This page explains what you can rely on flowss for and what you should not: the service commitments flowss does and does not make, how to read the public status page (everyone sees how many checks are degraded; only an administrator sees which) and its machine-readable twin, what keeps working when your connection drops, how app updates arrive, how backups work and why your own exports still matter, every service limit and outage message you are likely to meet, and how to get help from a person — with what to include so the answer comes quickly. For how your account and data are protected, see Security at flowss and Privacy and your data.
At a glance
| Question | Short answer |
|---|---|
| Is there an uptime guarantee or SLA? | No. The terms say so plainly. A 99.9% SLA is listed for Enterprise as "At GA" and is not available today. |
| Where do I check the service? | /status. It reports configuration checks, service set-up, real page-speed measurements and what has shipped. |
| Does a green badge mean nothing is down? | No. It reads "All configuration checks pass" and does not test the database, billing or AI providers. |
| Is my work backed up? | The database is backed up daily, but there is no point-in-time recovery and no restore guarantee. Keep your own exports. |
| What works offline? | A studio that is already open keeps working. Opening a studio, signing in, cloud sync, AI and server-rendered engines need a connection. |
| When do limits reset? | Daily limits reset at midnight UTC. |
| How do I reach a person? | The private support form at /contact#support-request, or email hello@flowss.ai. |
| How fast is support? | A person reads every message, during working hours. There is no instant-response guarantee. Paid plans get priority on the same address. |
| Where is self-serve help? | /support — documented help topics, with optional AI topic matching and local voice. |
What flowss commits to, and what it does not
flowss is built and run by a small team, and the commitments are deliberately modest so they can be kept.
| Commitment | Status |
|---|---|
| Uptime | No commitment. The terms state that flowss does not promise the service will be available, uninterrupted or free of errors, and that no service-level agreement has been published. |
| Service-level agreement | None today. "99.9 % SLA" appears in the pricing comparison for Enterprise only, marked "At GA". |
| Warranty | The service is provided "as is" and "as available", subject to your local consumer-protection law. |
| Paid features | flowss will not remove something you are paying for mid-term without telling you first. |
| Changes to the terms | Material changes are announced in-app or by email at least 14 days before they take effect. |
| Non-payment | If payments keep failing, an account may be moved to the Free plan. Your work is not deleted for non-payment; paid features stop and your documents stay where they are. |
| Support response | A person reads every message, in working hours. No response time is promised. |
| Security reports | Acknowledged within three working days, assessed within fourteen. See Security at flowss. |
Note: flowss also depends on other companies — payments, the database and sign-in, AI providers, and the outside render services some engines use. An outage at one of them can affect flowss even when flowss itself is running.
The status page
/status is public and needs no account. Its headline is "Counted, not claimed.": the inventory numbers are counted from the product as deployed, and the health badge is evaluated each time you load the page.
The health badge
| Badge | Meaning |
|---|---|
| All configuration checks pass (green) | Nothing is missing from the deployment's configuration in a way that would quietly change behaviour, and the server that answered you has recorded no recent email-sending failure |
| 1 health check is degraded or N health checks are degraded (amber) | That many configuration checks are failing. Which ones is shown only to flowss administrators |
Open What this health check measures under the badge for the exact wording. The important part: the badge is not an outage detector — in the page's words, "the database, billing and AI providers are not checked by this badge, so their being down does not turn it amber." The email-failure half is also best-effort, because the server that answers you may not be the one that saw a failure. No live connectivity test runs when you load the page. The amber count includes every failing check, even those whose details are withheld. A line such as "build 1a2b3c4" under the badge names the release that is running, which is useful to quote in a bug report.
Inventory tiles
Five tiles count what is deployed, straight from the product's own registries: Surfaces, Engines, Templates, Categories and Native renderers. They answer "has the thing I need actually shipped?" rather than "is it working right now?".
