Fireflies.ai

Breadth across classes, and every connector is available on every paid tier rather than gated by plan, so a small team gets the same connector surface as an enterprise one. CRM coverage spans Salesforce, HubSpot, Zoho, Pipedrive and Copper, with structured field writes rather than attached notes.

Conferencing covers Zoom, Google Meet and Microsoft Teams, and capture works through a meeting bot, a Chrome extension or a bot-free desktop app, so a team that will not admit a bot to calls still has a route. A GraphQL API is included on paid plans, and a Perplexity partnership supplies live web search. Admins control which integrations are enabled for the team.

Sourcefireflies.ai/pricing and guide.fireflies.ai admin controlsread 2026-09-04

A rules engine runs automated workflows triggered by meeting events, with named agents executing on top and results landing in systems of record rather than in a summary.

Rules fire on meeting events and route content to channels or user groups; CRM autofill writes structured fields into pipeline tools without manual entry; a Task Feed centralizes action items with ownership and due dates; and Voice Agents, AI Skills and Sales Assist execute work over the same layer.

That is multi-step and unprompted rather than assistive, and the chain can be inspected from outside through rule execution logs.

Sourcefireflies.ai/pricing and guide.fireflies.ai admin controlsread 2026-09-04

An internal corpus plus external retrieval. AskFred queries the entire meeting history in natural language over a searchable archive, with smart search and topic trackers structuring recurring themes, and the API exposes the same transcripts programmatically. Talk to Fireflies adds live external grounding, running a web search inside a meeting across more than sixty languages, so a question the archive cannot answer is answered from outside it in the moment. No citation or source attribution on generated answers is documented.

Sourcefireflies.ai/pricing and docs.fireflies.airead 2026-10-01

Real limits on what the system may do, and no gate before it does it. The key control is segregation: Integration Rules keep internal meetings out of client records, which prevents a failure most CRM-writing agents are exposed to, since an agent that syncs everything will eventually sync the internal call about a client into that client's record.

Administrators also enable or disable integrations and AI skills per team, the Rules Engine routes meetings to specific channels or user groups by policy, custom retention bounds what is kept, and a super admin role holds organization-wide control.

None of that is an approval step: no review, confirmation prompt, confidence threshold or hold is documented before CRM autofill writes structured fields into a system of record or before a rule fires, and the vendor's own framing is that these run without manual entry.

Sourceguide.fireflies.ai admin controls and fireflies.ai/pricingread 2026-09-04

Both certifications and access controls are documented. SOC 2 Type II and GDPR compliance are included on every plan, including Free, so the baseline attestation applies before an enterprise contract.

Enterprise adds HIPAA with a business associate agreement, FERPA with a signed data sharing agreement, SSO and SCIM provisioning, audit logs through the API, custom data retention, a super admin role with organization-wide visibility for compliance, and Private Storage on the customer's own bucket.

Admins control which integrations and AI skills are enabled for the team, so the capability surface itself is governed. Data is encrypted throughout.

Sourcefireflies.ai/security and guide.fireflies.ai admin controlsread 2026-10-01

A record of what the automation did. The API exposes fireflies_get_rule_executions, documented as retrieving automation rule execution logs on Enterprise, so a customer can query which rules fired and what happened rather than inferring it from the result.

Alongside sit audit logs retrievable through the API, a super admin role with organization-wide visibility across meeting content through dashboard and API for compliance, Team Insights, and conversation intelligence surfacing sentiment, key topics, question detection, speaker talk time and custom trackers. The rule execution logs cover the automation itself rather than only user activity. Conversation intelligence analyzes the meetings, not the agents, and nothing documents a trace of what CRM autofill wrote.

Sourcedocs.fireflies.ai and guide.fireflies.ai admin controlsread 2026-10-01

A durable meeting archive, not agent memory. Meeting history accumulates as a searchable corpus, Contacts and topic trackers structure recurring themes and people across it, and custom data retention policies govern its lifecycle. That archive is what AskFred queries, and a record store is not memory: no memory the agents write and read back as their own state, no per-user preference or recall carried between conversations, and no agent-level forgetting control is documented.

