Skip to main content
This guide covers the full Agent builder — filters, actions, field mapping with Mentions, trigger configuration, and execution modes.

Key concepts

ConceptDescription
IntegrationA connected external system (Salesforce, Snowflake, CSV, etc.)
FilterA condition that selects records (e.g., “Email is empty”)
ActionWhat BonData does for each matching record (create, update, notify, export)
TaskOne action for one record. 20 matching records = 20 tasks
BondA link between fields in different systems, used for cross-system filters

Create an Agent

1

Open the Agent builder

Go to Agents in the sidebar, then click Create New Agent. Give it a name and optional description.
2

Define filters

Add filter nodes to specify which records the Agent should act on.Filter logic:
  • AND — chain multiple filter nodes sequentially (use Then +)
  • OR — add conditions within the same node (use + Add condition)
  • Field comparison — compare fields across systems (e.g., HubSpot “Company Name” ≠ Salesforce “Account Name”)
Use Filter Insights (button on the filter’s right side) to see how many records match. Requires saving the Agent first.
3

Choose actions

Available actions:
  • Create Record — in any connected system
  • Update Record — modify fields on matching records
  • Slack notification — post to a channel
  • Generate CSV / Google Sheet — export matching data
  • HTTP Request — call an external API
4

Set execution mode

  • Automatic — tasks execute immediately
  • Manual — tasks queue for individual approval
5

Activate

Click Activate to start the Agent on your defined schedule.

Mentions: field mapping between systems

Mentions let you dynamically reference a field from one system when creating or updating records in another.

Basic usage

  1. In an Action config, click Mention
  2. Select the source integration and object
  3. Choose the field
BonData inserts that field’s value into the target action for each matching record. Example: Importing a CSV into Salesforce — mention the CSV’s “Account Name” column in the Salesforce Account “Name” field. Each row creates a new account using that row’s values.

Conditional Mentions

Transform values before writing them to the target system. Example: A CSV has an “OptedOut” column with values 0, 1, or null. Salesforce needs “Yes” or “No”:
  • If OptedOut = 1 → “Yes”
  • Else → “No”

Default values

Set a fallback for when the source field is empty. This prevents blank fields in the target system.

Trigger settings

Access trigger settings from the top-left of the Agent editor.

Trigger behavior

ModeBehavior
Change (default)Tasks only generate for new or changed records that match filters
AlwaysEvery run creates tasks for all matching records, regardless of whether they’ve changed
With Always mode: if 20 accounts match on Monday, a daily alert shows all 20 each day. A new match on Tuesday shows 21 on Tuesday. Use Change mode to see only new matches.

Run frequency

  • Default: every 15 minutes (when Agent is set to automatic)
  • Customizable to any interval, including cron expressions
  • All times are in UTC

Save, Test, Run

  • Save — persist your configuration
  • Test — simulate the Agent without triggering real actions
  • Run — activate the Agent on your schedule
New Agents may need a brief loading period before first execution. BonData must fetch both metadata and data for the entities involved. Check Monitoring & Troubleshooting for details.

FAQ

Yes. Each matching record can trigger multiple tasks (e.g., update a record AND send a Slack notification).
Yes. Filters and Mentions can reference any connected system.
Check the Task logs for affected records, then revert changes in the source system. BonData doesn’t store your full dataset.

Node reference

For a detailed guide to every node type — including Data Normalization, Find Duplicates, AI Enrichment, Web Search, and more — see the Nodes Overview reference.

Best practices

  • Start simple — one filter, one action, one field at a time
  • Use Test mode — validate before activating
  • Review periodically — data fields and business rules change
  • Watch task volumes — unexpected spikes may indicate overly broad filters