Documentation

Every Kinode product — what it does, how to use it, and what to do when it misbehaves.

What Kinode is

Kinode turns spoken audio and video into text you can read, search, repurpose, and act on — and then builds a family of products on top of that one capability.

At the center is a transcription engine. Paste a link from YouTube, Vimeo, TikTok, Twitch, Rumble, or a podcast feed, or upload your own file, and Kinode returns a clean, timestamped, punctuated transcript. Everything else in the product line is something useful you can do with that transcript: summaries and study kits, automatic show notes, SEO pages, brand monitoring, scheduled briefings, live meetings that transcribe themselves per speaker, and cryptographically sealed records.

The products share one account, one plan, one quota, and one transcript cache. A video transcribed in the workspace is instantly available to Monitor, Digest, Pages, Study, and the API without being fetched or billed again.

The product line at a glance

ProductIn one lineWhere it lives
TranscriptsThe core workspace: paste a link, get a transcript, generate insights, chat about it./app
Show NotesConnect an RSS feed; every new episode gets notes, chapters, quotes, and a newsletter draft./shownotes
StudyBundle lectures into a kit: one study guide, flash cards, and quizzes./study
PagesPublish transcripts as hosted, SEO-ready public pages./pages
MonitorTrack brand mentions inside the spoken audio of channels and feeds./monitor
DigestA daily or weekly briefing of everything new from your watched sources./digest
MeetingsBrowser video meetings that transcribe each speaker from their own mic./meetings
CertifyTurn a meeting transcript into a digitally signed, tamper-evident PDF./meetings
Support BotAI customer support, ticketing, and identity verification inside Discord./supportbot
PDFA full in-browser PDF editor with real X.509 digital signatures./pdf
ClipsFind the most clippable moments in any video./app
APIEvery capability above as a documented REST endpoint./api
One quota across everything

Transcription minutes and AI generations draw from a single plan quota shared by the workspace and the API. Because transcripts are cached, a source is only ever billed the first time anyone fetches it — repeat visits, insights, and API calls against the same video are free and instant.

Accounts & the workspace

Creating an account

Sign up at /signup with an email and password. There is a free tier for trying the core workspace; paid plans raise your monthly quota and unlock the products that watch feeds on a schedule (Show Notes, Monitor, Digest) and the API.

The workspace

The workspace at /app is where the core loop lives: paste a link or upload a file, watch it transcribe, read the result, and generate insights. Everything you transcribe is saved to your history at /history, so you can return to any transcript and its generated outputs without re-fetching.

Organizations & teams

An account can create an organization and invite members. Organizations share a plan and quota, and some products are org-aware: Study lets an owner assign a kit to the whole org and track completion; Support Bot verifies customers against org accounts. Manage members and invites at /organization.

Dark and light

The marketing site is light by default; the console (the workspace and every product page) is designed dark, with a theme toggle in the header. Your choice is remembered per device.

Core concepts

Five ideas explain how nearly everything in Kinode behaves. Read these once and the individual products make sense quickly.

Sources

A source is anything Kinode can turn into a transcript. Seven kinds run through the same pipeline:

SourceHow you provide it
YouTubePaste a video, playlist, or channel URL.
Vimeo, TikTok, Twitch, RumblePaste the video URL.
PodcastsPaste an RSS feed URL, or a direct episode link.
Your own filesUpload audio or video directly at /upload.

The transcription pipeline

When a source has a caption track, Kinode fetches it and cleans it up with AI — restoring punctuation, capitalization, and paragraph breaks — so it reads like a document instead of a caption dump. When a source has no captions, it goes through speech-to-text. Either way, every line keeps a timestamp that links back to that moment in the original.

Insights

Once a transcript exists, one click generates an insight from it: a summary, chapter markers, key quotes, a study guide, a quiz, flash cards, highlights, suggested clips, or a translation into 100+ languages. There is also a grounded chat that answers questions using only the transcript and cites timestamps — the fastest way to find what someone said in an hour of video.

Caching & billing

Transcripts are cached globally. A given video is fetched and billed once; every later visit, insight, or API call against it is free and instant, for you and for everyone. If you work with the same channels repeatedly, this is the single biggest lever on your cost.

Why the cache matters

Products that watch feeds — Monitor, Digest, Show Notes — ride the same cache. A video any Kinode user has ever transcribed returns instantly and doesn't count against your quota. This is why watching busy channels stays affordable.

Quota & credits

Your plan sets a monthly quota of transcription and AI generation. The workspace and the API draw from the same pool, so there's no separate API bill to reason about. When quota is exhausted, API calls return 402 and the workspace prompts you to upgrade. See Pricing for the current tiers.

Transcripts Core

Paste a link or upload a file and get a clean, timestamped transcript — then generate summaries, chapters, translations, or an MP3, or ask the chat questions about what was said.

Transcripts is the workspace the rest of the product line is built on. It fetches the caption track when a video has one and runs speech recognition when it doesn't. Either way it restores punctuation, capitalization, and paragraphs, so what you get reads like a document instead of a caption dump. Every line keeps its timestamp and links back to that moment in the source.

Once the transcript exists, one click generates a summary, chapter markers, key quotes, a study guide, a quiz, flash cards, highlights, or a translation into 100+ languages. The grounded chat answers questions using only the transcript and cites timestamps, which is the fastest way to find what someone said in an hour of video.

How it works

  1. Paste a link or upload a file. YouTube, Vimeo, TikTok, Twitch, Rumble, a podcast RSS feed, or your own audio or video at /upload.
  2. Kinode fetches or transcribes. Caption tracks are cleaned up with AI; sources without captions go through speech-to-text. A progress indicator shows the job running.
  3. Read, search, and generate. The transcript arrives timestamped and readable. Click any insight, or ask the chat a question.
  4. Export or automate. Copy it, download it, convert it to MP3, or do all of it through the REST API from n8n, Zapier, Make, or your own code.