Services
A card for each optional service, marked configured or not configured:
| Card | What the card says it covers |
|---|---|
| Accounts & cloud sync | "Sign-in, saved diagrams, collaboration, per-account quotas." |
| Transactional email | "Invites, receipts, security notices, waitlist confirmations." |
| Billing | "Checkout, customer portal, subscription webhooks." |
| Hosted AI | "Server-side generation. Bring-your-own-key works either way." Marked configured when either hosted AI provider is set up |
| Living diagrams | "GitHub App — diagrams that follow a repository." |
"Configured" means the service's keys are present, not that the provider is answering right now.
Speed, as visitors measured it
This section summarises Core Web Vitals reported anonymously by real page loads over the last 7 days (no IP addresses, no page addresses, and Do-Not-Track is honoured). The figures are recalculated at most every five minutes, so a reload within that time usually shows the same numbers.
| Measure | What it is | Good at or under |
|---|---|---|
| LCP | Largest Contentful Paint — when the main content is on screen | 2.5 s |
| INP | Interaction to Next Paint — how quickly the page answers a click or key | 200 ms |
| CLS | Cumulative Layout Shift — how much the page moves under you | 0.1 |
| TTFB | Time to First Byte — how fast the server begins to answer | 800 ms |
Each card shows the 75th-percentile value (the number the web is judged on), the median and the number of loads, rated good, needs improvement or poor (poor starts above 4 s LCP, 500 ms INP, 0.25 CLS and 1.8 s TTFB). A table below breaks LCP, INP and CLS down by workstation. In the page's words, "A vital is only recorded once a page is hidden, so today's loads arrive later today." A measure is shown only once at least 20 real page loads have reported it; until any measure qualifies, the section reads "Collecting." rather than showing an estimate.
Surfaces and engine coverage
The rest of the page lists the main public surfaces — Landing, Studio, Evidence, Weave, Science, BPMN, Sketch, Study, Templates, Docs, Use cases, Embed, Blog, Pricing and Attributions — each marked live and linked, and a table of every engine with its id and how many templates use it, sorted by template count. The footer offers "Heard something feels off? Email" with a link that opens an email to hello@flowss.ai with the subject "Something looks wrong".
Machine-readable health
If you run your own monitoring, the status page's check is also served as JSON at /api/health. It is free, needs no sign-in, runs no connectivity checks, is never cached (Cache-Control: no-store) and always answers HTTP 200 while the service is up — the verdict is in the body, not the status code.
| Field | Meaning |
|---|---|
ok | Always true when the service answers |
status | ok, or degraded when any configuration check fails — alert on this field |
release | The short identifier of the running build, or null |
services | supabase, email, stripe and github as true or false, plus ai with anthropic and openai — the same "configured" flags as the status page |
time | When the answer was generated |
Deeper connectivity checks and the names of failing checks are available only to flowss operators.
Working offline and receiving updates
flowss can be installed as an app from /install, and keeps working through a dropped connection in the tab you already have open.
| Situation | What happens |
|---|---|
| Your connection drops while a studio is open | A banner says "You’re offline. Keep your current work open. AI, cloud saves and sharing need a connection." The editor keeps working |
| You open a page with no connection | The You’re offline page explains that "an editor that is already open keeps working, but opening one needs a connection", and offers Open home and Library |
| The app cannot load at all offline | A fallback page labelled Connection paused ("Your next idea can wait a moment.") says "a studio that is already open keeps working. Opening a studio needs a connection." and offers Open Library and Try home |
| Things that always need a connection | Signing in, opening a studio, cloud sync, AI features, sharing and collaboration, and the engines that render on a server (PlantUML and the Kroki-backed engines) |
Studio pages are never stored for offline use, because each needs your signed-in session. Export important local work before closing a tab while offline. The details are in Cloud sync, offline work and devices.
When a new version is ready
When flowss is updated while you are working, a card titled A fresh version is ready appears at the bottom right: "Finish your work, then update when you’re ready. This window stays open until you choose."