Sourcefireflies.ai/pricing and guide.fireflies.ai admin controlsread 2026-10-01

The customer chooses where the data lives, including whether a recording exists at all. Private Storage keeps meeting content in the customer's own storage bucket in their preferred location rather than in shared vendor storage, as documented in the admin controls guide, and a transcript and summary only mode captures text without storing audio or video. Custom data retention policies govern the lifecycle of what is kept. Private Storage and custom retention are Enterprise features.

Sourceguide.fireflies.ai admin controls and fireflies.ai/pricingread 2026-10-01

Named capabilities that ship ready to run, from which an administrator selects. Voice Agents, AI Skills, Sales Assist, AskFred, CRM autofill, topic trackers, sentiment and question detection all arrive built, and the Rules Engine lets a customer compose workflows over them without code.

The admin controls guide documents administrators enabling or disabling integrations and AI skills individually according to team requirements, so the AI Skills are a set chosen from rather than a fixed bundle that arrives switched on. It is an administrative toggle over a set, and no published gallery lets a customer browse and adopt configurations.

Sourceguide.fireflies.ai admin controls and fireflies.ai/pricingread 2026-09-04

Event-driven on three surfaces. The Rules Engine fires on meeting events; the bot joins automatically from the calendar without anyone starting it; and Talk to Fireflies is invoked by voice during a live meeting by saying Hey Fireflies, or by a slash command, in Zoom, Google Meet or Teams, returning a web search result inside the running call across more than sixty languages, available on the free plan. The API returns real-time captions while a meeting is live, as well as processed transcripts afterwards. Outputs go to CRM writes, channel and user-group routing, the Task Feed and Slack.

Sourcefireflies.ai/pricing and docs.fireflies.airead 2026-10-01

The customer does not choose the model, and no provider is named for what powers the product. No model list, routing policy, per-workspace selection or bring-your-own-model option appears on the pricing page, the security page, the admin controls guide or the developer documentation. The only provider named anywhere is Perplexity, whose partnership powers Talk to Fireflies, the live in-meeting web search. That is a supplier for one retrieval feature and says nothing about which models produce summaries, run AI Skills or drive Voice Agents.

Sourcefireflies.ai/pricing and fireflies.ai/securityread 2026-09-04

Three external surfaces, all documented. A GraphQL API at api.fireflies.ai/graphql with published developer documentation at docs.fireflies.ai, bearer token authentication and cursor pagination, included on paid plans and the documented route for audit log and transcript retrieval as well as integration. Programmatic meeting sharing and an upload mutation for audio.

And an MCP server with a published tool reference, including fireflies_get_rule_executions for retrieving automation rule execution logs on Enterprise. The API also exposes live meeting state, returning real-time captions for a meeting in progress rather than only processed transcripts, so a customer can build on the in-meeting surface rather than only use it through the product. No SDK or client library is published.

Sourcedocs.fireflies.ai and api.fireflies.ai/graphqlread 2026-09-04

Nothing lets a customer test the agents before or after they act. Team Insights and conversation intelligence analyze the conversations rather than the agent output, so what is measured is how meetings went, not how well the system read them. No evaluation harness, scored test set, benchmark, regression check when a rule or skill changes, sandbox or dry run is documented on the pricing page, the security page, the admin controls guide or the developer documentation.

That matters here: CRM autofill writes structured fields into a system of record without manual entry, and rules fire on meeting events unattended, so the outputs that most need testing are the ones reaching Salesforce and HubSpot. Rule execution logs show what happened afterwards, but seeing is not testing, and nothing lets a team try a change against a known set first.

Sourcefireflies.ai/pricing and docs.fireflies.airead 2026-09-04

Capture happens inside conferencing tools the customer has licensed, and everything else moves through interfaces.

A meeting bot joins Zoom, Google Meet or Teams through those platforms' own participant model, a Chrome extension and a bot-free desktop app offer other capture routes, and actions leave through native CRM connectors, the Rules Engine, the Task Feed and the GraphQL API.

No browser control, page navigation, portal autofill, form filling or agent-driven computer use appears on the pricing page, the admin controls guide or the developer documentation. A bot attending a meeting is a participant, not an operator.

Sourcefireflies.ai/pricing and docs.fireflies.airead 2026-09-04