Meeting notes · checked 2026-10-19

Can I vibecode Fireflies.ai?

KINDA· weekend project

What happens after the call fits in one sitting: a recording goes through whisper.cpp, speakers are labelled, a language model writes the summary and action items, and everything lands in a searchable local database. The weekend goes on the bot that joins Zoom, Meet and Teams from your calendar, because a headless browser has to sign in, wait in the lobby and keep up with each platform's changes, and some platforms block it. Skipping the bot and recording your own audio works for one person. Team workspaces, conversation analytics and CRM sync stay behind.

$18/moprice per seat

$216you'd save a year

weekendbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyIntermediate · Some system setup or keys.

TimeA weekend: the pipeline takes a few hours, recording your calls is the rest

Claude Code, Codex or CursorAny of them can run the prompt.
Python 3.12 and ffmpegwhisper.cpp builds a native module; Apple Silicon makes transcription fast.
About 1.5 GB of diskFor the medium whisper model.
A language model endpointAn OpenAI-compatible key, or Ollama if transcripts must stay local.
A Hugging Face tokenOnly for speaker labels: pyannote asks you to accept its model terms first.

Before you buildCheck Fireflies Free first: unlimited transcription with limited AI summaries costs nothing. It stops being enough once your stored meetings pass 400 minutes.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a meeting transcript archive to replace Fireflies. Requirements:

- Python 3.12: FastAPI with HTMX pages and a Typer CLI. `meetnotes serve` runs
http://localhost:4770; `meetnotes add <file>` ingests one recording.
- Input: watch ~/MeetingRecordings/ (`watchdog`) for .mp3, .m4a, .wav, .mp4, .webm and .mkv, and
accept uploads on the web page. `ffmpeg` converts everything to 16 kHz mono WAV first.
- Transcription: whisper.cpp through `pywhispercpp`, model and language from .env (default
medium, auto-detect), with timestamps. Label speakers with `pyannote.audio` when HF_TOKEN is
set; otherwise label every line "Speaker".
- Notes: send the transcript in chunks to an OpenAI-compatible endpoint (LLM_BASE_URL,
LLM_API_KEY, LLM_MODEL in .env) and write a summary, decisions, action items with owner and
due date when they were spoken, and open questions. Save transcript and notes as Markdown in
~/MeetingNotes/YYYY-MM-DD-title.md; the files are the source of truth.
- Storage: sqlite3 with FTS5 and `sqlite-vec` in ./data/meetings.db: meetings, transcript
segments (start, end, speaker, text) and action items with a done flag. `meetnotes reindex`
rebuilds the database from the Markdown files.
- Pages: a meeting list with search that jumps to the timestamp in an HTML5 audio player; an
action-items page with checkboxes; export of one meeting to Markdown or SRT.
- Ask: a box that takes a question across all meetings, retrieves the 8 best segments (FTS5 plus
embeddings from the same endpoint) and answers with citations as meeting and timestamp.
- Single user. No accounts, no telemetry, local-first. Audio never leaves the machine; only
transcript text goes to the LLM endpoint, and the README must say so and show how to point
LLM_BASE_URL at Ollama to keep everything local.
- Out of scope: a bot that joins Zoom, Meet or Teams, calendar integration, CRM, Slack or task
syncing, accounts and sharing.
- README: how to record a call (the platform's own record button, or OBS), the ~1.5 GB
whisper medium download, and the Hugging Face licence step that pyannote needs.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Fireflies.ai.

Add automatic meeting capture to the transcript archive in this folder. Requirements:

- Read an iCal feed (ICS_URL in .env) every 5 minutes and find events in the next 10 minutes that
  have a Zoom, Google Meet or Teams link.
- For each, ask a self-hosted Vexa instance (VEXA_URL and VEXA_API_KEY in .env, following the
  endpoints in Vexa's README) to send a bot, and remember the request so it is never sent twice.
- When the meeting ends, fetch the transcript and feed it through the same notes and indexing
  pipeline as an uploaded file; mark the source as "vexa".
- An opt-out list of event titles or domains the bot must never join.
- Update the README, including a warning that attendees should be told a bot is recording.

Add a weekly digest to the transcript archive in this folder. Requirements:

- A command `meetnotes digest` that collects last week's meetings, their decisions and every
  action item still open, and sends one email through SMTP (`smtplib`; SMTP_HOST, SMTP_USER,
  SMTP_PASSWORD and DIGEST_TO in .env).
- One section per meeting with the title, date and a three-line summary linking to its page.
- A launchd plist and a cron line in the README to run it Monday at 08:00.
- Update the README.

Add task export to the transcript archive in this folder. Requirements:

- A command `meetnotes tasks` that writes open action items to TASKS_FILE (a Markdown file in .env)
  as `- [ ] text (owner, due date) [[meeting]]` lines, and reads ticked boxes back to mark them done.
- Keep a stable id per item in an HTML comment so edits to the text never create duplicates.
- Only items assigned to YOUR_NAME in .env, or all of them with a flag.
- Update the README.

