Fireflies.ai

What it is

Fireflies.ai is a bot-based meeting notetaker. fred@fireflies.ai joins Zoom, Google Meet, and Microsoft Teams calls off your calendar, transcribes in 100+ languages, and writes summaries, action items, and topic trackers. Krish Ramineni and Sam Udotong founded it in 2016; it is headquartered in San Francisco and crossed a $1B valuation in 2025.

It stopped being only a notetaker on 2026-07-21, when Fireflies shipped the AI Sales Suite: Deal Intelligence (at-risk deals, competitive threats, missing stakeholders, each linked to the call evidence behind it), CRM Autofill for Salesforce and HubSpot, Sales Assist live prompting during a call, AI Scorecards against MEDDIC, BANT, SPICED or a custom methodology, and Meeting Prep briefings. Deal Intelligence lands on Business, at $19 a seat. That is the reason the entry exists: functionality the revenue-intelligence platforms sell on a quote-only contract, priced inside a notetaker seat.

Two meters, not one

The seat price is the number buyers quote each other. The credits are the number that decides the bill.

Seats, per user per month (monthly / annual, from Fireflies’ own pricing page):

• Free — $0. Unlimited transcription, limited AI summaries, 400 minutes of storage per team, 20 AI credits.
• Pro — $18 / $10. Unlimited AI summaries, 8,000 minutes of storage per seat, video recording, downloads, Voice Agents, AI Skills, Email Assistant, unlimited integrations, 20 AI credits.
• Business — $29 / $19. Unlimited storage, conversation intelligence, admin team analytics, multi-language mode, user groups, 30 AI credits.
• Enterprise — $39, annual only. Rules engine, SSO and SCIM, audit logs, HIPAA, private storage, custom data retention, super-admin role, 50 AI credits.

Transcription and standard summaries are not metered. AI Skills, Voice Agents, Sales Assist, CRM Autofill, custom summary sections, Magic Soundbites, and the Personal Assistant’s Daily Digest and Meeting Prep all draw AI credits — which is to say every part of the 2026 agentic layer draws credits. Fireflies’ documentation describes the credits bundled with a plan as a one-time grant, not a monthly refill; recurring use runs on a separate AI credits subscription sized between 50 and 10,000 credits per month. Published packs run 50 for $5, 200 for $20, 1,000 for $90, 5,000 for $375, and 10,000 for $600.

Do the arithmetic before signing. Twenty seats on Business annual is $4,560 a year. Whether the agentic features on top cost $240 or $7,200 a year is set by the credit slider, and no seat tier caps it.

Best for

• RevOps leaders who want scorecards, deal risk, and CRM autofill on Salesforce or HubSpot without a revenue-intelligence platform contract, and who will accept metered AI in exchange
• Recruiting and CS teams running high call volume across languages who need one searchable archive with unlimited storage from $19 a seat
• Teams feeding meeting history to an agent over MCP or the GraphQL API, per user rather than org-wide

Alternatives — and when to pick them instead

• Gong — pick it when the forecast is the deliverable. Fireflies scores calls and flags at-risk deals; Gong is the system of record for pipeline inspection and a managed coaching program, and it is the market-share incumbent buyers benchmark against. The trade is a quote-only platform contract against a $19 seat.
• Granola — pick it when a bot in the participant list is the blocker: candidates who decline recording, in-person conversations, client calls where a visible notetaker changes the room. Granola captures from the device at $14 on Business and has no bot to decline.
• Fathom — the fastest-growing entrant in the segment and the pick when a free tier must carry real usage. Unlimited recording and summaries at $0, against Fireflies Free capping team storage at 400 minutes and metering every AI action.
• Otter.ai — pick it when transcription minutes and a searchable archive are the whole requirement and the agentic layer is not. Business is $19.99 a seat annually with unlimited transcription minutes.

Watch-outs

• Credit auto-upgrade is on by default. Run out mid-cycle on a paid plan and Fireflies subscribes you to the next credit tier automatically rather than blocking. Guard: pause auto-upgrade under Settings → Billing → AI Credits before rollout, pilot 30 days with a fixed pack, and read actual consumption off the team credit-usage report before you size the subscription.
• MCP is per-user, with no workspace control. Each person connects Claude or ChatGPT with their own API key, only their own meetings are reachable, and Fireflies states team-level support is still in development. Zero-day retention does not extend to third-party MCP connections — data crossing into an external client is governed by that client’s terms. Guard: if meeting history has to reach a shared agent, do it through the API under one service account, and gate MCP connections in your AI-tool policy instead of assuming an admin toggle exists.
• The API and MCP share one rate limit, and it is tier-dependent. 50 requests a day on Free, 500 a day on Pro, 60 a minute on Business and Enterprise. Guard: an agent reading transcripts on a Pro seat hits 500 a day quickly — price Business at $19 for anything automated, and treat Pro as a human-only tier.
• Governance costs double the seat. SSO, SCIM, audit logs, HIPAA, private storage, and custom retention are all Enterprise at $39, against $19 on Business. Guard: if Security or Legal will require SSO and a retention window, price Enterprise from the start — moving up mid-term is a renegotiation, and Enterprise is annual-only with no monthly escape.
• A bot appears in the participant list, and calls cap at 3 hours. Recording is visible to everyone present, which makes consent your obligation in two-party-consent jurisdictions including California, Illinois, Pennsylvania, Washington, and the EU under GDPR. Video captures at 720p. Guard: put notice in the invite template and the call opener before rollout, and split all-day workshops or panel interviews across sessions rather than losing the tail of a 4-hour recording.