- Finish or export what you are doing.
- Choose Update & reload. A confirmation asks: "Save or export any unfinished work in your open flowss windows first. Update and reload this window now?"
- Confirm. The window reloads on the new version.
Choose Later (or the close button, "Remind me later") to keep working; a small Update available button stays in the corner until you are ready, and choosing it brings the card back. The card and the button appear only while you are online. While the update applies, the button reads "Updating…". If the update takes more than about 15 seconds you see "The update is taking longer than expected. Your current work remains open; try again when ready."
On /install, Check for updates checks on demand ("Checking…") and reports "Update available — save your work before reloading.", "You’re using the latest available app version." or "Couldn’t check. Reconnect and try again." The same line otherwise reads "App support is active in this browser.", "Preparing app support…" or, when this browser window cannot run app support, "Use a normal, secure browser window to enable app support." Check for updates is available only once app support is active.
Backups and keeping your own copy
| Layer | What it gives you | What it does not |
|---|---|---|
| Your browser | An immediate local copy of everything you edit | Protection against clearing site data, a lost device, or a browser that evicts storage |
| Cloud sync | Your workspace on every device you sign in on | A history you can roll back to; it is a convenience, not custody |
| Database backups | Daily backups, kept for our database provider's retention period, to recover the service from a failure | Point-in-time recovery, a restore guarantee, or restoring one account or one diagram — deleted accounts are not restored |
| Your exports | A copy you control | — |
Every diagram exports from its studio as SVG, PNG, PDF, TikZ and JSON source, and the JSON round-trips back in. Account → Your data → Download my data downloads everything flowss holds about you as one file. For important work, keep both. See Exporting your work, Version history and undo and Privacy and your data.
Service limits and outage messages
Daily limits reset at midnight UTC. In the messages below, N stands for the number the message shows — your current cap — plan for your plan's name, and code for a status number. Higher plans raise the caps for hosted AI, server-side exports, TikZ conversion, PlantUML and Kroki rendering, and Evidence literature search. AI points are covered in AI points and limits.
| Message (as shown) | Cause | What to do |
|---|---|---|
| "We couldn't check the request limits just now, so we haven't run this. Nothing has been charged — please try again in a moment." | The limit counters were briefly unreachable, so the request was refused rather than risked | Try again in about 30 seconds |
| "Hosted AI on flowss has reached today's safety limit. It resets at 00:00 UTC. …" | The platform-wide daily safety limit for hosted AI was reached; nobody's personal quota | Wait until midnight UTC, or use your own AI key if your plan allows. No points were taken |
| "You've reached today's hosted-AI limit for one account on flowss. It resets at 00:00 UTC. …" | One account's share of the day's hosted AI was used | As above |
| "Hosted AI is temporarily unavailable because its safety budget could not be reserved. Nothing was charged. …" | A safety check for hosted AI could not run | Try again later; your own key still works where your plan allows one |
| "Daily limit for AI requests on your own key reached (N/day). It resets at midnight UTC — nothing was sent to your provider." | The abuse ceiling for requests on your own AI key | Wait until midnight UTC |
| "Daily limit for agent runs on your own key reached (N/day). …" | The separate ceiling for agent runs on your own key | Wait until midnight UTC |
| "Server-side export limit reached for your plan (N/day on plan). It resets at midnight UTC; a higher plan raises the cap." | Too many server-side exports today | Wait, or upgrade; exports your browser can produce itself are not affected |
| "TikZ conversion limit reached for your plan (N/day on plan). …" | Too many AI TikZ conversions today | Wait, upgrade, or use your own AI key, which is never metered here |
| "PlantUML render limit reached for your plan (N/day on plan). …" | Too many PlantUML renders today; each pause in your typing is a render | Wait until midnight UTC, or upgrade |
| "Diagram render limit reached for your plan (N/day on plan). …" | The same for the Kroki-backed engines | As above |