The insights you can generate

InsightWhat you get
SummaryA tight paragraph-level summary of the whole piece.
ChaptersTimestamped chapter markers you can paste straight into YouTube or a podcast host.
Key quotesThe most quotable lines, verbatim, with timestamps.
Study guide / Quiz / Flash cardsLearning material built from the transcript (see Study for kits of many videos).
Suggested clipsFive clippable moments with hooks and captions (see Clips).
TranslationThe transcript rendered into 100+ languages.
MP3Any transcribed video or episode converted to audio for offline listening.

How-to: transcribe a YouTube video

  1. Open the workspace at /app.
  2. Paste the URL into the box and submit. A playlist or channel URL expands into its videos.
  3. Wait for the pipeline. A cached video appears instantly; a new one shows a progress state while it fetches or transcribes.
  4. Generate what you need. Click Summary, Chapters, or any other insight; open the chat to ask questions.

What people use it for

  • Research without the watching — skim an hour-long lecture as text in minutes, then pull quotes with timestamps.
  • Content repurposing — write the blog post, newsletter, or social copy from an accurate transcript.
  • Accessibility & compliance — give every published video a readable, searchable text version.
  • Competitive intelligence — transcribe competitor launches and earnings interviews, then search what was said.
A video is fetched once

After the first fetch, every repeat visit, insight, and API call against that video is free and instant. If you re-run the same channels often, this matters a lot for cost.

Common issues

The transcript looks rough or has errors

The source had no captions

Speech-to-text is used when a video ships no caption track. Accuracy is high but not perfect on heavy accents, crosstalk, or poor audio. The caption-track path (when one exists) is cleaner because it starts from the creator's own text.

Music or noise stretches

Long musical intros and ambient noise can produce filler lines. Skip to the spoken section using the timestamps.

A link won't transcribe

The video is private, age-gated, or region-locked

Kinode can only fetch what's publicly reachable. Download the file and use Upload instead.

The URL is a channel or playlist

That's supported — it expands into videos — but a very large channel returns many items. Narrow to the specific video if you only want one.

Quota is exhausted

New (uncached) fetches need quota. If you've hit your plan limit the workspace says so; upgrade at Pricing or wait for the monthly reset.

Show Notes Automated

Connect your podcast RSS feed once. From then on, every new episode gets show notes, chapter markers, pull quotes, and a newsletter draft on its own.

Most podcasters put off show notes until the last minute, because writing them is tedious and the episode is already done. Show Notes does the writing for you. Connect your RSS feed once, and each new episode is transcribed and turned into publication-ready notes: a summary written for a listener deciding whether to press play, timestamped chapters, verbatim pull quotes, and a list of links and mentions taken from the audio itself.

Each episode also gets a newsletter draft with a subject line, a hook, and a call to action, so the episode email becomes a quick review instead of an hour of writing. Notes appear in your dashboard as soon as processing finishes. Paste them into your podcast host, or pull them through the API.

How it works

  1. Connect your feed. Paste your podcast's RSS URL once at /shownotes. The newest episode processes immediately so you can see the output.
  2. Kinode watches the feed. The feed is checked every 30 minutes. New episodes get transcribed whether or not captions exist anywhere.
  3. Notes generate themselves. Summary, chapters, quotes, links, and a newsletter draft appear for each episode.
  4. Publish anywhere. Copy from the dashboard into your host, or pull via API into your publishing pipeline.

How-to: connect a feed

  1. Find your RSS URL. Every podcast host exposes one (often labeled "RSS feed" in your show settings). It usually ends in .xml or /feed.
  2. Paste it into Show Notes and save. The latest episode is fetched and processed right away as a preview of the output.
  3. Leave it running. Kinode re-checks every 30 minutes and processes each new episode automatically.
Works even with no captions

Podcasts rarely ship captions, so Show Notes transcribes the audio directly. You don't need to provide a transcript — Kinode makes one.

What people use it for

  • Solo podcasters — ship notes and the episode email in minutes instead of losing an evening per episode.
  • Podcast networks — keep note quality consistent across every show without hiring writers per feed.
  • Producers & VAs — have client-ready notes waiting when an episode drops instead of the day after.
  • SEO for episodes — chapters and summaries give search engines real text to index for every release.

Common issues

New episodes aren't appearing

The feed URL was wrong

Paste the RSS/XML feed, not your show's web page. If your host has both a "website" and an "RSS feed" link, you want the second one.

The episode just published

The feed is polled every 30 minutes, and transcription takes a few minutes after that. A brand-new drop can take up to about half an hour to appear.

The plan doesn't include Show Notes

Feed-watching products need a paid plan. Check your tier at Pricing.

Study Teams-ready

Bundle lectures, tutorials, or a whole playlist into a study kit and get one merged study guide, Anki-ready flash cards, and quizzes — all built from the actual transcripts.

Video is how a lot of people learn now, but you can't skim it and you can't test yourself on it. Study fixes both. Paste up to 12 lecture links into a kit, and Kinode merges their transcripts into one body of material: a study guide with key concepts and review questions, Anki-style flash cards, and a multiple-choice quiz where every answer traces back to something the lecturer said.

The built-in review mode schedules your flash cards with spaced repetition. Cards you miss come back sooner; cards you know come back later. A tutor chat answers questions using only the kit's transcripts and cites timestamps, so it can't feed you an answer that wasn't in the lecture.

For teams, an org owner can assign a kit to the whole organization and watch completion: who opened the guide, who took the quiz, and what they scored. Quizzes are graded server-side and members never see the answer key, so Study holds up for real onboarding and training.

How it works

  1. Build a kit. Name it and paste video links, one per line (up to 12). Kinode fetches or transcribes each source.
  2. Generate the material. One click each for the merged study guide, the flash-card deck, and the quiz.
  3. Study it. Read the guide, drill cards in spaced-repetition review, take the quiz, or ask the tutor.
  4. Assign it (teams). Org owners assign kits to members and track guide views and quiz scores per person.
  5. Take it with you. Export cards as CSV for Anki, or the whole kit as Markdown for your notes app.

