 check body.status is "held" not reached # no body, no header, no value, no link — the run is in the store — two runs later — Subject: [Sonda] Payments API · hourly — recovered Run 2026-10-21 10:00 · 12 of 12 passed
FailedWhen the failures in a row reach the threshold: one for the first failure, more so a flaky endpoint does not wake anyone over one blip. A refused run counts — the schedule did not run.
RecoveredOnce, when a run is fine again after an alert went out. Still failing every N failures while it keeps failing, or never. Each run says what went out for it.
By mailThrough the organization's own mail server. Plain text: the project, the schedule, the failed requests and the checks that did not pass. Up to 20 addresses.
By webhookA POST of JSON with the event, the run and the failed requests, signed with HMAC-SHA256 when a secret is set. Never follows a redirect.
The summaryA workspace's digest by mail, daily at a time or weekly on days, set by its administrators: failing now first, then the worst success rate, then the paused and the ones that did not run — each with its runs, rate, average, slowest and alerts.
Where it goesYour mail server, the address you give, your store. Nothing of it reaches LockFlare.
/ ADVANCED RUNS

Sign in first. Fetch the list. Run it for each one.

An advanced schedule sends sources before the target — requests of the project — and reads their JSON answers into the variables the target uses.

Sources, up to five

Sent before the target, each through the same engine: its folder's auth and scripts, the environment, what the sources before it filled. Up to 20 variables each, read from the answer by path.

A sign-in, read and dropped

The first source signs in; its token fills a variable for the sources after it and for the target. A source's answer is read and dropped — a sign-in's token is never kept.

The loop

One source can loop the target over an array of its answer: once per element, up to 500, with loopIndex counting from 1. An empty array is a good run with nothing to run over.

A source that fails

No 2xx, no JSON, or a path that finds nothing: the run is refused, the record says which source and why, and the alerts count it.

/ WHO MAY, AND AS WHOM

A schedule gives nobody more than they have.

The server checks before every run, with the permission files as they stand now — not as they stood when the schedule was saved.

Schedule runsOne permission, per project: create, change, pause, remove, run now. A schedule belongs to the project, not to its author: any holder or an administrator changes or stops it. App runners never see the runs.
Signed by its saverEvery save signs what the schedule runs with the saver's key, and the server runs nothing whose signature does not verify. The saver's Send is checked at every request the run sends; a request the saver may not send is named and left out.
Encrypted projectsThe server runs from a copy of the project sealed for its own key, written from the store and never from a computer — unpublished changes never reach it. Every file is checked against the project's signed head before a run.
The valuesThe shared environment and the Team secrets the saver may read, resolved once at the save and sealed for the server. A value the saver cannot fill is named on the page.
Hidden stays hiddenA request hidden from the team never runs on a schedule, an administrator's schedule included. It is counted, never named.
The server's diskA request whose body is a file of someone's computer is skipped on the server and named. The server never puts a file of its own into a request.
/ LOCKFLARE SONDA

Schedules and alerts, in the free edition, up to 10 seats.

Schedules and alerts, listed →