| "Literature search limit reached for your plan (N/day on plan). …" | Too many Evidence literature searches today | Wait or upgrade |
| Any of the above with "(N/day for anonymous use)" and "Sign in to lift this cap" | The request came without a signed-in session (for example an embedded diagram viewed by a visitor), so it counts against a shared per-network allowance | Sign in, or open the diagram in flowss; for client-side exports, use the toolbar's PNG or SVG, which is not metered |
| "You've reached today's online icon search limit. The built-in icons still work — or try again tomorrow." | Too many online icon searches from your network today | Use the built-in icon sets, or try tomorrow |
| "Rate limit reached (N/day). Try later." | A daily API or Evidence analysis limit | Wait until midnight UTC; for the API see The REST API |
| "Too many checkout attempts. Please wait a while and try again — nothing has been charged." | Too many checkout starts today | Wait and retry |
| "Too many billing requests. Please wait a while and try again." | Too many billing-portal openings today | Wait and retry |
| "Too many invitations from here today. Share a project link instead, or try again tomorrow." | The daily limit on invitation emails, per sender and per invited address (teams say "Share the team link instead") | Share the link, or invite tomorrow |
| "Too many account operations today. Try again tomorrow." | Account export and similar operations | Try after midnight UTC |
| "Feedback limit reached (N/day). Please try again tomorrow." | Too much feedback from one network today | Email hello@flowss.ai instead |
| "The diagram render service could not be reached." | The outside render service for a Kroki-backed engine is down or unreachable | Try again later, or switch to an engine that draws locally |
| "The diagram render service is rate-limiting this deployment. Try again shortly." | The outside render service is throttling flowss | Wait a few minutes |
| "The diagram render service refused the request (code). This is a configuration or access problem, not your diagram." | The render service refused access | Report it with the code; your diagram is fine |
| "PlantUML server returned code" | The PlantUML render server failed | Try again later |
| "PlantUML source is too large (>30,000 characters)" | PlantUML source over the size cap | Split the diagram |
| "Request body is too large (>N KB)." | A request exceeded the size allowed | Send less at once, or split the document |
| "Email sign-in is briefly unavailable: too much email has gone out in a short time, …" | Sign-in email is paused across the service | Use Continue with GitHub or Continue with Google, or try later |
When something goes wrong in the app
| What you see | What it means | What to do |
|---|---|---|
| This page hit an unexpected error (or the name of the studio followed by "hit an unexpected error") under "Something went wrong" | One part of the page crashed. "The rest of flowss is fine and your local work is saved." | Choose Try again to re-render, or Library to go back. If it repeats, report it with the message and the "digest" line shown in the box |
| flowss hit an unexpected error — "The app failed to load." | The whole app failed to load | Choose Try again, or Home (which reloads the app from scratch). If it persists, check /status (it shows everyone whether any check is degraded; only an administrator sees which) and contact support, quoting the "digest" line if one is shown |
| This canvas is blank. — "We couldn't find that page. Try one of the workstations instead." | The address does not exist | Use one of the studio links or the status page |
| Sync banner "Cloud backup is failing — your work is saved on this device only." | Three or more cloud saves in a row failed | Check your connection, then export anything important; see Cloud sync, offline work and devices |
Crash reports are recorded automatically so the team can fix them; you do not need to report every one, but a report with steps is what gets a recurring problem fixed fastest.
Getting help
The help page
/support answers common questions from reviewed flowss documentation. "This help selector cannot see your account, change your work or manage payments."
- Choose a topic under What can we help with?: Create your first diagram, Save and find your work, Export a diagram, Sign-in and confirmation help, Billing and cancellation, AI access and limits, Privacy and account data or Something is not working.
- Or type into Or describe what you need (up to 500 characters — "Keep it general. Do not include passwords, API keys, payment details or private project content.") and choose Find help. The match is made in your browser.
