More control, scale, and transparency for AI-powered operations and integrations
In September, Pipefy releases advanced AI governance, integration scalability, and process control. The updates expand visibility into costs and executions, simplify recurring tasks, and make the management of automations, data, and pipe changes safer and more predictable.
See all releases:
Pipefy iPaaS: High-volume file processing, real-time queue visibility, and temporary AWS credentials
High-volume operations and large data integrations face bottlenecks when automation engines attempt to load massive files all at once or when visibility over execution queues is limited. Version v88.4 of the Pipefy Integrations (iPaaS) engine delivers improvements focused on stability at scale, efficient handling of large files and workloads, faster troubleshooting, enhanced security, and a smoother flow-building experience.
What's new?
- Incremental CSV batch processing: CSV files can now be processed incrementally in batches of up to 10k rows, allowing large imports to run in parallel without loading the full file into memory.
- Efficient large file handling: File processing no longer requires loading the entire file into memory when reading or writing across Amazon S3, Google Drive, Dropbox, OneDrive, SharePoint, Azure, and SFTP.
- Execution queue visibility: Metrics showing queue wait time versus active execution time, alongside a snapshot of the current queue state.
- Temporary AWS credentials support: Connections using Amazon S3, AWS Bedrock, and AWS Secrets Manager can now use temporary AWS credentials instead of storing permanent access keys.
- Faster data mapping experience: Field searches from previous steps run significantly faster within the workflow builder.
- Expanded connectors and actions: New native apps available, including Microsoft SQL Server, RingCentral, QuickBooks Online, Microsoft Teams, Workday, PolyDoc, LetMePost, DataForB2B, LinkupAPI, and Sofya, alongside new actions across existing connectors.
- Smoother builder experience: Enhancements to long lists, formulas, keyboard shortcuts, tables, result downloads, HTML/image previews, connection validation, and missing-app warnings.
- Engine stability fixes: Structural fixes addressing stuck automations, duplicate scheduled runs, and generic memory error reporting.
Company Overview: Month-over-Month AI Credit Chart
Scalable processes and complex automation ecosystems demand clear financial visibility to prevent budget surprises and maintain operational continuity. Company Overview expands its analytical depth for multi-organization administrators by bringing historical AI usage tracking to the multi-org level. This update removes visibility gaps around agent and assistant adoption across the ecosystem, providing IT leaders with precise data to forecast budgets, monitor consumption trends, and optimize resource allocation over time.
What's new?
- Month-over-Month AI Credits Consumption Chart (Multi-Org): a new reporting layer in the multi-org interface showing historical monthly AI credit usage;
- Unified view within the Company Overview ecosystem: this visualization joins existing governance capabilities, including integration usage stats, proactive limit alerts, API call reports by operation or token, and consolidated multi-org metrics.
AI Governance: Clear execution tracing for advanced agent capabilities
Critical operational processes require thorough auditability over automated decisions to prevent errors and bottlenecks in daily operations. The update to AI Agent logs introduces five new tracing nodes, allowing full inspection of every lookup, search, calculation, or reasoning step before any action is taken on a card. To trace these steps, go to AI Agents > Logs > See Details > Tracing.
What's new?
- Data Lookup: displays the data sources consulted and records returned from databases during execution.
- Web Search: reveals the exact search queries and internet results used by the agent.
- Calculations & Analysis: shows the code executed in a secure sandbox and its resulting values.
- Max Effort: outlines the agent multi step iterative reasoning path for complex evaluations.
- Plain Text: verifies that plain text instructions and rules loaded successfully.
AI Agents: Google Drive and Microsoft OneDrive as real time knowledge sources
Process management loses efficiency and exposes operations to errors when critical information like internal policies, price sheets, and contract templates must be manually downloaded and re uploaded into artificial intelligence tools. Pipefy AI Agents now read documents stored directly in Google Drive and Microsoft OneDrive. Your team connects files where they already live, keeping the knowledge base in sync automatically, with no manual copies going stale.
What's new?
- Google Drive as a knowledge source: direct connection to files in Google Docs, Google Sheets, Google Slides, PDF, DOCX, TXT, CSV, and xlsx, available under the Add document menu inside the agent's Knowledge Base. Auto sync is enabled by default and can be turned off per file.