What people use it for

  • Exam prep — turn a semester's lecture playlist into one guide, one deck, and a self-test.
  • Employee onboarding — assign training videos as a kit and see who finished and how they scored, without an LMS.
  • Certification study — merge scattered tutorials into a single syllabus with drillable cards.
  • Journal clubs & bootcamps — everyone gets the same guide and quiz from the same material.
Everything comes from the lectures

The guide, cards, and quiz are built strictly from what was said in the source videos. If the lecturer didn't cover it, it isn't in the kit — which is exactly what you want for faithful study material.

Common issues

A kit won't take another video

Twelve is the ceiling

A kit holds up to 12 sources so the merged guide stays coherent. For a larger course, split it into themed kits (e.g. one per module).

Quiz scores aren't showing for my team

The kit wasn't assigned

Tracking only appears for kits an org owner has assigned to members. Assign it from the kit's page, then completion and scores populate as members take it.

Members opened it signed out

Scores are recorded against an account. Members need to be signed in to the org for their results to count.

Pages SEO

Publish every transcript as a hosted public page with the full text, chapters, and summary, so your back catalog can rank for the words spoken in it.

A creator's back catalog is full of long-tail search queries that were spoken out loud and never written down, which means Google can't see them. The big podcasts solve this with transcript pages on their websites. Almost nobody else does, because building and maintaining a transcript site is real work. Kinode already produces the transcript, the chapters, and the summary — Pages just publishes them.

Create a site, paste a video link, and the page is live: full transcript, chapters, summary, embedded player, canonical URL, and Article plus VideoObject structured data, all server-rendered so crawlers see everything. The site index lists every page you've published, and every page links back to it.

How it works

  1. Create a site. Name it and pick a slug at /pages. That becomes your public namespace at /p/your-site.
  2. Publish transcripts. Paste a video URL. Videos you've already transcribed publish instantly; new ones get fetched first.
  3. Enrich (optional). Generate a summary and chapters in the workspace and the public page picks them up on its own.
  4. Get indexed. Submit your site index to Search Console once. New pages are linked from it automatically.

What makes the pages rank

FeatureWhy it helps
Server-rendered HTMLCrawlers see the full transcript text without running JavaScript.
Canonical URLsEach page has one authoritative address, avoiding duplicate-content dilution.
Article + VideoObject structured dataRich-result eligibility and clear signals about what the page is.
Embedded player (YouTube)Visitors read and watch in one place, improving engagement signals.
Linked site indexInternal linking spreads crawl equity across your catalog.
One namespace per project

Run multiple sites — one per show or channel — each with its own public index at /p/<slug>. Agencies run one per client and hand over the traffic report.

What people use it for

  • Podcasters — every episode gets a page that can rank for the topics discussed in it.
  • YouTubers — years of videos become crawlable text without touching your own website.
  • Course creators — public lesson transcripts pull search traffic toward your paid material.
  • Agencies — run a transcript site per client show and hand them the search-traffic report.

Common issues

Pages aren't showing up in Google

Indexing takes time

New pages have to be crawled before they rank. Submit your site index (/p/your-site) to Google Search Console once to speed discovery; new pages are linked from that index automatically.

The video isn't published yet

Publishing a page requires the transcript to exist. If you pasted a brand-new URL, it's fetched first — the page goes live once transcription finishes.

Monitor Automated

Watch the channels and feeds you choose, and let discovery sweep the rest of YouTube for new videos mentioning your keywords. Everything gets transcribed and searched inside the spoken words — and every hit comes with the exact quote and a link to the moment.

Media monitoring tools read titles, descriptions, and comments. But the mention that matters is usually in the audio: a founder name-dropped in minute 43 of a podcast, or your product panned halfway through a review video. Text-only monitors can't see any of that. Monitor transcribes the media itself and searches your keywords inside the spoken words.

You choose channels and podcast feeds to watch, and discovery sweeps the rest of YouTube for fresh uploads matching your terms. New videos and episodes are usually transcribed within about 30 minutes of publishing. Every hit records the verbatim quote, the timestamp, and a link that jumps to the moment, so what you forward to your team is the sentence itself rather than a guess.

How it works

  1. Define keywords. Brand names, product names, executives, competitors, ticker symbols — anything worth catching.
  2. Pick sources + discovery. Watch specific channels and feeds, and let discovery scan new YouTube uploads platform-wide.
  3. Kinode transcribes new releases. Roughly every 30 minutes, new videos and episodes are fetched and transcribed, with or without captions.
  4. Get hits with proof. Each mention carries the quote, the timestamp, and a jump link. Review them in the dashboard at /monitor as they land.

What people use it for

  • Brand & PR teams — find out a YouTuber mentioned you hours after upload, with the exact quote ready to share.
  • Investor relations — catch your ticker discussed across finance podcasts as episodes drop.
  • Competitive tracking — watch competitors' channels and read how they position against you, verbatim.
  • Crisis monitoring — during an incident, follow the spoken coverage as well as the written headlines.
Discovery vs. watched sources

Watched channels and feeds are checked exhaustively. Discovery is a broader sweep of new YouTube uploads matching your keywords — useful when your company is in the news and you can't predict who'll cover it.

Common issues

Too many irrelevant hits

The keyword is ambiguous

A short or common term (a one-word brand that's also an ordinary word) matches unrelated speech. Use the fuller name, or a distinctive phrase, to cut noise.

A mention you know about didn't show

It's outside your sources and discovery window

Discovery covers new uploads. A mention in an old video, or on a platform other than the ones watched, may not be swept. Add the specific channel or feed as a watched source.

The wording differed

Matching is against the spoken words. If the speaker used a nickname or abbreviation, add it as an additional keyword.

Digest Automated

A daily or weekly briefing of everything new from your watched channels and connected feeds. Each item is transcribed, summarized, and linked — and the briefing lands in Discord, a webhook, or your dashboard.

Keeping up with a beat — an industry, a set of competitors, a roster of creators — normally means watching hours of content or accepting that you'll miss things. Digest removes that tradeoff. Point Monitor at the channels and Show Notes at the feeds you care about, pick a time, and every day or week Kinode delivers a briefing of what was said, summarized to the paragraph, with links to each source.

