FlowRunner
PricingContact
Theme
Start Free

Connect AI agents to Crisp, the customer messaging platform. Agents list and read chat conversations, post operator replies into the visitor's widget in real time, and create or update contact profiles on your Crisp website.

9 actions Basic auth available
Crisp website ↗ Platform Documentation ↗ Capability data verified 2026-07-28
A sweep runs against the website's inbox every minute during business hours
List Conversations surfaces sessions with new inbound visitor messages
List Messages pulls the full transcript for each active session, in order
The agent classifies the conversation: routine question it can answer, or a topic that needs a person
Send Message posts the drafted operator reply into the visitor's widget for routine questions
For refunds, security topics, or frustrated visitors, the support lead reviews the draft before it appears in the widget
Escalated sessions post to the support channel with the transcript attached

What This Integration Enables

Live chat makes a promise the widget cannot keep by itself: that someone is there. Crisp puts the conversation surface on your site; this connector puts a FlowRunner agent behind it. Agents read conversations and full transcripts, post operator replies that appear in the visitor's widget in real time, and keep contact profiles current, so the promise of the little green dot holds at midnight the way it does at midday. - Answer routine inbound chats with drafted operator replies, instantly - Read full conversation transcripts to summarize, triage, and log support demand - Sync chat contacts into the CRM, the mailing list, or the reporting sheet - Create and update contact profiles from forms and external sources - Escalate the conversations that need judgment, with the transcript attached The widget is a stage, and everything posted to it is live. That is why FlowRunner's [human-in-the-loop](/concepts/human-in-the-loop) orchestration decides which replies flow and which wait for review. What separates this from a chatbot bolted onto the widget is where the intelligence lives. A FlowRunner agent answering through Crisp is a full workflow participant: it can check the order system before answering a shipping question, look up the subscription before discussing billing, and consult the documentation before making a claim, because every other connector in your stack is available to the same flow. The reply that lands in the widget is grounded in your systems, not in a canned script, and when the agent is not confident, the escalation path is part of the same flow rather than a dead end that promises an operator will be along shortly.

Without FlowRunner

Chat rewards whoever is watching Visitors get instant answers at 2 PM and silence at 7 PM, and the difference is staffing
Transcripts evaporate What visitors actually asked lives in scrollback nobody mines for patterns
Chat contacts stay in chat People who wrote in never make it to the CRM or the mailing list

With FlowRunner

First response in seconds, always Agents draft and send replies to routine questions the moment they arrive
Every conversation becomes data Transcripts summarize into logs and reports, and recurring questions surface as trends
Contacts flow to the stack Chat visitors become CRM records and list members while the conversation is still warm

Use Case Scenarios

The overnight chat that gets answered anyway

At 11 PM a visitor asks whether the product supports single sign-on. The sweep catches the session in List Conversations, pulls context with List Messages, and the agent drafts a factual answer from the documentation. Send Message posts it into the widget within the minute. The next morning, the conversation and its resolution are already logged, and the visitor did not bounce to a competitor who happened to be awake.

Chat contacts become CRM records mid-conversation

When a visitor shares an email, the agent checks List People for an existing profile, calls Create Person for new contacts or Update Person to enrich known ones, then mirrors the record into [HubSpot](/integrations/hubspot) and enrolls it in the right list in [Mailchimp](/integrations/mailchimp-marketing). Sales sees the chat history behind the lead. The conversation becomes a relationship record instead of a closed tab.

Support demand, measured from transcripts

Nightly, the agent walks the day's sessions with List Messages, summarizes each conversation, and logs topic, resolution, and duration to [Google Sheets](/integrations/google-sheets). Recurring questions cluster into a weekly digest posted to [Slack](/integrations/slack): what people keep asking, what the docs fail to answer, which pages generate confusion. The support inbox stops being anecdote and starts being a dataset the product team argues from.

Human-in-Loop Highlight

Send Message posts as an operator into the visitor's chat widget in real time. There is no draft state, no recall, and no delay between the action firing and the words appearing on a customer's screen under your brand. For an order status question, that immediacy is the whole point. For a refund demand, a security question, or a visitor who is already angry, an instant wrong answer is a screenshot on its way to social media. So FlowRunner agents split the queue: routine classifications flow straight through Send Message, and flagged topics pause with the transcript and the drafted reply in front of the support lead. The lead edits or approves, and only then does the reply hit the widget. Speed where speed is safe, judgment where it is not.

Agent processes routinely
Detects the widget
Clear match Continues automatically
Ambiguous Routes to human via preferred channel
Human decides
Agent resumes with decision

Agent Capabilities

9 actions

Conversations

5
  • List Conversations Lists the website's conversations by most recent activity, with full-text search across messages, contact details, and metadata. The sweep behind every chat automation.
  • Get Conversation Retrieves a session's full metadata: status, assigned operator, contact profile, participants, segments, and last message. The context read before routing.
  • Create Conversation Creates a new empty conversation and returns its session ID, ready for a first message via Send Message. Used to open outbound sessions programmatically.
  • Send Message Sends a text message into a conversation as an operator, appearing in the visitor's widget in real time. The send this page's human gate exists for.
  • List Messages Retrieves a conversation's messages chronologically, with type, direction, content, and timestamp. The transcript read behind summaries, triage, and logging.

People

4
  • List People Lists the website's contact profiles, searchable by name and email, roughly 50 per page. The matching step before any profile write.
  • Get Person Retrieves a profile in full: email, name, company, geolocation, segments, and custom data. The enrichment read for CRM sync.
  • Create Person Creates a contact profile with an email and optional name, returning the people ID. Fails if the email already exists, which makes duplicates loud instead of silent.
  • Update Person Updates a profile's email, name, phone, or company, changing only the fields provided. The maintenance write behind profile enrichment.

Frequently Asked Questions

What can FlowRunner do with Crisp?

FlowRunner agents can run List Conversations, Get Conversation, and Create Conversation in Crisp, plus 6 more actions.

Does connecting Crisp to FlowRunner require OAuth?

No. Crisp connects to FlowRunner with basic authentication (a username and password), no OAuth flow required.

Can Crisp trigger a FlowRunner workflow automatically?

Crisp doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with Crisp

$100 in credits. No card required. Connect in minutes.