- Optionally tick "Use the flowss Support Agent to select a help topic. I agree to send this question to Anthropic when I choose Find help." This needs you to be signed in; the flowss Support Agent only picks a topic, and flowss does not save the question. When it is off, the page says "The flowss Support Agent is currently off. Topic matching runs in your browser and does not send your question to a model."
- Read the answer, labelled "Documentation topic" or "Selected by the flowss Support Agent", and choose Read the source for the full page.
- If nothing fits, choose Contact a person. "Your question is not forwarded automatically."
To use your voice, tick "Enable local voice for this visit". Dictate question listens for up to 30 seconds using your device's own English speech recognition ("Microphone on. Listening locally…"); choose Stop dictation to finish early, then review the transcript before choosing Find help. Read answer aloud reads the answer with a local voice (Stop reading ends it). Voice stops when you switch to another tab. Clear resets everything.
Questions that look like they contain a password, key, card number or email address, that mention a security problem, or that ask for a person (words such as "human", "person" or "agent") are never sent to the AI; the page says "This question is better handled privately by a person. Nothing was sent to the AI provider." and points you to Contact a person.
Sending a private support request
The form on the contact page, Send a private support request, keeps a reference you can track. You need to be signed in with a confirmed email address; if you cannot sign in, email instead.
- Choose a Topic: Product, Bug, Billing, Privacy, Security or Other.
- Enter a Subject (5 to 160 characters).
- Describe What happened? (20 to 6,000 characters): the page, the steps, and what you expected.
- Tick "I have reviewed this message and want to send it to the flowss support team."
- Choose Send to a person.
You see "Saved as FS-…" with a reference, and either "A receipt has been sent to your account email." or "Email notifications are pending; your request is saved for human review." Your recent requests lists up to 20 requests with their status — open, in progress, waiting on customer or resolved — and date. Keep the reference and reply to the receipt email to follow up. Requests are visible to you and to flowss administrators.
Emailing support
Everything goes to one monitored address, hello@flowss.ai. The contact page's subject cards open an email with the right subject already filled in, which gets it to the right place faster:
| Card | Subject | Use it for |
|---|---|---|
| Something is broken | flowss — Bug report | The page, what happened, what you expected |
| How do I…? | flowss — Product question | Features, exports, workflows |
| Billing, plans and refunds | flowss — Billing | Plans, invoices, payments; include your account email and checkout reference |
| Privacy and your data | flowss — Data request | Deletion requests, corrections, privacy questions |
| Security | flowss — Security report | Vulnerabilities, sent privately with steps to reproduce |
| Teams, education and everything else | flowss — Teams | Classrooms, team rollouts, partnerships |
What to expect
- A person reads every message. Replies come in the team's working hours rather than instantly; no response time is quoted.
- Where your plan includes priority support — priority email support from Starter upwards, and a named customer success manager on Enterprise — it uses the same address, prioritised. See Plans and what they include.
- Bug reports always get an answer, even when it is "confirmed, it's on the list".
- Writing to support does not add you to a mailing list.
- Never send a password, an API key, a one-time code or a sign-in link to support. flowss's own sign-in emails say "We will never ask you for your password or for this link."
Writing a bug report that gets fixed fast
- Name the studio and page, and the approximate time it happened (with your time zone).
- List the steps that cause it, starting from a fresh page if you can.
- Say what you expected and what happened instead, quoting any message word for word.
- Add the "digest" line from an error box, and the build identifier from /status, if you have them.
- Include your browser and operating system.
- Attach a minimal, redacted example only if needed — never confidential project content, passwords or keys.
Feedback from inside the app
For quick suggestions, open the Flow Dock (⌘J), choose Feedback at the bottom, type up to 2,000 characters in "Something missing? Something broken? Tell us —" and choose Send feedback. You see "Thank you — feedback received." Signed-in feedback carries your account so the team can reply; it is not a support ticket and has no reference. Use the support form for anything you need answered.
Staying informed
The blog carries release notes and engineering write-ups. Material changes to the terms or the privacy policy are announced at least 14 days before they take effect. The status page's tiles show what has shipped as of the build you are using.
