Pipefy

Connect AI agents to Pipefy business processes. Agents create cards from forms or tickets, move them between phases, update field values, post comments, and run arbitrary GraphQL against anything the dedicated actions do not cover.

10 actions API key available

Pipefy website ↗ Platform Documentation ↗ Capability data verified 2026-07-28

A purchase request arrives through the intake form with vendor, amount, and cost center

Get Pipe reads the procurement pipe's phases and start-form field IDs

Create Card opens the request in the intake phase, with the form answers mapped into fields as field_id and field_value pairs

The agent reads the amount against the approval policy: under threshold routes one way, over threshold another

Move Card To Phase advances under-threshold requests toward fulfillment and places the rest in Manager Review

The budget owner opens the card in Manager Review, reads the request in full, and approves or rejects it

Create Comment logs the decision and its reasoning on the card, and Update Card Field records the PO number

The requester's channel gets the outcome with the card link, so nobody asks procurement for a status

What This Integration Enables

A Pipefy pipe is not a to-do list, it is a policy that runs. Companies encode their procurement, hiring, and intake processes into phases precisely so that a card sitting in Approved means someone checked. FlowRunner agents respect that semantics: they do the clerical movement, creating cards from upstream events, syncing fields, advancing work whose conditions are objectively met, and they stop at the phases that exist to hold judgment, because a phase is a promise about what happened there. - Create cards automatically from form submissions, emails, and support tickets, with answers mapped to start-form fields - Advance cards between phases as conditions are met, and revert them when downstream checks fail - Keep card fields synchronized with external systems of record - Log hand-offs, decisions, and cross-system context as comments where the process's audit trail lives - Reach the rest of the Pipefy schema through arbitrary GraphQL when the dedicated actions do not cover the operation

Without FlowRunner

Process by memory The purchasing policy lives in a PDF, and whether a request followed it depends on who handled it

Cards stall between owners A request clears one desk and waits days for someone to notice it needs the next one

Status means asking The only way to know where a request stands is to interrupt whoever touched it last

With FlowRunner

Policy runs as workflow Every request enters through the same form, the same fields, the same phases, no side doors

Movement is event-driven Cards advance the moment their conditions are met, and only their approval phases wait, for approvers

The pipe answers status Phase position is the status, readable by agents, reportable on schedule, visible to the requester

Use Case Scenarios

The intake pipe with no side door

Submissions from [Typeform](/integrations/typeform) become cards the moment they arrive: the agent reads the pipe's structure with Get Pipe once, then calls Create Card per submission with every answer mapped to its field. Requests that arrive by email get parsed and enter through the same door. The intake phase becomes the single, complete queue, and the requests that used to arrive as hallway favors now have card IDs and timestamps.

The pipe that reports itself

Every Monday, the agent pages through List Cards for the procurement pipe, reading each card's title, phase, and field values, and checks the approval phases with Get Phase for anything aging past its SLA. The snapshot lands in [Google Sheets](/integrations/google-sheets), and a digest posts to [Slack](/integrations/slack) naming the cards stuck in review with their owners. Cards that cleared into Done during the week get announced as they land, because finished work deserves a heartbeat too.

The card that mirrors the outside world

A vendor invoice reaches the ERP and its payment clears. The agent finds the matching card with List Cards, confirms its state with Get Card, writes the payment reference with Update Card Field, posts the confirmation as a Create Comment, and advances the card to Paid with Move Card To Phase. The process record in Pipefy and the financial record in the ERP tell the same story without a person reconciling them at month end.

Human-in-Loop Highlight

Two operations in this connector carry authority beyond their syntax. Move Card To Phase can place a card in an approval phase's outbox without the approval ever happening, so FlowRunner agents never move cards across a judgment phase; they move cards to those phases and wait for the human whose name the phase implies. And Run GraphQL is a loaded escape hatch: it executes arbitrary mutations, including card deletion, which the dedicated actions deliberately do not offer. When a cleanup workflow concludes that 40 duplicate cards should be deleted, the agent drafts the exact mutation, lists the 40 card titles and their field values, and a person reads the list before anything runs. A deleted card takes its comments, history, and audit trail with it, and no one gets that back with a second mutation.

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

10 actions

Pipes

Get Pipe Retrieves a pipe by ID with its phases and start-form fields. The structural read that supplies phase IDs for movement and field IDs for card creation, done once per pipe rather than guessed per card.
List Organizations Lists the organizations the authenticated user belongs to, with the pipes inside each. The discovery step when organization and pipe IDs are not already known.

Phases

Get Phase Retrieves a single phase with the cards currently sitting in it, paginated. The read behind SLA sweeps and queue-depth reporting on approval columns.

Cards

Create Card Creates a card in a pipe's first phase with a title and optional start-form field values as field_id and field_value pairs. The write that turns an inbound request into a governed work item.
Get Card Retrieves a card in full: title, current phase, assignees, labels, and every field value. The state read before any decision or cross-system sync.
List Cards Lists a pipe's cards with pagination, returning title, phase, and field values. The reporting and reconciliation workhorse.
Move Card To Phase Moves a card to a different phase in the same pipe, advancing or reverting the work item. Powerful precisely because phase position carries meaning, which is why judgment phases stay human.
Update Card Field Updates a single field's value on a card, including array values for connector and checklist fields. The sync instrument that keeps cards honest against external systems.

Comments

Create Comment Adds a comment to a card with author and timestamp. Where agents log decisions, hand-offs, and context so the card's history reads as an audit trail.

Advanced

Run GraphQL Executes an arbitrary Pipefy GraphQL query or mutation with optional variables and returns raw data. The escape hatch for tables, labels, webhooks, and deletions, and the action this page's human gate watches most closely.

Frequently Asked Questions

What can FlowRunner do with Pipefy?

FlowRunner agents can run Get Pipe, List Organizations, and Get Phase in Pipefy, plus 7 more actions.

Does connecting Pipefy to FlowRunner require OAuth?

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

Can Pipefy trigger a FlowRunner workflow automatically?

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

Start building with Pipefy

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

Start Building Free Talk to us