📍 New capability in closed beta. Available starting November 18, 2026, for a select group of customers. |
What is Pipefy Copilot
Pipefy Copilot is Pipefy's agentic process expert. It lets you build, optimize, run, and get insights from your pipes just by chatting in plain language, with no need to click through configuration screens.
Just ask, and Pipefy Copilot can set up and build:
Processes (pipes), including the phases they need;
Card fields and conditional display rules;
Automations and AI Agents, including general instructions, knowledge base, and behaviors;
Database tables;
Interfaces.
Beyond building, Pipefy Copilot handles day-to-day process work (creating and moving cards, filling in fields, adding comments), answers questions about your pipe data, and helps you design and improve processes.
Pipefy Copilot doesn't replace the configuration screens. Everything it does can be reviewed and adjusted in the usual screens, and no Pipefy feature depends on it to work.
Availability during the closed beta
📢 Closed beta:
Starting November 18, 2026: Pipefy Copilot becomes available to a select group of customers, rolling out gradually. Participating organizations are notified in the product, and Pipefy Copilot comes turned on, with no extra setup required.
During the closed beta: Pipefy Copilot is free to use, with no usage limits. There's no set end date for the closed beta. Dates and terms for the official launch will be announced in advance.
Because it's in closed beta, Pipefy Copilot is still being evaluated. Stability, availability, and response times may vary, and features may be adjusted or changed along the way.
If your organization hasn't received an invitation, you can request early access. Customers with a CSM can reach out to them directly. Everyone else can submit a request through Pipefy's support channels.
Where to find Pipefy Copilot
Pipefy Copilot lives behind an icon in the top-right corner of the screen, next to your profile picture. How it behaves depends on where you are:
Where | How Pipefy Copilot works |
Home | Your starting point for creating a new process or asking questions about your organization. |
Inside a pipe | The conversation starts in the context of that pipe. Pipefy Copilot helps with the specifics of the process on your screen. |
Interfaces | Especially well suited for creating new interfaces. |
What Pipefy Copilot can do
The table below sums up what you can ask Pipefy Copilot to do in each area of the platform, with a sample request. For the detailed, area-by-area list, see the Pipefy Copilot FAQ.
Area | What you can ask for | Sample request |
Processes (pipes) | Create a complete process in one go, duplicate, change settings, delete, find processes, and understand how a pipe is structured. | "Create a purchase request process with the phases Request, Manager Approval, Finance Approval, Purchase, and Done." |
Phases | Create phases (including between two existing phases), rename them, adjust due dates and descriptions, delete them, and check which phases a card can be moved to. | "Create a Legal Review phase between Negotiation and Signature." |
Fields and display rules | Create, edit, and delete start form or phase fields, including connection fields. Show or hide fields based on the value of other fields. | "Only show the Justification field when the amount is over $10,000." |
Labels | Create, rename, recolor, and delete labels, and add or remove labels on a card. | "Create a red Urgent label in the Purchasing pipe." |
Cards | Create, fill in, move, comment on, connect, and delete cards. | "Move card 123 to Finance Approval." |
Finding and analyzing cards | Search, filter, and count cards by phase, label, assignee, dates, title, and field values. | "How many orders over $5,000 are in the approval phase?" |
Connections between processes | Create, edit, and delete connections between pipes (parent and child), connect cards, and look up connected cards. | "Connect the Purchasing pipe to the Payments pipe, with Purchasing as the parent." |
Database tables | Create, edit, and delete tables and columns, and create, edit, search, and delete records. | "Create a Vendors table with tax ID, legal name, contact, and category." |
Automations | Create "when X happens, do Y" rules with conditions, and list, explain, edit, turn on, turn off, and delete rules. | "When a card reaches Approved and the amount is over $10,000, move it to Executive Approval." |
AI automations | Create rules where AI fills in card fields based on a prompt, validating the prompt before creating the rule. | "When a card is created, use AI to summarize the description in the Summary field." |