Because Digest rides on the transcript cache, an item summarized once is never paid for twice. And any video in your digest is one click from its full transcript, chapters, and chat in the workspace.

How it works

  1. Watch your sources. Add channels in Monitor and feeds in Show Notes. Digest uses both.
  2. Set the schedule. Daily or weekly, pick the hour, and optionally add a Discord or JSON webhook at /digest.
  3. Kinode writes the edition. New items are transcribed, summarized, and composed into one briefing.
  4. Read it anywhere. In Discord, in your own tooling via webhook, or on the Digest page.

Delivery options

ChannelSetup
DashboardNothing to configure — every edition is saved on the Digest page.
Discord webhookPaste a channel webhook URL; each edition posts there on schedule.
Generic JSON webhookKinode POSTs the edition as JSON to your endpoint for custom routing.
Only what's new

Each edition covers what appeared since the last one. Nothing repeats, so a daily digest never re-summarizes yesterday's videos.

What people use it for

  • Competitive intel — every competitor podcast and channel, summarized into one morning read.
  • Creator research — keep up with your niche without watching your niche.
  • Team briefings — pipe the digest into a Discord channel the whole team reads.
  • Investor monitoring — a weekly roll-up of every earnings interview and finance show you follow.

Common issues

The digest arrived empty

Nothing new published in the window

Digest only includes items that appeared since the last edition. A quiet day produces a short or empty briefing — that's expected.

No sources are connected

Digest draws from your Monitor channels and Show Notes feeds. If neither has sources, there's nothing to summarize. Add them first.

The Discord webhook isn't posting

The webhook URL is wrong or revoked

Re-copy the webhook URL from the Discord channel's Integrations settings. Deleting and recreating the webhook invalidates the old URL.

Meetings Live

Host live video meetings in the browser — scheduled or instant, open or passcode-protected. Each attendee's own microphone transcribes their words under their name, so the record of who said what is accurate.

Most meeting-AI tools record one audio stream and use AI to guess where one voice ends and the next begins. That guessing breaks down when people sound alike or talk over each other. Meetings avoids the problem: every attendee's own microphone transcribes their own words under their own name, live, seconds behind speech. Nobody's identity is inferred, so the transcript can serve as the actual record of the meeting.

Meetings run in the browser with multi-party video and screen share. Schedule one with a link and waiting room, use your permanent personal room, or start one on the spot. Every line carries wall-clock time. When it ends, formal minutes with attendees, decisions, and action items write themselves, and uploaded recordings can be exported with names and timestamps as a subtitle track or burned into the video.

How it works

  1. Start or schedule. Instant meeting, personal room, or scheduled link with an optional passcode and waiting room, from /meetings.
  2. Everyone talks, everyone transcribes. Each participant's mic feeds their own live transcript, line by line, under their name.
  3. End the meeting. Minutes, a summary, and the speaker-labeled transcript are ready when you hang up.
  4. Export the record. SRT, TXT, Markdown, Word, or the video itself with names and clock time embedded.

Meeting types & access

TypeWho can join
Open linkAnyone with the link. Best for public or low-stakes calls.
PasscodeAnyone with the link and a generated passcode.
Sign-in requiredOnly Kinode-authenticated users you admit.
Waiting roomAttendees wait until the host admits them, on any of the above.

Exports

After a meeting you can export the record as minutes (attendees, decisions, action items), a summary, or the raw speaker-labeled transcript in SRT, TXT, Markdown, or Word. For uploaded recordings, Kinode can embed the speaker names and wall-clock time as a subtitle track, or burn them directly into the picture. To seal the record cryptographically, see Certify.

Rename a speaker once

Correct a speaker's name a single time and every export — minutes, SRT, Word, burned-in video — follows the change.

What people use it for

  • Client calls — send minutes with decisions and action items while the meeting is still fresh.
  • Interviews & research — verbatim, speaker-accurate quotes with wall-clock timestamps you can cite.
  • Depositions & disputes — who said what is captured per speaker at the moment of speech, not reconstructed later.
  • Distributed teams — everyone who skipped the call can read the record in two minutes.

Common issues

My words aren't being transcribed

The browser didn't get microphone permission

Per-speaker transcription needs each attendee's mic. Allow microphone access when the browser prompts; if you dismissed it, re-enable it in the site permissions and rejoin.

The wrong input device is selected

If you have multiple microphones, pick the active one in the meeting's device settings.

An attendee can't join

The meeting requires sign-in or a passcode

Open-link meetings admit anyone; passcode and sign-in-required meetings don't. Share the passcode, or admit the person from the waiting room.

They're on an unsupported/old browser

Meetings uses in-browser video (WebRTC). Use a current Chrome, Edge, Firefox, or Safari.

Certify Signed

One click turns any meeting transcript into a digitally signed PDF — speaker-attributed and timestamped, sealed with an X.509 signature from the Kinode Signing Authority. If anyone edits the file afterwards, verification fails.

A transcript in a Word file can be edited by anyone who opens it, which limits what it can prove. Certify renders the meeting's speaker-attributed, timestamped record to PDF and signs it cryptographically, so the document itself shows whether it has been altered since certification.

The two halves matter together. Meetings captures attribution live, with each speaker transcribed from their own microphone, and the Signing Authority (the same X.509 infrastructure behind PDF Studio) seals the export. That gives disputes, HR records, client sign-offs, and compliance archives a record that holds up on its own.

How it works

  1. Hold the meeting. Live in Meetings, or upload a recording and generate the speaker map.
  2. Click "Certified PDF". The transcript renders to a clean PDF and is signed by the Kinode Signing Authority.
  3. Store it anywhere. The proof travels inside the file — nobody needs a Kinode account to verify it.
  4. Verify on demand. The verifier reports the signer, the timestamp, and whether the bytes are intact.

Verifying a certified record

Drop the file into Kinode's verifier at /pdf/sign, or open it in Adobe Acrobat with the Kinode root CA installed. Either way you get an instant verdict: who signed, when, and whether a single byte has changed. Every page also carries the certification watermark and the signing timestamp.

