Monitoring for teams without an SRE
Flowhawk follows what your customers do, like signing up or checking out, not just your servers. When checkout starts failing, you see how many orders it cost, which deploy most likely caused it, and the traces behind it. The agent then digs in and suggests a fix, which only runs if you approve it.
Start with $10 of usage free for 14 days. Stripe checks your card but charges nothing. After that, pay as you go from $30. No per-seat fees. Works with OpenTelemetry.
These screenshots come from a demo shop where a bad deploy broke checkout.
Most monitoring tools are organised around servers and services. Flowhawk is organised around your business flows, such as sign-up, checkout and payouts, so an alert tells you which one is broken and how many customers it affects.
Flowhawk suggests flows from the first few hours of traffic, or your coding agent drafts them from your routes over MCP. You check them in an editor and mark which ones matter most. After that, every span and log is tagged with its flow and step.
Traces, logs, errors, metrics, sessions, AI calls and database queries share the trace ID. You can go from a log to its trace, from a point on a chart to example traces, and from an alert to its flow.
When an incident opens, Flowhawk builds a page for it with the orders or sign-ups affected, the failing step, the last change before it and the traces. Click any number to see the query behind it.
You can ask the agent questions. It also looks into incidents and suggests changes as diffs for you to review. If it needs more evidence, it can raise sampling on a broken flow for a while.
Flows, alerts, dashboards, collection settings and page templates are YAML files kept in git. The UI and the agent edit the same files, and every change is a diff you can review and undo.
Claude Code, Cursor and other coding agents can read production data, incidents and flows over MCP while they work on a fix.
There are SDKs for Node.js, Python, Go, Java, .NET, Ruby and PHP, and for the browser, React Native, Flutter, iOS and Android. They are thin wrappers around OpenTelemetry. You set sampling per flow in Flowhawk, so you can keep fewer healthy requests and every error without redeploying.
// Node.js: npm install @flowhawk/sdk-node
require("@flowhawk/sdk-node").init({ service: "orders" });
# Python: pip install flowhawk-sdk
import flowhawk_sdk
flowhawk_sdk.init(service="recommendations")
// Go
flowhawk.Init(flowhawk.Options{Service: "payments"})
// Java and Kotlin
Flowhawk.init(Flowhawk.options("orders"));You top up a balance in dollars, and storage and agent work are paid from it. There are no plans and no per-seat fees, so the whole team can use it. You can try it first with $10 of usage for 14 days, and when you top up, your workspace and data carry on as they are. If you top up and it isn't for you, ask within 30 days and we refund what you haven't spent.
$10 for 14 days
Every feature, on your own services. Stripe checks your card but doesn't charge it, and we only charge it if you top up. There is one trial per person, card and company. You can send up to 2 GB a day, and on-call texts and calls are off. Everything pauses when the $10 or the 14 days run out.
Try it free$30 first top-up
Includes every feature and 30 days of retention. Later top-ups start at $10, or you can turn on auto-reload. A 10-person team with light use spends about $30 a month, the same as the first top-up.