AI Agents | Create a complete AI Agent with general instructions and 1 to 5 behaviors (trigger, instructions, and actions), a knowledge base from text or table lookups (documents are added in the agent screen), and extra capabilities. Validate, turn on, turn off, edit, and delete agents. | "Create an agent that, when a purchase card is created, reads the attached quote, fills in the amount and vendor, and moves the card to Approval." |
Card emails | Send emails from the card's inbox, use existing email templates, and read a card's emails. | "Has the vendor replied to the email on card 123?" |
Webhooks | Create, list, edit, and delete webhooks that notify another system when something happens in the pipe. | "Create a webhook that notifies our ERP whenever a card is moved." |
Reports | Create pipe and organization reports with columns, filters, and formulas. Edit, delete, and export reports and a pipe's log. | "Create a report of orders approved this month with amount, requester, and vendor." |
Monitoring | Check the run history of AI Agents and automations, organization usage, and AI credit consumption. | "Why did the triage agent fail yesterday?" |
Interfaces | Create and edit HTML interfaces with real-time data. | Click here for a step-by-step guide to creating interfaces with Pipefy Copilot. |
Process design and improvement | Design a process before building it, analyze an existing process, assess whether an automation is worth it, and get answers to questions about using Pipefy. | "What can be improved in the Purchasing pipe?" |
Organization info | Look up the plan, number of members, and number of pipes. | "How many members and pipes does our organization have?" |
Three key points:
Pipefy Copilot always acts with the user's permissions. It can only do what the user's account can already do in Pipefy.
In advanced automations (iPaaS), Pipefy Copilot sets up flows that run inside Pipefy. It does NOT access or configure the connected external systems: that part is up to the customer.
Pipefy Copilot is an AI and can make mistakes. Always review what it proposes and double-check the results.
How Pipefy Copilot works: plan and approval
Pipefy Copilot never changes anything without telling you first. Before making any change, it presents a plan in business terms, explaining what will change and where. The flow is always the same:
You make a request.
Pipefy Copilot reads what it needs and responds with a plan.
You approve or decline.
If you approve, Pipefy Copilot asks for a final confirmation to publish [confirm the on-screen button name]. Until then, nothing changes in the process.
Once you publish, the changes are applied in the order laid out in the plan.
If you decline, nothing changes. Adjust your request and ask for a new plan.
Each type of action is handled differently:
Action type | What happens |
Queries (questions, searches, counts, analyses) | No approval needed. Reading never changes anything. |
Creating and editing | Requires plan approval. Once approved, changes are applied directly to the process. |
Deleting (pipe, phase, field, card, rule, agent, table, record, report, webhook, and other items) | Requires a specific confirmation and can't be undone. |
Actions that go outside Pipefy, such as sending an email | Require confirmation and can't be undone once carried out. |
New automations and AI Agents | Start running right after they're created. |
Existing automations and AI Agents, edited at your request | New ones start running when the plan is published. In some cases, such as when the pipe already has cards a rule would affect, Pipefy Copilot leaves the item turned off and flags it in the summary. Editing an existing automation or agent keeps its current state. |
If part of a plan fails, Pipefy Copilot tells you what was applied, what wasn't, and what to do about it. For example, if the pipe was created but one of the phases failed, it tells you which phases already exist. After a partial failure, check the process.
⚠️ Heads up: there's no undo button or draft mode yet (except when creating new interfaces, which can already be built as drafts and published later). Once approved, changes take effect immediately, so review the plan before approving. To reverse a change, ask Pipefy Copilot itself (for example, "remove the phase you created") or adjust it in the screen. Deletions and sent emails can't be undone.
Permissions and security
Pipefy Copilot always acts as the user, with the user's permissions, and never uses elevated permissions to complete a task. If the account doesn't have permission to do something, it explains why and doesn't try a workaround. Each conversation is limited to the organization the user is in.