Why it's tamper-evident

The signature covers the exact bytes of the PDF. Change one character — a name, a number, a date — and the cryptographic check fails. There's no way to alter the content and keep the signature valid.

What people use it for

  • Disputes & legal — a speaker-attributed record whose integrity the other side can check for themselves.
  • HR & compliance — disciplinary meetings and policy briefings archived as sealed records.
  • Client sign-offs — certify the call where scope was agreed, before memories diverge.
  • Board minutes — the canonical record of what was decided, sealed the day it happened.

Common issues

Verification says "altered" on a file I didn't edit

Something re-saved the PDF

Some tools rewrite a PDF when you "save" or "print to PDF," which changes the bytes and breaks the signature. Keep and share the original signed file; don't re-export it.

It was opened and saved by another editor

Even adding an annotation and saving changes the file. Verify the untouched original; annotate a copy if you need to mark it up.

Adobe shows the signature as unknown/not trusted

The Kinode root CA isn't installed

The signature is valid; Adobe just doesn't recognize the issuer yet. Install the Kinode CA certificate (see PDF → the CA you can install) and it validates cleanly.

Support Bot Discord

A support desk that lives where your customers already are. AI answers the common questions, real issues become tickets, identity gets verified by customer ID and PIN, and hard problems go to your team.

If your customers are already in your Discord, that's where your support desk should be. Support Bot answers common questions from your knowledge base, opens private ticket threads for anything real, and verifies who it's talking to by customer ID and PIN, so account-specific help doesn't leak across users.

Everything mirrors to a web dashboard: ticket history, statuses, priorities, and the full conversation. When the AI reaches the edge of what it knows, it escalates to your team rather than improvising, and the handoff carries the whole thread so nobody has to ask the customer to repeat themselves.

How it works

  1. Install the bot. Invite it to your server and point it at your docs and FAQ content, from /supportbot.
  2. AI handles the front line. Common questions get answered in-channel or in DMs, with sources.
  3. Real issues become tickets. Private ticket channels with status and priority, mirrored to the web dashboard.
  4. Humans take the hard ones. Escalations arrive with full context; your team answers from Discord or the dashboard.

Identity verification

For account-specific help, the bot links a Discord user to a real account via customer ID and PIN. This keeps one customer from seeing another's account details, so billing and order questions can be answered safely in a shared community.

What people use it for

  • SaaS communities — handle the how-do-I questions automatically and keep engineers on real bugs.
  • Game studios — resolve account and purchase issues with PIN-verified identity instead of trusting DMs.
  • Creator memberships — give paying members a support lane that doesn't depend on you reading every channel.
  • Indie products — a one-person company gets a 24/7 first-response desk.

Common issues

The bot connects but doesn't read messages

Message Content Intent is off

In the Discord developer portal, under Bot → Privileged Gateway Intents, enable Message Content Intent. Without it the bot sees empty message bodies.

Slash commands don't appear

Global registration is still propagating

Commands registered globally can take up to an hour to appear. Setting a specific guild (server) ID registers them immediately for that server.

The AI gives weak or wrong answers

The knowledge base is thin

Answers are grounded in the docs and FAQ you point it at. Add or improve that content; the bot escalates rather than guessing when it's out of its depth.

PDF Studio In-browser

A complete PDF editor that runs entirely in your browser. Edit the text and images already in a PDF, fill forms, rearrange pages, password-protect or unlock files, and sign documents with real X.509 digital signatures — with nothing uploaded for the editing itself.

Most PDF editors ask you to upload the document first, which is a problem when the document is a contract, a medical record, or a term sheet. PDF Studio does its editing entirely in your browser: text edits, image edits, freehand drawing, highlights, whiteout, page operations (rotate, reorder, delete, merge, split), and form filling all happen client-side, and the file never leaves your machine. Open it at /pdf.

Two operations do use a server, because they need real cryptography: digital signing/verification (the Kinode Signing Authority) and password protection/removal (industry-standard AES encryption). Both are described below, and both are explicit actions you choose.

Opening a document

Drag a PDF onto the editor, or use the file picker. The document renders locally and you're editing immediately. If the file is encrypted, PDF Studio handles it automatically — see Protect & unlock.

The toolset

ToolWhat it does
SelectClick any object you've added to move, resize, or delete it.
Edit contentClick text or images already in the PDF to change them (see below).
TextClick anywhere to drop a new text box; choose font, size, weight, and color.
Draw / HighlightFreehand ink and translucent highlighting.
ShapesRectangle, ellipse, line, and arrow with stroke, fill, and opacity.
WhiteoutCover content with a solid block (background-matched when used via Edit content).
ImagePlace a PNG or JPG anywhere on the page.
FormsAppears when the PDF has an interactive form — fill its fields.

Editing existing text

The Edit content tool is what makes this a true editor rather than an annotator. Switch to it and every line of text and every image already in the PDF gets a hover outline.

  1. Click a line of text. It opens in place, pre-filled with the original words at a matched size, font family (mapped to Helvetica, Times, or Courier), weight, and color sampled from the page.
  2. Retype it. When you click away, the original text is removed from the PDF's content stream — so the old words are gone from search and copy-paste too, not just hidden — and your new text takes its place.
  3. Or clear it to erase the line entirely.
Real removal, with a safety net

Where it can prove the removal is safe, Edit content deletes the original text operators outright. In rare layouts where that isn't provably safe (text sharing one drawing instruction, or shared background artwork), it falls back to covering the line with a background-matched patch so the result still looks right.

Editing existing images

With Edit content active, click an image that's part of the PDF to select it:

  • Move — drag it; a live preview follows the cursor.
  • Resize — drag a corner handle.
  • Replace — swap in a new PNG/JPG at the same spot.
  • Delete — remove it from the page (press Delete, or use the toolbar action).

Moves and resizes rewrite the image's placement inside the PDF without re-encoding the picture, so there's no quality loss. Page rotation and unusual page origins are handled correctly.

Text wins overlaps

When a thin rule or a background box sits on top of a line of text, the text target takes your click, so you can always edit the words. Images remain clickable anywhere text isn't.