Start for $30| Each GB of telemetry you send, kept for 30 days (billed as $2.00 per compressed GB, which holds about 10 GB) | about $0.20 |
| Agent investigation | $1.25 |
| Agent question | $0.25 |
| Generated page | $0.20 |
| Uptime check or heartbeat beyond the first 10, per month | $0.20 |
| On-call text message / phone call | $0.10 / $0.25 |
| Alerts, dashboards, flow rollups, push and email pages | free |
| Light use, 10-person team | |
|---|---|
| 25 GB of telemetry sent | $5.00 |
| 10 agent investigations | $12.50 |
| 50 agent questions | $12.50 |
| 10 people, alerts, dashboards, on-call by push and email | $0.00 |
| Total | $30.00 |
| Tool | Entry price | Seats | AI agent | Free start |
|---|---|---|---|---|
| Flowhawk | Pay as you go, about $30 a month for a light team | No fee | $1.25 an investigation, $0.25 a question | $10 free for 14 days, and unspent money back within 30 days |
| Better Stack | $30 a month for 40 GB each of logs, traces and metrics | $34 a month per on-call responder | Billed per token, price shown in the app | Free plan |
| Sentry | $29 a month (Team) with 50k errors, 5M spans and 5 GB of logs | No fee | Seer, an add-on at extra cost | Free plan for one user |
| Honeycomb | Free up to 20M events, then from $150 a month (Pro) | No fee | Included | Free plan |
| SigNoz | $49 a month minimum, then $0.30 per GB of logs or traces | No fee | Included | Free trial |
| Datadog | $18 per host for infrastructure plus $36 per host for APM, logs $0.10 per GB plus indexing | No fee | AI credits from $500 a month | Free trial |
Competitor prices are their own published list prices on the date above, billed monthly where a page gives both. They change, and each tool counts usage differently, so check their pages for your case.
Prices exclude VAT and sales tax, which are added at checkout where they apply, and every top-up comes with an invoice. Top-ups stay valid for 12 months. Annual invoices, longer retention and SSO add-ons are available on request.
Every option has the same app and agent. What changes is where your telemetry is stored and what you pay for storage.
The default. We run everything on shared servers in Germany, with ClickHouse replicated across two machines and encrypted nightly backups. You only send data.
Keep your telemetry on ClickHouse Cloud or your own server, written there by our collector. Storage isn't charged, and the agent and everything else cost the same. Ask us to turn it on for your workspace.
We run ClickHouse inside your AWS, Google Cloud or Azure account, and your cloud provider bills you for the machines. It costs $300 a month from your balance. This is in early access, so ask us at hello@flowhawk.dev.
There is no self-hosted edition of the whole platform yet.
Telemetry, config and backups are stored in Germany, on Hetzner servers in Falkenstein and Nuremberg. When you ask the agent a question or it investigates an incident, the parts it needs are sent to Anthropic's API, which may process them outside the EU. Payments go through Stripe and email through Resend. The full list is on the subprocessors page, and the DPA covers GDPR.
Yes. The SDKs and the collector redact common secret and personal attributes by default, even before your policy loads, and strip LLM prompts unless you turn prompt capture on. You can add your own drop and redact lists per flow.
No. Flowhawk takes standard OpenTelemetry over OTLP with an ingest key, so the instrumentation you have keeps working. Our SDKs add flow and step tags and remote control of collection on top.
Anything that touches your systems, like a rollback or a config change, is a proposal that runs only after someone approves it. On its own, it can only raise sampling on a broken flow for a limited time, within limits you set. It also investigates new incidents by itself, at $1.25 each, up to 20 a day by default; you can lower that limit or set a monthly spend cap.
We email the owners when it runs low and when it reaches zero. A grace period covers about a day of usage. After that, new telemetry and the agent pause until you top up, and the data you already have stays readable. Auto-reload and a monthly spend cap are in Settings.
Yes. A new workspace gets $10 of usage for 14 days. Stripe checks your card first and charges nothing, and we only charge it if you top up. There is one trial per person, card and company. During the trial you can send up to 2 GB a day, and on-call texts and calls are off. When the $10 or the 14 days run out, new data and the agent pause until you top up, and your data stays. If you top up and change your mind, ask within 30 days at hello@flowhawk.dev and we refund what you haven't spent.
Flowhawk runs on ClickStack, the open-source ClickHouse observability stack, so storage and queries are fast and cheap underneath. ClickStack gives you the data and a UI to search it, and you build the dashboards and alerts yourself. Flowhawk adds the parts on top: your business flows, a page for each incident, the agent, and collection that it sets up and adjusts for you.
Yes. See where it runs: both are available on request, and with your own ClickHouse you pay no storage charges.
Add an SDK and send some traffic. Flowhawk drafts your flows from the first few hours of it.