- Microsoft OneDrive as a knowledge source: connect a Microsoft account and paste the link to a OneDrive file right from the Add document menu. Supports PDF, DOCX, XLSX, TXT, CSV, PPTX, DOC, XLS, and PPT.
- Content preview: for both connectors, you get a preview of the file before committing it to the Knowledge Base.
- A more complete Knowledge Base: the agent's knowledge base now combines Documents/PDFs, Pipes & Databases, Plain Text, Google Drive, and Microsoft OneDrive in one place.
Automated add-on consumption alerts: email and Webhooks
Admins and Super Admins now receive automated alerts whenever API calls and Automations consumption hits 80%, 90%, 100%, and 150% of plan limits. Alerts arrive via email for direct visibility and can also be configured through Webhooks, delivering the same notice directly to your internal systems and automation workflows.
What's new?
- Automated email notifications at four critical thresholds: 80%, 90%, 100%, and 150% of consumption limit.
- Configurable Webhooks that send these same alerts in real time to internal systems, communication platforms, or custom software.
- Email preference management in User Settings; Webhook configuration in Account Settings.
- Initial coverage for API calls and Automations, with continuous expansion to remaining add-ons.
Suggested names for automations and conditional fields
Automations and conditional fields now come with a suggested name. The name is built from what you configured. Anyone opening the list understands what each one does without opening its setup.
Until today, a new automation started with an empty name. The person who created it knew what it did. Anyone inheriting the process, or the same person months later, had to open each one. In a pipe with dozens of automations, the list stopped being readable.
What's new?
- Suggested at creation. The name comes from the trigger and action you configured. For conditional fields, it comes from the condition and the affected field.
- Works on what you already have. When editing an existing automation or conditional field, ask for a suggested name and bring the whole pipe up to the same standard.
- Both objects. Automations and conditional fields.
- Always editable. It's a suggestion, not a rule. Prefer your own naming? Just edit it.
Pipefy AI: Lite and BYOM Now Cost Fewer Credits
We adjusted credit consumption for the Lite and BYOM (Bring Your Own Model) options in Pipefy AI. Running an automation with Lite now costs half of what it did, and BYOM runs cost a third less.
The change is already live for the entire base, no configuration needed.
What's new?
- Pipefy AI Lite: consumption drops from 2 credits to 1 credit per run.
- Bring Your Own Model (BYOM): consumption drops from 1.5 credits to 1 credit per run.
- Standard and Pro stay at 2 credits per run (Pro is in its promotional period until 10/08, then moves to 3 credits).
AI Agent Logs: new filters and date range calendar make it faster to find executions
Investigating AI Agent executions just got faster. AI Agent Logs now include additional filters, a new Card ID column, and a calendar mode for selecting date ranges, making it easier to find the exact execution you need without scrolling through long lists.
What's new?
- Date range calendar: filter executions by a specific period directly from the date picker.
- Card ID column: see which card each execution is linked to and jump straight to it.
- Triggers column and filter: identify what started each execution and filter by trigger type.
- Improved keyword search bar, now displaying results correctly.
[BETA] Pipe Snapshots and Restore via API: save your process setup and return to the latest stable version
Changing the setup of a live process can be unsettling: a structural change might break active rules and bring operations to a halt. Pipe Snapshots brings the same concept as the UI's version history to the API. Each snapshot is a saved version of the pipe's setup, and the workflow can be restored to the latest snapshot if a change does not go as planned.
What's new?
Snapshot and restore capabilities are available through GraphQL API calls in the Pipefy Developer Portal:
- Full pipe snapshot (Create Pipe Snapshot): generates a structured JSON file containing setup elements such as phases, fields, automations, conditionals, and more.
- Restore to the latest stable version (Restore Pipe to Last Snapshot): reverts the pipe's setup to the latest saved snapshot. This action cannot be undone.
- View and manage history (Query & Rename Pipe Snapshots): list a process's saved snapshots and rename them to keep versions organized.