Filling forms

When a PDF contains an interactive form (an AcroForm), a Forms tool appears and the fields become live inputs — text boxes, checkboxes, radio buttons, and dropdowns — laid directly over the page. Type into them, and the values are written into the PDF when you download. Field values are part of the undo history, so you can undo/redo edits like anything else.

  1. Switch to the Forms tool. The fillable fields highlight.
  2. Fill them in. Text, checkbox, radio, and dropdown fields all work.
  3. Download. The values are baked into the form fields of the exported PDF, visible in any viewer.

Page operations

OperationHow
Rotate / reorder / deleteFrom the page thumbnails rail.
Insert blank pageFrom the thumbnail menu.
MergeAppend every page of another PDF.
SplitExtract a page range (e.g. 1-3, 5) as a new PDF.
Export PNGRender the current page to a PNG image.

Watermarks

Add a diagonal text watermark (a preset like DRAFT or your own text), set its opacity and color, and choose whether it applies to all pages or a range. The watermark is drawn into the PDF when you download, split, or export a PNG.

Protect & unlock (passwords)

The Secure button opens a panel with two modes. Both use industry-standard AES-256 encryption on the server (the file is processed and returned; it isn't stored).

Protect a PDF

  1. Open Secure → Protect.
  2. Set a password. A password to open encrypts the whole document; a separate owner password (optional) controls permissions without blocking opening.
  3. Choose what's allowed — printing, editing, copying text, annotations, form filling, and page assembly. Anything you leave unchecked is blocked in the protected file.
  4. Protect & download. Kinode flattens your edits, applies any watermark, encrypts with AES-256, and downloads the protected copy.

Remove restrictions

Secure → Remove restrictions strips a document's security so you can edit and print it.

  • Permission-only locks (a file that opens fine but blocks editing or printing) are removed with no password needed.
  • Open-password files need that password entered once; then the restriction is removed.
What "remove restrictions" can and can't do

If a PDF only has an owner/permissions password — the common "you can view but not edit or print" case — Kinode removes it without any password, because the file was never actually locked from opening. If a PDF has a genuine open password, you must supply it; there's no way to bypass real encryption, and Kinode won't try to.

When you open an encrypted file in the editor, this runs automatically: permission-only files unlock silently, and open-password files prompt you once for the password before editing begins.

Sign & certify (digital signatures)

Signing goes further than a drawn squiggle. Open /pdf/sign:

  1. Add a visual signature. Draw one, type one, or upload an image, and place it on any page.
  2. Apply a cryptographic signature. The Kinode Signing Authority embeds an X.509 digital signature (PKCS#7 — the same standard Adobe uses) that proves who signed and when.
  3. Verify anytime. Drop any signed PDF into the verifier for an instant verdict: intact or altered, signed by whom, at what moment. Change one character of a signed file and verification fails.

The CA you can install

Download the Kinode CA certificate once and install it in your OS or in Adobe Acrobat. From then on, every document your organization certifies validates cleanly, and Adobe shows a trusted green check instead of an "unknown issuer" warning.

What people use it for

  • Contracts & NDAs — redline, sign, and return agreements without routing them through a third-party server.
  • Sensitive documents — medical, legal, and financial files get edited where they sit: on your machine.
  • Internal sign-offs — certify approvals with signatures your whole org can verify against one CA cert.
  • Form finishing — fill, whiteout, stamp, protect, and flatten forms in minutes.

Common issues

Clicking text doesn't open an editor

Wrong tool

Editing existing page text needs the Edit content tool (not Select or Text). Switch to it, then click the line — it highlights on hover when the tool is active.

The text is actually an image

Scanned documents have no real text — the "words" are pixels in an image. Edit content can move or replace that image, but there's no text to retype. (OCR of scans isn't part of the in-browser editor.)

An image can be deleted but not moved

It's shared or repeated artwork

An image drawn several times on a page, or shared across pages, is protected from moving to avoid corrupting the other uses. It's still selectable, and the toolbar explains why with a hint. You can delete or replace it.

"This PDF is password-protected"

It has an open password

Enter the password when prompted and editing continues. If you don't have it, the file can't be opened — that's real encryption, not a Kinode limitation.

It only blocks editing/printing

Permission-only locks are removed automatically on open (no password needed). If you're doing it manually, use Secure → Remove restrictions.

A signed file verifies as altered

It was re-saved after signing

Any tool that rewrites the bytes — "print to PDF," another editor's save, even adding one annotation — breaks the signature. Share the original signed file; annotate a copy if needed.

Adobe shows "unknown issuer"

The signature is valid; install the Kinode CA certificate so Adobe trusts the issuer.

A very large PDF is slow or won't open

In-browser editing has a size ceiling

Editing happens in your browser's memory, so there's a practical limit (around 200 MB) for the editor and a 100 MB limit for the encrypt/unlock operations. Split a huge file first, or reduce its size.

Clips Preview

Kinode reads the transcript and finds the five most clippable moments in any video — with timestamps, verbatim hooks, ready-to-post captions, and hashtags. Rendered vertical clips with burned captions are coming next.

Turning long-form video into shorts usually means scrubbing through footage or paying an editor. The slow part is finding the moments, and finding is something a transcript is good for. Kinode already has the transcript, so it can score every stretch of a video for hook strength and self-containment, then hand you the five best moments with timestamps.

Moment detection is live now as the Suggested clips insight in the workspace. Paste any video, click once, and get five candidate clips with verbatim opening hooks, notes on why each works, post-ready captions, and hashtags. The full Clips product adds rendering: 9:16 crops with word-synced burned captions. It's in development, and waitlist members get first access.

How it works today

  1. Transcribe any video. The same paste-a-link workflow as everything else.
  2. Generate Suggested clips. One click. Five moments, each 15–60 seconds, spread across the video.
  3. Post. Jump to the timestamp, cut in your editor, and use the caption and hashtags as written.
Rendering is coming

9:16 export with word-synced burned captions and style presets is in development for the full Clips release. Join the waitlist to get it first: support@usekinode.app.