Some areas require specific permissions or settings:
To | You need |
Change a pipe's structure | To be a pipe admin |
Build iPaaS flows | Advanced Automations enabled for the organization |
Create AI automations | AI enabled in the pipe |
AI credits
During the closed beta, chatting with Pipefy Copilot doesn't use any AI credits. AI Agents and AI automations created with its help, however, work like any other AI Agent or AI automation in your organization: once they start running tasks, they use credits based on your plan and the AI model chosen for each behavior.
Usage | AI credit consumption |
Chatting with Pipefy Copilot (questions, plans, creating, and editing) | None during the closed beta |
Running an AI Agent created by Pipefy Copilot | Based on the AI model for each behavior (Lite, Standard, or Pro) and the specific actions performed |
Running an AI automation created by Pipefy Copilot | Based on your plan's AI automation consumption rules |
Pipefy Copilot can also help you keep track of usage: just ask how many AI credits have been used this month and on which features. You can also track usage in the Admin Panel.
What Pipefy Copilot doesn't do yet
During the closed beta, Pipefy Copilot can't:
Set which phases a card can be moved to, beyond the sequence created along with the phases;
Change access credentials, roles, or members of pipes or the organization when the user doesn't have permission to do so;
Reorder a phase that already has cards;
Create or edit email templates. It uses existing templates and, when you ask for a new one, provides the name, subject, and body so you can create it in the screen;
Create AI Agents with scheduled triggers. As an alternative, it suggests triggering the agent when a card enters a phase;
Attach files to cards or records;
Read the contents of attached files. For that, it suggests using an AI Agent;
Build a knowledge base from documents, such as a PDF. Documents are added in the agent screen;
Look up or change data in systems outside Pipefy during the conversation;
Work in Teams, Slack, or WhatsApp;
Send notifications or suggestions on its own, without a user request;
See or change anything the user doesn't have access to.
Responses and conversations
Pipefy Copilot responds with text and, when helpful, with tables, charts, alerts, and lists. It tries to show where each piece of information comes from and to say when it can't see something. Even so, it can make mistakes: double-check important numbers in the pipe itself. When it can't see something, it says so instead of claiming it doesn't exist. Every response ends with a suggested next step.
Languages: Pipefy Copilot replies in the language of your latest message — English, Portuguese, or Spanish.
Conversation memory: the conversation carries over when you switch screens, reload the page, or go from Home to a pipe. To start fresh, use New chat.
History: past conversations are available in your chat history.
Long conversations: Pipefy Copilot automatically summarizes the older parts. For a new topic, use New chat: responses will be faster and more accurate.
Performance and limits
Response time: simple questions take a few seconds. Requests that read a lot of data or make several changes can take up to about a minute. While it works, Pipefy Copilot shows what it's doing.
Large processes: queries work even in pipes with thousands of cards. Questions about the history of many cards take longer. In large, complex processes, review the plan carefully before approving.
Exports: reports and automation histories are generated in the background. Pipefy Copilot lets you know when the file is ready or where to find it.
Usage limit: during the closed beta, usage is free and unlimited.
Getting the most out of Pipefy Copilot
Pipefy Copilot works best when you describe the outcome you want, not the technical steps to get there. The table below shows how to write better requests:
Best practice | How to apply it |
Describe the outcome, not the steps | "Orders over $10,000 go through Finance" works better than "create a conditional field." |
Start with the design | For a new process, ask "help me design" before "create," especially when decisions are still open, such as who approves or what the threshold is. |
Read the plan before approving | Check the pipes, phases, and fields mentioned, and which automations and AI Agents will be created or changed. |
Test before going live | Use a test card to validate automations and AI Agents. To test outside the live process, ask Pipefy Copilot to duplicate the pipe and apply the change to the copy first. |
Check the results | If an AI Agent or automation doesn't behave as expected, ask Pipefy Copilot to check the run history and explain what happened. |
One conversation per topic | Avoid mixing topics in the same conversation. |
Give feedback | Use the thumbs-down on incorrect responses and tell us what you expected. |