Version history: save, download, and track your pipe setup
Version history lets you save and download any pipe's setup directly in Pipefy. It helps IT governance teams see how a process was configured on a specific date and provides a record for safer adjustments before high-risk operational changes.
What's new?
- One-click version creation: save a version of your pipe's setup under Manage → Version history. You can rename each record to make it easier to identify.
- JSON file download: download a version's JSON file to inspect the setup, compare configurations from different points in time with a diff tool, or share it with an AI agent.
- Completion notification: receive an in-app notification confirming that processing is complete and the version is ready in the list. The notification appears when you refresh the page.
- Pipe traceability: view each version's ID, author, and date, with events automatically recorded in the history under Activities.
Card Duplication: create full copies in seconds
Cards that repeat, like recurring requests or similar tasks, can now be duplicated directly in Pipefy. The feature is already available to all accounts and removes the need to rebuild a card from scratch every time the process repeats.
What's new?
- Duplicate card: open any card, click the menu in the top right corner and select Duplicate (only visible to users with permission to create cards in that pipe).
- Choose the copy scope: decide whether to copy only the start form fields, or all fields and card attributes, including assignees, labels, and due dates.
- Duplicated cards always land in the first phase: every card created through duplication automatically starts in the pipe's first phase, following the normal process flow.
- First version limitation: for now, attachments and connections are not duplicated, with the exception of connection fields.
Pipefy Piece: batch operations for large data volumes
You can now process thousands of cards and records at once through the Pipefy Piece, without building loops to handle each item individually inside a flow. The new batch operations cover both sending data into Pipefy and exporting information from a pipe or database into a single file, making integrations, migrations, and initial data loads much faster.
What's new?
- Batch Upsert: send a CSV or JSON file to create or update thousands of cards or records in a single operation. Use up to 3 combined fields to identify existing items, or match by ID or title, and choose whether unmatched items should be created or skipped.
- Batch Export: export all cards from a pipe or records from a database into a single CSV or JSON file. Use a saved report or an advanced filter to define what gets exported, and choose which columns are included.
- Processing tracking: track the status of each batch operation and access a row-level error report, reprocessing only the rows that failed without re-uploading the whole file.
- Wait for batch completion: use this when the next steps in your flow need the final processing results before continuing.
- Duplicate protection: operations use an idempotency key to prevent duplicate submissions or exports.
Pipefy Integrations: more stable automations and an updated connector set
We've upgraded the automations engine behind Pipefy Integrations. Most of the work happens behind the scenes, but it translates into more reliable flows, clearer error messages, better performance, and a more up to date set of connectors.
What's new?
- Clearer action labels: actions now show whether they read, search, write, or delete data, with delete actions highlighted in red.
- Cleaner step forms for HTTP Request, Google Sheets, and Custom API Call.
- Automations are sorted alphabetically, with friendly messages instead of blank screens when something fails to load.
- Over 270 updated connectors, including QuickBooks Desktop, HubSpot Private App token, and a new Google Drive trigger.
- Subflows scheduled for retry no longer cause the parent flow to fail.
- OAuth2 connections with special characters can now be saved correctly.
- Publishing a scheduled flow no longer breaks its schedule.
- Older flows and templates can be imported without errors.
- Large flows run faster and lighter.
- Newly installed connectors no longer fail with a "piece not found" error.
- Error messages now point to the actual cause instead of a generic error.
- Members can no longer access connections from other projects.
- Synchronous webhooks now return 500, instead of 200, when a flow fails. If any system relies on that response, it's worth reviewing that integration.
- HTTP headers are preserved exactly as entered, with no automatic conversion to lowercase.
- The Kimai connector now uses Bearer token authentication. If you already use Kimai, you'll need to reconnect.
AI Credits: see execution costs before and after you run
AI Agents now show you what they cost to run, before and after execution. You get an AI Credits estimate when testing a behavior, and the real consumption broken down by node once the agent runs, giving you full visibility into the cost of every configuration.
What's new?
- Test Behavior cost estimate: see how many AI Credits a tested configuration would use before publishing it. Testing does not consume AI Credits.
- Execution Summary: see the total AI Credits consumed by each agentic run, right in the Summary.
- Cost per node: understand how AI Credits were distributed across individual execution nodes.
