HTTP jobs that fire on time, every time. POST a URL and a fire time — a one-off job, a cron schedule, or an ordered buffer — and Fliq schedules, retries, and rate-limits it, survives crashes, and keeps every attempt on record. No queues to run. No workers to babysit.
100,000 executions free, every single day
jobs picked up in under 2 seconds
nothing gets lost — every attempt retried & recorded
nothing to run — no queues, brokers, or cron boxes
Fliq replaces the lot with one REST API — and shows you every attempt it ever made.
Everything is plain HTTP in, plain HTTP out. Pick the shape that fits, and Fliq handles execution, retries, ordering, and history.
POST /jobs
POST a URL, method, headers, body, and a time. Fliq executes it on the dot, retries with exponential backoff and jitter, and records every attempt — status code, duration, error. Cancel it, query it, or get a webhook when it completes.
idempotency keys · per-job timeouts up to 1h · webhook callbacks
POST /schedules
Standard five-field cron expressions that fire real jobs. Pause and resume on demand, trace every run back to its schedule, and forget thundering catch-up bursts after downtime — missed runs skip, by design.
5-field cron · pause / resume · per-run job history
POST /buffers/:id/items
Point a buffer at a rate-limited API and push as fast as you like. Fliq delivers in strict submission order, at your configured rate, one item in flight at a time — and honors Retry-After when the target pushes back.
strict FIFO · 1–1000 req/s · retries that don’t break order
No SDK required, no infrastructure to stand up. One authenticated request and your job is durable, retried, and auditable.
# pay an invoice tomorrow at 9am, retry up to 5 times
$ curl -X POST https://api.fliq.sh/jobs \
-H "Authorization: Bearer fliq_sk_live_…" \
-d '{
"url": "https://api.stripe.com/v1/invoices/in_1JX/pay",
"method": "POST",
"scheduled_at": "2026-06-12T09:00:00Z",
"max_retries": 5,
"backoff": "exponential",
"webhook_url": "https://you.dev/hooks/fliq"
}'
201 Created
{ "id": "job_8f2…", "status": "pending" }Create a buffer with a target URL and a rate — say 5 req/s. Push ten thousand items. Fliq delivers them in submission order, at most that fast, one in flight at a time. Failures retry in place without breaking order, and a 429’s Retry-After is honored without burning the item’s retry budget.
queueing items…
Fliq is young, so we don’t quote invented uptime numbers — we show our live production status and design for failure instead.
Every call carries a stable X-Fliq-Delivery-Id across retries and
crash recovery — dedupe on your side with one header.
Workers heartbeat every 10 seconds. A reaper notices the silence, rescues the job, and retries it. Your job outlives our process.
Pending jobs are claimed within about two seconds of their fire time — indexed polling, no batch windows.
Exponential 30s → 1h cap with ±25% jitter, or linear — your choice, per job. Retries that don’t stampede your target.
Optional X-Fliq-Signature (HMAC-SHA256 over timestamp, method,
URL, body) so your endpoint can verify every request’s origin.
Every execution is recorded — worker, status code, duration, error — and queryable per job. When something fails at 3am, you can see exactly why.
When a job exhausts its retries, Fliq pings your Slack or a webhook — once, on permanent failure, not on every retry. The signal a reliability tool owes you.
Fix the endpoint, then re-run a failed job or buffer item with one call. The original is kept for audit; the replay drains in order, with its own clean history.
Success rate, p95 duration, and per-day execution counts — jobs vs buffers — over any window. See what you’re paying for and what’s breaking, without leaving Fliq.
Job claiming is FOR UPDATE SKIP LOCKED. The rate limiter is a token
bucket stored in a row, refilled in the same atomic UPDATE that claims
the next item — correct across replicas, with no coordinator to operate. One
database. Boring, the way infrastructure should be.
One credit per execution attempt. Retries included in the math, so the bill is never a surprise. See full pricing →
Free during the beta — generous limits, real production API.
One attempt to call your target URL. Retries count as separate executions — one credit each, no hidden multipliers. $1 buys 100,000 executions, and the free tier refreshes 100,000 every day at UTC midnight.
At-least-once. Every call carries a stable X-Fliq-Delivery-Id header that survives retries and crash recovery, so your endpoint can deduplicate safely. Buffers additionally guarantee strict submission order with one item in flight at a time.
Workers heartbeat every 10 seconds while executing. A reaper process detects the silence within about 30 seconds, rescues the orphaned job, and retries it with the same delivery ID. Nothing is lost and nothing fires twice without you being able to detect it.
Fliq honors the Retry-After header and waits exactly as long as the API asked — without burning the item’s retry budget and without breaking submission order. Successors wait their turn; nothing is reordered or dropped.
Yes. Enable request signing and every outbound call includes X-Fliq-Timestamp and X-Fliq-Signature headers — an HMAC-SHA256 over the timestamp, method, URL, and body. Rotate the secret any time via the API.
Yes — the Go backend and the dashboard are open source at github.com/fliq-sh. Fliq is free during the beta: 100,000 executions a day, no card required.
Yes. Fliq ships an MCP server, so agents can schedule jobs and rate-limit their own outbound API calls in natural language — useful for agents that talk to Stripe, Shopify, or LLM providers without tripping rate limits.