API Developers

Everything the workspace does is a documented REST endpoint: transcripts, bulk fetches, playlist and channel expansion, uploads, AI insights, MP3 export, and grounded chat. Authenticate with a Bearer key, get JSON back.

If you can click it, you can curl it. The API draws from the same plan quota as the workspace, so there's no separate API pricing to reason about, and the transcript cache works for your code the same way it works in the app — a video any Kinode user has ever fetched returns instantly and doesn't bill you.

Authentication

Create a key under Account → API keys and send it as a Bearer token. Keys are scoped, and you can mint and revoke them at any time.

Authorization: Bearer kinode_live_xxxxxxxxxxxxxxxx

Endpoints

Method & pathWhat it does
POST/api/transcriptsFetch transcripts for up to 50 video IDs in one call. Cached fetches are instant.
POST/api/playlistExpand a playlist URL or ID into its video IDs (up to ~250 items).
POST/api/channelsLook up channel metadata and uploads playlists for up to 50 handles. Plus/Pro tiers.
POST/api/uploadStream your own video or audio file; returns a job that resolves to a transcript.
POST/api/mp3/:videoIdConvert any transcribed video to MP3; GET the same path to download.
POST/api/insight/:videoId/:kindGenerate summaries, chapters, key quotes, study guides, quizzes, translations, and more.
POST/api/chatAsk questions grounded in a video's transcript, with optional multi-turn history.
GET/api/transcribe/:videoIdRead a cached transcript and every generated output for a video.
GET/api/jobs/:idPoll job status: pending, running, done, error.

Examples

Bulk transcripts — cURL

curl -X POST https://usekinode.app/api/transcripts \
  -H "Authorization: Bearer $KINODE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"ids": ["dQw4w9WgXcQ", "jNQXAC9IVRw"]}'

Upload your own file — cURL

curl -X POST https://usekinode.app/api/upload \
  -H "Authorization: Bearer $KINODE_KEY" \
  -H "Content-Type: application/octet-stream" \
  -H "X-File-Name: standup-recording.mp4" \
  --data-binary @standup-recording.mp4

Summary — JavaScript

const res = await fetch("https://usekinode.app/api/insight/dQw4w9WgXcQ/summary", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.KINODE_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({}),
});
const { text } = await res.json();   // markdown summary

Grounded chat — Python

import os, requests

r = requests.post(
    "https://usekinode.app/api/chat",
    headers={"Authorization": f"Bearer {os.environ['KINODE_KEY']}"},
    json={"videoId": "dQw4w9WgXcQ", "question": "What are the key takeaways?"},
)
print(r.json()["answer"])

Conventions

  • Errors are JSON: { "error": "message" } with a meaningful HTTP status. 402 means quota or credits are exhausted.
  • Caching is per video: repeat fetches of the same content are instant and never re-billed.
  • Rate limits apply per key; batch endpoints accept up to 50 items per call.
  • Async work (uploads, AI generation) returns a job or a pending list — poll /api/jobs/:id until it resolves.

Common issues

401 Unauthorized

Missing or malformed header

The header must be exactly Authorization: Bearer kinode_live_.... A revoked key also returns 401 — mint a fresh one under Account → API keys.

402 Payment Required

Quota exhausted

The key drew your plan's pool to zero. Upgrade at Pricing or wait for the monthly reset. Cached fetches don't count, so re-requesting the same videos still works.

A response is stuck "pending"

The work is asynchronous

Uploads and AI generation return a job or a pending list. Poll GET /api/jobs/:id until status is done (or error). Don't re-submit — that starts a second job.

Integrations

Because every capability is a REST endpoint with Bearer auth, Kinode drops into no-code automation tools and your own services alike.

ToolTypical flow
n8nNew upload → transcript → summary → post to Slack/Notion, entirely no-code.
Zapier / MakeTrigger on an event, call the Kinode endpoint, route the JSON result onward.
WebhooksDigest and other products POST JSON to your endpoint for custom handling.
Your own codeAny language that can make an HTTP request. Bulk endpoints handle backfills.

Common patterns

  • Product features — add "transcribe this video" to your own app without building a transcription stack.
  • Data pipelines — backfill a channel's entire history into your warehouse with bulk calls and playlist expansion.
  • Automation — new upload → transcript → summary → Slack or Notion, no-code via n8n or Zapier.
  • Research corpora — assemble searchable text datasets from thousands of videos, cached and deduplicated.

See /integrations for connector details and /api for the full endpoint reference.

Plans, quota & billing

One plan powers every product. Your quota is a shared monthly pool of transcription minutes and AI generations, used by the workspace and the API alike.

How quota works

  • Shared pool. The workspace and the API draw from the same quota. There's no separate API bill.
  • Cache-friendly. A source is billed the first time it's fetched. Repeat visits, insights, and API calls against a cached video are free.
  • Resets monthly. Quota refills on your billing date. Hitting the limit returns 402 from the API and an upgrade prompt in the app.

Tiers

The free tier covers the core workspace for trying things out. Paid tiers raise quota and unlock the scheduled-watching products (Show Notes, Monitor, Digest) and the API. Some capabilities (channel lookups, higher batch sizes, webhooks) are Plus/Pro. See Pricing for current tiers and limits, and manage billing at Account.

Refunds

See the Refund Policy for details on eligibility and how to request one.

Common issues

I hit my limit mid-month

New fetches consumed the pool

Only uncached fetches and AI generations use quota. Upgrade at Pricing for a bigger pool, or lean on already-cached content until the reset.

Troubleshooting

Cross-cutting problems and their causes. Each product section above also has its own "Common issues" — the search box at the top left finds any of them by symptom.

A link won't transcribe

The source isn't publicly reachable

Private, age-gated, region-locked, or deleted videos can't be fetched. Download the file and use Upload instead.

Quota is exhausted

New (uncached) fetches need quota. If you've hit your plan limit, upgrade at Pricing or wait for the reset. Already-cached videos still work.

It's a whole channel or playlist

Those expand into many videos, which is intended — but paste the single video URL if you only want one.

