FlowRunner
PricingContact
Theme
Start Free

Start outbound AI phone calls through a CallX account and search the call log history. Agents place a call, read the outcome, and pass the transcript to whatever comes next in the flow.

1 trigger 4 actions API key available
CallX website ↗ Platform Documentation ↗ Capability data verified 2026-08-03
A form submission creates a lead in the CRM
Agent reads the submission fields and the history behind the email domain
Agent derives the lead's local time from the number and checks it against permitted calling hours
Agent looks for an open support ticket or an existing customer record on the same number
Agent calls Make a Call for the leads that clear every check, and CallX dials with the chosen assistant
Agent posts the leads it held back with the specific rule each one failed
A person decides whether each held lead gets dialled, because CallX has no action that recalls a queued call

What This Integration Enables

CallX is deliberately small, and that is the reason to choose it. There are no campaign objects, no contact database, no rosters, no schedules. There is Make a Call, and there is the call log you read afterwards: search it, pull the AI-generated summary with its extracted key points, and pull the full transcription as speaker-tagged segments or plain text. Four operations, and none of them ask you to move your data into somebody else's product.

That narrowness makes it a good fit for a very particular architecture. The lead list stays in the CRM. The rules about who gets called, when, and by which assistant stay in the flow, where the rest of the business logic already lives. CallX is a step, not a platform. An agent decides a call should happen, places it, and reads back what was said. The tradeoff is worth stating plainly, because it is where the human-in-the-loop argument bites: with no campaign scheduler and no cancel operation, every guard rail this connector has is one you build into the flow, and every call it places is final the moment it is queued.

Without FlowRunner

Speed comes at the cost of judgment The fastest setup calls everyone, including the customer who just opened a support ticket
Judgment comes at the cost of speed Adding a review step means leads wait in a queue nobody watches
Call content stays in the calling tool Summaries and transcripts live where the CRM cannot see them

With FlowRunner

Clean leads move immediately Anything that clears the rules is dialled without waiting for a person
Edge cases surface as decisions Held leads arrive with the rule they failed and a yes or no attached
Content lands in the record Summary, key points and full transcript are written back where the team works

Use Case Scenarios

Form to phone call in seconds, with the exceptions held back

A demo request lands. The agent enriches it, works out the lead's local time from the number, checks whether the same number belongs to an existing customer with an open ticket, and for the leads that clear every rule calls Make a Call with the assistant that handles inbound demand. The response returns the call SID, call log ID and direction, which the agent writes onto the CRM record so the call is traceable before it has even connected. The leads that did not clear go to a person with the reason attached. Most of the volume moves at machine speed; the handful that would embarrass you do not.

Every call summarised into the record

The On New Call Log trigger fires as a call finishes. The agent calls Get Call Summary for the summary text and extracted key points, writes them onto the opportunity in Pipedrive, and posts the key points into the owning rep's channel in Slack so the human follow-up starts from what was actually discussed. Calls whose ended reason indicates the assistant hit a wall are flagged as automation exceptions rather than filed as completed.

Transcripts pulled for quality review and dispute handling

A customer disputes what they were told on the phone. The agent runs Search Call Logs filtered by the number and date range, finds the call, and pulls Get Call Transcription as speaker-tagged segments with timestamps. The reviewer reads exactly who said what and when, in order, rather than listening to a recording with a stopwatch. The same search filters, by assistant, direction and ended reason, drive the weekly quality sample.

Human-in-Loop Highlight

The gate is Make a Call, and it deserves one because of what this connector does not include. There is no Stop Call here, no scheduled-batch state to cancel, no draft campaign to review. Once Make a Call returns a call SID, CallX is dialling, and the API offers no way to take it back. That makes the decision to dial the last reversible moment in the whole flow. So the agent puts real work into that moment rather than treating it as a formality. Leads that clear the rules go through untouched. Leads that trip one stop, and the agent asks a question with the specific reason in it: "Rachel Ng submitted a demo request from an account with an open P1 support ticket opened yesterday. Local time is 7:42pm. Call her now with the inbound assistant, hold her until 9am tomorrow, or route her to her account manager instead of an AI call?" Somebody answers, and the call happens or it does not. The point is not that a person approves calls. It is that the one operation with no undo is the one operation the flow refuses to guess on.

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

Agent Capabilities

4 actions

Calls

1
  • Make a Call Creates a new outbound call from a CallX phone number to a customer's phone number, optionally specifying which AI assistant handles it. Returns the queued call with its call SID, call log ID, from and to numbers, and direction. There is no companion operation to cancel a queued call, so the decision to place it is the final one.

Call Logs

3
  • Search Call Logs Retrieves call logs with optional filters for date range, free text search, call direction, assistant and ended reason. Paginated up to 100 entries per page. Each log includes the caller, phone number, start and end time, duration, cost in cents and ended reason.
  • Get Call Summary Retrieves the AI-generated summary for a specific call log, including the summary text and extracted key points. Optionally returns the summary as plain text instead of a structured object. Used to write a readable account of the call into the CRM without shipping a transcript into a text field.
  • Get Call Transcription Retrieves the transcription for a call log as an array of transcribed segments, each with the speaker, the spoken text and a timestamp. Optionally returns plain text. Used for quality review, dispute handling and archiving what was actually said.

Triggers

1 triggers

Event Triggers

1
  • On New Call Log Fires when a new call log appears on the account, including inbound calls the assistant handled. Used to pull the summary and transcription, write them into the record, and route follow-up to the team based on how the call ended.

Frequently Asked Questions

What can FlowRunner do with CallX?

FlowRunner agents can run Make a Call, Search Call Logs, and Get Call Summary in CallX, plus 1 more action.

Does connecting CallX to FlowRunner require OAuth?

No. CallX connects to FlowRunner with an API key, no OAuth flow required.

Can CallX trigger a FlowRunner workflow automatically?

Yes. CallX supports 1 trigger that can start a FlowRunner workflow automatically.

Start building with CallX

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