Tips
- Bookmark /status, but remember the green badge checks configuration, not live availability.
- If one engine fails to render while others work, the outside render service may be down; an engine that draws locally will keep working.
- Before clearing browser data, switching devices or closing a tab offline, check the sync status and export anything important.
- Apply app updates when you are at a natural stopping point; the update card waits for you.
- Use the support form rather than plain email when you want a reference and a visible status.
- When a limit message names a number, that is your plan's cap for today; it resets at midnight UTC.
Limits and known constraints
- No uptime commitment, SLA or guaranteed response time.
- The status badge does not detect outages of the database, billing or AI providers, and no live connectivity test runs on the public page.
- No point-in-time recovery; deleted accounts and individual documents are not restored from backups.
- Opening a studio needs a connection; studio pages are not available offline.
- The support form needs a signed-in account with a confirmed email; otherwise use email.
- AI topic matching on the help page needs you to be signed in and only selects a help topic.
- Local voice on the help page works only in browsers with English speech support already installed, and has no remote fallback.
Troubleshooting
| What you see | What it means | What to do |
|---|---|---|
| "Sign in to submit and track a request, or use the support email above if you cannot sign in." | You are not signed in | Sign in, or email hello@flowss.ai |
| "Choose a topic, use a subject of 5–160 characters and describe the issue in 20–6,000 characters." | A support form field is out of range | Adjust the subject or description length |
| "Confirm your account email before submitting, or email support directly." | Your email address is not confirmed | Confirm it, or email support |
| "Too many support attempts today. Please email support." or "Too many support requests today. Reply to your existing request or email support." | The daily limit on support requests | Reply to your existing request's receipt, or email |
| "This request reference already belongs to a different draft. Start a new request." | A retried submission was edited in between | Start a new request |
| "Support requests are unavailable right now. Email support instead; your draft has not been submitted." | The support form's service is down | Email hello@flowss.ai |
| "Could not submit your request. Please retry with the same draft." | A temporary failure | Retry; the draft is kept and will not be duplicated |
| "Sign in to use the flowss Support Agent. Help topics and human support remain available." | AI topic matching needs an account | Sign in, or untick the flowss Support Agent and use local topic matching |
| "The flowss Support Agent has reached its daily limit. Use the help topics or contact a person." | The flowss Support Agent's allowance is used up | Use the topics or contact a person |
| "The flowss Support Agent is temporarily unavailable. Use the help topics or contact a person." | The AI provider could not be used | As above |
| "Review the transcript, then choose Find help. It has not been submitted." | Dictation finished | Check the text, then choose Find help |
| "The flowss Support Agent could not answer. Use the help topics or contact a person." or "The flowss Support Agent is unavailable. Use the help topics or contact a person." | The AI provider did not answer, or AI help is switched off | Use the topics or contact a person |
| "Local dictation is unavailable in this browser. Please type your question." or "This browser cannot guarantee local dictation. Please type your question." | Your browser has no on-device speech recognition | Type the question |
| "An English on-device speech pack is not available. Please type your question; no pack was downloaded." | On-device recognition exists but has no English pack installed | Install English speech support in your browser or system settings, or type |
| "Dictation stopped or microphone permission was denied. You can type instead." | The microphone was refused or dictation ended early | Allow the microphone for this site, or type |
| "No local English voice is available yet. Read the answer or try listening again." | No local voice is installed for read-aloud | Read the answer on screen |
| "Sign in to send a support request." | Your session ended before you sent the form | Your draft stays in the form; copy it, sign in again, and resubmit |
| "Couldn't send feedback." | The feedback request failed | Retry, or email support |
| "That update is no longer available. Check for updates from the install page." | A newer update replaced the one offered | Use Check for updates on /install |
| "Couldn’t check. Reconnect and try again." | The update check could not reach flowss | Reconnect and retry |
