Node catalog
Nodes are the building blocks of an AI agent. You add them to a flow from the Nodes library in the builder, connect them in order, and each one does a single job — start the flow, read a file, extract data, call an external system, branch on a condition, route work to a person, or mark an item complete. Outcomes like classification, extraction, matching, routing, human review, and write-back are produced by composing these nodes, not by separate features.
This catalog is the map. It lists every node you can see in the builder, grouped by the same categories the Nodes library uses, with a short description, what each node takes in and produces, related use cases, and a link to the detailed guidance for that category.
How to read this catalog
- Category — the group the node appears under in the Nodes library: Triggers, Apps, Utilities, Actions, Controls, and InsurGPT.
- What it does — the node's purpose in one line.
- Inputs → Outputs — a short summary of what the node consumes and what it makes available to the nodes that follow.
- Related use cases — where the node typically appears. Some nodes are general-purpose and appear in most flows.
- Status — whether the node is available to use today:
- Available — appears in the Nodes library and can be configured and run.
- Not available — visible in the Nodes library but disabled. It cannot be added or configured, and has no setup guidance until it is enabled.
Detailed setup, configuration fields, and examples for each node live on the category pages linked in each section below. Only Available nodes have setup guidance.
Triggers
Every flow starts with exactly one trigger. The trigger starts the flow, ingests the initial input, and injects it as normalized data for the nodes that follow. See Configure triggers for setup and configuration.
| Node | What it does | Inputs → Outputs | Related use cases | Status |
|---|---|---|---|---|
| Manual File Upload | Start a flow by uploading one or more files by hand | Files you upload → the files and a count, available to downstream nodes | On-demand processing; testing any agent | Available |
| Outlook Trigger | Start a flow when a matching email arrives in a connected mailbox | A filtered inbound email → the message and its attachments | Submission Intake; any email-driven process | Available |
| HTTP Trigger (Webhook) | Start a flow from an inbound webhook | — | — | Not available |
| Recurring Schedule or Timer | Start a flow on a schedule | — | — | Not available |
| SFTP Trigger | Start a flow when files arrive over SFTP | — | — | Not available |
Apps
App nodes connect a flow to external systems — reading from them, writing back to them, or acting on their data. See App integration nodes for setup and configuration.
| Node | What it does | Inputs → Outputs | Related use cases | Status |
|---|---|---|---|---|
| Outlook: Auth | Authorize the connection to an Outlook mailbox | Connection settings → an authorized session the other Outlook nodes use | Any email-driven process | Available |
| Download Outlook Email | Retrieve an email's contents and attachments | A message reference (from the Outlook trigger) → the email body and its attachment files | Submission Intake | Available |
| Send Outlook Email | Send an email from the connected mailbox | Recipients, subject, and body → a send confirmation | Acknowledgements; routing correspondence | Available |
| Mark Outlook Email as Read | Mark a message as read | A message reference → confirmation | Inbox management within a flow | Available |
| Move Outlook Email | Move a message to another folder | A message reference and a destination folder → confirmation | Inbox management within a flow | Available |
| Move Outlook Email to Folder | Move a message to a named folder | A message reference and a folder name → confirmation | Inbox management within a flow | Available |
| Search for a Guidewire Claim | Look up an existing claim in Guidewire | Search criteria → matching claim record(s) | Claim-to-Policy Comparison; claims processing flows | Available |
| Create a Guidewire Claim | Create a claim record in Guidewire | Claim field values (often from extraction) → the created claim and its identifier | Claims processing flows | Available |
| Update a Guidewire Claim | Update an existing claim in Guidewire | A claim identifier and field values → the updated claim | Claims processing flows | Available |
| Document Management | Connect to a document management system | — | — | Not available |
| Claims Processing | Connect to a claims processing system | — | — | Not available |
| Policy Administration | Connect to a policy administration system | — | — | Not available |
| Cloud Storage | Connect to a cloud storage system | — | — | Not available |
| Customer Management | Connect to a customer management system | — | — | Not available |
Utilities
Utility nodes handle the general-purpose work inside a flow — preparing files, checking values, calling APIs, running custom logic, and routing work to a person. See Utility nodes for setup and configuration.
| Node | What it does | Inputs → Outputs | Related use cases | Status |
|---|---|---|---|---|
| Read Files | Prepare ingested files so later nodes can process them | The ingested files → normalized documents ready for extraction | Most document flows | Available |
| Field Validation | Check extracted field values against rules | Fields and validation rules → a pass/fail result per rule, used to branch or flag | Document validation flows | Available |
| Flag for Human Review | Pause the flow and route the work item to a person for review or approval | The item and the fields to review → the reviewed, approved, or corrected result | Any flow with human-in-the-loop review | Available |
| API Request | Call an external HTTP API | Request configuration (URL, method, headers, body) → the API response, available downstream | Custom integrations | Available |
| HTTP File Upload | Upload a file to an external HTTP endpoint | A file and the endpoint configuration → the upload response | Custom integrations | Available |
| Custom Code Blocks | Run a custom code snippet for logic the built-in nodes don't cover | Data from upstream nodes → the value the code returns | Custom transformations | Available |
Actions
Action nodes change the state of the work item itself. See Action nodes for setup and configuration.
| Node | What it does | Inputs → Outputs | Related use cases | Status |
|---|---|---|---|---|
| Mark Item as Complete | Mark the work item complete to signal that processing has finished | The current item → the item marked complete | The final step of most flows | Available |
Controls
Control nodes shape the path execution takes through a flow — branching and looping. See Control nodes for setup and configuration.
| Node | What it does | Inputs → Outputs | Related use cases | Status |
|---|---|---|---|---|
| Conditional Branching | Split the flow into two paths based on a single condition | A condition over flow data → execution continues down the true or false branch | Routing by document type or value | Available |
| Multiple Condition Branches | Route the flow down one of several branches by matching conditions | A set of conditions → execution continues down the first matching branch | Routing by document type or value | Available |
| For Loop | Repeat a set of nodes once for each item in a collection | A collection → the loop body runs for each element | Processing multiple documents or line items | Available |
| Pause Flow Execution | Pause the flow for a set period before continuing | — | — | Not available |
InsurGPT
InsurGPT nodes apply Bevaya's insurance-trained document intelligence — segmenting page streams, extracting structured data, and generating insights from documents. See InsurGPT nodes for setup and configuration.
| Node | What it does | Inputs → Outputs | Related use cases | Status |
|---|---|---|---|---|
| Split Documents | Segment a multi-document page stream into separate documents | A page stream → the individual documents it contains | Claim Indexing; Claim File Summarization | Available |
| InsurGPT: Custom | Extract structured data (such as FNOL fields) from insurance documents and classify the document type | Documents and the fields or prompt to extract → the extracted fields and a document classification | Submission Intake; document extraction flows | Available |
| Insights | Generate business insights or a summary from documents | Documents and a prompt → the generated insights | Claim File Summarization | Available |
Sticky notes
A Sticky Note is an annotation you place on the builder canvas to leave a comment or label for the people building the flow. It is not a processing step, does not run as part of the flow, and is added from the canvas rather than the Nodes library.
Where to go next
- Configure triggers — set up the trigger that starts your flow.
- App integration nodes — connect to Outlook, Guidewire, and other external systems.
- Utility nodes — prepare files, validate fields, call APIs, and route work to a person.
- Action nodes — change the state of the work item.
- Control nodes — branch and loop within a flow.
- InsurGPT nodes — segment, extract, and summarize documents.
- Create an AI Agent — assemble nodes into a working flow.
- Build AI Agents — the full build lifecycle.