What you lose

A bot that joins Zoom, Google Meet and Teams from your calendar without you doing anything
Shared workspaces, channels, comments and clips for a team
Notes and action items pushed into CRMs, Slack and task apps
Conversation analytics such as talk time, topics and sentiment on Business
Mobile and desktop apps and a Chrome extension for recording without the bot

Why people still pay

The bot means nobody has to remember to record: it appears in every call, and the transcript and summary are waiting in the archive, Slack or the CRM afterwards. Teams also pay for one searchable archive in place of a folder per person. A script you start by hand loses that habit, and the meeting you forgot to record is the one you needed.

Moatintegrations collaboration execution polish

Do the math

Fireflies.ai or your own copy?

Seats

Over

Your usage

Count my time12h ×$/h

Fireflies.ai, 1 yr$216

Your own copy, 1 yr$41running + build session

You keep$175pays for itself in the first month

Estimate · Transcription runs locally for free. Summaries cost a few cents per meeting through an API, or nothing with Ollama. The build is token usage on your agent plan.

Leaving Fireflies.ai

Take your data, then cancel.

1 · Take your data out of Fireflies.ai

1Open Fireflies' guide How to export your Fireflies account data and fill in its data request form.
2Approve the confirmation email titled Action Required - Confirm Your Data Export Request. Nothing is processed until you do; check spam if it doesn't arrive.
3Within about 24 to 48 hours an email brings a download link to a ZIP file.
4The ZIP holds audio as MP3 and video as MP4 where available, transcripts and AI summaries as TXT with speaker identification, attendee details and metadata as JSON.

2 · Cancel Fireflies.ai

1In the web app click Settings, then Account.
2In the Subscriptions section click Cancel next to your active plan.
3Choose a reason in the feedback pop-up and click Continue. AskFred may suggest alternatives: choose Proceed With Cancellation.
4Click Cancel Subscription on the last prompt, or Cancel Team Subscription for a team plan.

Cancelling keeps your data: paid features run to the end of the billing period, then the account moves to Free. Fireflies promises access to meetings only within the Free limits, which include 400 minutes of storage per team, so export first.
The export link works for 7 days; after that you submit a new request.
Fireflies states that it does not offer refunds on subscriptions or AI credit purchases.
Deleting the account is a separate step that permanently removes the workspace and its data. Cancelling the plan does not do that.

Checked 2026-10-19 against Fireflies.ai's docs: 1, 2, 3, 4

The escape hatch

Don't feel like building it? These are already free.

MeetilyOpen sourceMIT desktop notetaker with local Whisper or Parakeet transcription, speaker labels and Ollama summaries; nothing has to leave your machine.★ 32k · active Sep 2026 · one-click install
macOS, WindowsVisit ↗AnarlogOpen sourceMIT privacy-first meeting notepad that works locally; its authors now build a separate product, and this community app stays maintained.★ 9.5k · active Oct 2026 · one-click install
macOSVisit ↗VexaOpen sourceApache-2.0 meeting transcription API with bots that join Google Meet, Teams and Zoom and stream transcripts over WebSocket; run it yourself or use the hosted version.★ 2.9k · active Oct 2026 · docker to self-host
WebVisit ↗

All 5 free alternatives to Fireflies.ai ›

The numbers

Fireflies.ai pricing

Plan	Monthly	Annual, per month	What you get
Free	$0	$0	Unlimited transcription, limited AI summaries, 400 minutes of storage per team, 20 AI credits, recording of Zoom, Meet and Teams, upload of audio and video files, Chrome extension, desktop and mobile apps.
Pro	$18	$10	Unlimited AI summaries, 8,000 minutes of storage per seat, 20 AI credits, video recording, downloads of transcripts, summaries and recordings, action items, unlimited integrations.
Business	$29	$19	Unlimited storage, 30 AI credits, conversation intelligence, team analytics for admins, user groups, multi-language mode, public meeting access.
Enterprise	—	$39	50 AI credits, rules engine, SSO and SCIM, audit logs, HIPAA compliance, private storage, custom data retention, a dedicated account manager.

Free tier: 400 minutes of storage per team, limited AI summaries and 20 AI credits.
Billing: Monthly or annual per seat; every teammate you add to the workspace counts as a seat. Fireflies says it does not refund subscriptions or AI credit purchases. Individual meeting recap links can be shared with non-members at no cost.
Hidden costs: AI credits are a monthly allowance for the AI features, and more can be bought as a separate package under Settings, Billing, AI Credits (annual is about 20% cheaper than monthly). Customising the bot's name is a paid add-on, and Pro's storage caps at 8,000 minutes per seat.
Source: fireflies.ai

Also on the list

Granola$14/mo · Meeting notesYES

Superwhisper$8.49/mo · DictationYES

Descript$24/mo · Audio & videoKINDA

Questions

Fireflies.ai, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

Free has unlimited transcription with limited AI summaries and 400 minutes of storage per team.

Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: A bot that joins Zoom, Google Meet and Teams from your calendar without you doing anything, Shared workspaces, channels, comments and clips for a team.

About $18 a month on the Pro plan per seat, $216 a year (checked 2026-10-19).

A bot that joins Zoom, Google Meet and Teams from your calendar without you doing anything; Shared workspaces, channels, comments and clips for a team; Notes and action items pushed into CRMs, Slack and task apps; Conversation analytics such as talk time, topics and sentiment on Business; Mobile and desktop apps and a Chrome extension for recording without the bot. If any of those matter to you, keep paying.