The transcript has errors

No captions existed

Speech-to-text is used when a source ships no caption track. Accuracy is high but drops on heavy accents, crosstalk, and poor audio.

Music or noise sections

Long musical or ambient stretches can add filler lines. Use the timestamps to jump to the spoken content.

A watched feed or channel isn't updating

The window hasn't elapsed

Feeds and channels are polled about every 30 minutes, and transcription adds a few minutes. A brand-new item can take up to ~30 minutes to appear in Show Notes, Monitor, or Digest.

The wrong URL was added

Use the RSS/XML feed for podcasts, not the show's web page; use the channel URL for YouTube, not a single video.

The plan doesn't include it

Scheduled-watching products (Show Notes, Monitor, Digest) need a paid plan.

A Discord or JSON webhook isn't firing

The URL was revoked

Deleting and recreating a Discord webhook changes its URL. Re-copy it from the channel's Integrations settings.

Nothing new to send

Digest only posts when there are new items since the last edition. A quiet period produces nothing — that's expected.

I can't edit the text in a PDF

Wrong tool

Existing page text needs the Edit content tool, not Select or Text. See Editing existing text.

The page is a scanned image

Scanned "text" is pixels, not characters. You can replace the image, but there's no text to retype in the in-browser editor.

The file is encrypted

Permission-only locks are removed automatically on open; open-password files prompt for the password. See Protect & unlock.

A signed PDF verifies as altered or untrusted

It was re-saved after signing

Any tool that rewrites the bytes breaks the signature. Share the original signed file; annotate a copy instead.

Adobe shows "unknown issuer"

Install the Kinode CA certificate so Adobe trusts the issuer. The signature itself is valid.

A meeting isn't transcribing a speaker

Microphone permission was denied

Per-speaker transcription needs each attendee's mic. Allow microphone access and rejoin; check the selected input device.

Everything says I'm out of quota

The monthly pool is spent

Workspace and API share one pool. Upgrade at Pricing, or use cached content (which doesn't count) until the reset. The API returns 402 in this state.

API calls return 401

Header or key problem

Send Authorization: Bearer kinode_live_... exactly. Revoked keys 401 — mint a new one under Account → API keys.

Glossary

Transcript
The timestamped, punctuated text of a source. The thing every other product is built on.
Source
Anything Kinode can transcribe: a YouTube/Vimeo/TikTok/Twitch/Rumble video, a podcast feed or episode, or an uploaded file.
Insight
Something generated from a transcript with one click: summary, chapters, quotes, study guide, quiz, flash cards, highlights, suggested clips, or translation.
Grounded chat
A chat that answers only from a transcript and cites timestamps, so it can't invent facts the source didn't contain.
Cache
The global store of transcripts. A source is fetched and billed once; every later use is instant and free.
Quota
Your plan's monthly pool of transcription and AI generation, shared by the workspace and the API.
Kit
A Study bundle of up to 12 sources merged into one guide, flash-card deck, and quiz.
Discovery
Monitor's platform-wide sweep of new YouTube uploads matching your keywords, beyond the sources you explicitly watch.
Managed / owner (org)
An organization shares one plan; the owner can invite members and assign Study kits.
AcroForm
A PDF's interactive form. When present, PDF Studio shows a Forms tool with live fields.
Flatten
Baking overlays (text, drawings, form values) into the PDF's pages so they're part of the file, done on download.
Open password
A password required to open (decrypt) a PDF. Real encryption — it can't be bypassed without the password.
Owner / permissions password
A password that restricts editing/printing but not opening. Kinode can remove these without a password.
X.509 / PKCS#7
The certificate and signature standards behind Kinode's digital signatures — the same ones Adobe uses.
Signing Authority
Kinode's certificate authority that issues the signatures used by PDF Studio and Certify.
Certify
Rendering a meeting transcript to a signed, tamper-evident PDF.
Bearer key
The API token (kinode_live_...) sent in the Authorization header. Minted and revoked from your account.

Questions people ask

Do my files get uploaded when I edit a PDF?

No. PDF editing — text, images, drawing, forms, page operations — runs entirely in your browser. Only two explicit actions use a server: applying/verifying a digital signature, and password-protecting or unlocking a file. Both are things you choose, and the file isn't stored.

How does billing work across all these products?

One plan, one quota. Transcription minutes and AI generations come from a single monthly pool shared by the workspace and the API. Because transcripts are cached, a source is only billed the first time it's fetched.

What does "cached" mean for me?

A video any Kinode user has ever transcribed returns instantly and doesn't use your quota. Re-opening a transcript, generating another insight from it, or calling the API against it again is free.

Can Kinode remove a PDF password I don't know?

Only if it's an owner/permissions password — the "view but not edit or print" kind — because that file was never locked from opening. A real open password can't be bypassed; you have to supply it.

How accurate is the per-speaker transcription in Meetings?

Each attendee's own microphone transcribes their own words under their own name, so attribution isn't guessed. That's why the transcript can serve as the actual record — and why Certify can seal it as evidence.

Do I need captions on my podcast for Show Notes?

No. Podcasts rarely have captions, so Kinode transcribes the audio directly. Connect the RSS feed and everything else is automatic.

How fast do watched feeds update?

Feeds and channels are checked about every 30 minutes, and transcription adds a few minutes. A brand-new item usually appears within half an hour in Show Notes, Monitor, and Digest.

Is there an API for everything?

Yes — transcripts, bulk fetches, playlist and channel expansion, uploads, every insight, MP3 export, and grounded chat, all with Bearer-key auth. See the API section.

Can a signed document be verified without a Kinode account?

Yes. The proof travels inside the PDF. Anyone can drop it into Kinode's verifier, or open it in Adobe with the Kinode root CA installed, and see the signer, the time, and whether it's intact.

What's coming next?

Clips rendering — 9:16 vertical export with word-synced burned captions — is in development. Moment detection is already live as the Suggested clips insight. Join the waitlist at support@usekinode.app.

Nothing in the documentation matches that. Try a different word, or contact support.