Skip to main content
App Builder agents are only as useful as the tools they can call. This page documents the full toolkit available to every App Builder agent: read access across every connected source, write access into the merchant’s notification and ticket systems, platform-level write actions, and the reasoning tools that let an agent ask Ask Viq questions or pull Vortex Mind reports as part of a workflow. The toolkit is shared, not per-app. Every App Builder agent has access to every tool. The agent picks the minimum tools required for the workflow at plan time. The merchant can swap or add tools manually in the plan view.

​
Read tools, every connector wired into Nerve Centre

App Builder reads from the same data layer that powers Nerve Centre and Vortex Mind. Any connector you have authenticated in your workspace is available as a read tool. The major read surfaces: Any connector that surfaces KPIs on Nerve Centre exposes those KPIs as a read tool. The agent can pull the latest values, query a date range, or filter by dimension. The data shape matches the data shape API so an agent that knows how to read for one connector reads for all of them.

​
Write tools, notification and ticket channels

The write side is where App Builder turns insight into action. Write tools fall into four buckets.

​
Notification channels

​
Ticket and issue creation

​
Platform write actions

These are the highest-trust tools because they modify the merchant’s storefront or marketplace state directly. Approval gates are recommended for production runs.

​
Generic webhook

For any system that exposes an HTTPS endpoint, the Webhook tool lets the agent call it with a structured JSON payload and use the response in subsequent steps. This is the escape hatch when you have a custom internal system the agent should integrate with.

​
Reasoning tools, agent-callable AI surfaces

These are the most distinctive tools in the App Builder toolkit. They let an agent delegate part of its work to other AI surfaces inside the Vortex IQ AI OS.

​
Ask Viq as a tool

The agent can ask Ask Viq a plain-English question and use the answer in the workflow. This is enormously powerful because it means the agent does not need to know the SQL, the Elasticsearch query, the GA4 dimension key, or the Stripe API filter. It just asks Ask Viq, and Ask Viq translates the question into the right query against the right source. Examples of agent-to-Ask-Viq calls:
  • “Ask Ask Viq: what was last week’s revenue compared to the week before?”
  • “Ask Ask Viq: which SKU had the highest week-over-week sales growth?”
  • “Ask Ask Viq: how many orders are stuck in Pending in CloudHub right now?”
  • “Ask Ask Viq: which keywords drove the most conversions on Google Ads in March?”
The agent uses the answer text in the next step (paste into a Slack message, include in an email, attach to a ticket).

​
Vortex Mind reports as triggers and tools

The agent can:
  • Subscribe to a Vortex Mind report. Trigger the app whenever the Daily Revenue Leakage, Decline Recovery, Checkout Conversion Failure, Customer Recovery, Ads Revenue Intel, Paid Traffic Waste, or Payment Performance report produces a new finding above a severity threshold.
  • Run a Vortex Mind report on demand. As part of the workflow, run a report and use the findings in the next step (“run the Daily Revenue Leakage report, take the top 3 findings, post them to #ops”).
  • Read a Vortex Mind finding’s evidence. Pull the underlying evidence (the supporting data points, the affected SKUs, the customer cohorts) and use them in subsequent tool calls.

​
Vortex Memory retrieval

The agent can query Vortex Memory for past run output, archived reports, uploaded files, or knowledge base entries. Common patterns:
  • “Read the last week’s Tuesday Slack summary from Vortex Memory and compute the trend against this week.”
  • “Find the brand DNA document in Vortex Memory and use it as context for drafting the customer email.”
  • “Pull the last incident report from Vortex Memory and link to it in the Jira ticket.”

​
Tool composition, sub-agents

Complex apps decompose into sub-agents. A sub-agent is a scoped agent with its own toolkit and its own prompt, called by the parent agent for a specific sub-task. This pattern was already present in V1 Agent Hub (the Connector Agent orchestrated five sub-agents to discover API credentials, the Ask Viq Agent routed to specialist sub-agents). App Builder makes it explicit and editable. Common sub-agent patterns:
  • Read sub-agent. Pulls and aggregates data from one or more connectors, returns a structured result to the parent.
  • Draft sub-agent. Drafts the human-facing output (email body, Slack message, ticket description) from structured input.
  • Approval sub-agent. Routes a draft to the merchant for approval and waits for the response.
  • Branch sub-agent. Picks one of several next steps based on a condition (severity, threshold, customer cohort).
The agent uses sub-agents when the workflow has multiple distinct cognitive steps, each with its own toolkit. The merchant sees the sub-agent structure in the plan view and can edit each sub-agent independently.

​
Permissions and credential scope

Tools inherit credentials from the connectors authenticated in the workspace. App Builder agents do not get more access than the connector itself has. If your Shopify connector is read-only, the Shopify write tools are not available to App Builder. If your Slack connector is scoped to specific channels, the agent can only post to those channels. You can further restrict an individual app’s access in the plan view: “this app can only read from Shopify orders, not from Shopify customers.” The agent respects per-app scopes at run time.

​
Frequently asked questions

Can the agent call tools I have not authenticated? No. The agent only sees tools whose underlying connectors are authenticated in the workspace. If you want the agent to use Twilio, you connect Twilio first. What if a tool fails mid-run? The agent’s error handler runs. By default, the run logs the error to Vortex Memory, posts an alert to the merchant, and either retries (transient errors), routes to a fallback channel (channel errors), or aborts (data errors). The error policy is editable per app. Can I add a custom tool that App Builder does not have? For most use cases, the generic Webhook tool covers it: the agent calls your HTTPS endpoint with a structured payload. For deeper integration, contact support to request a first-class tool. Are there rate limits on tool calls? Yes, both Vortex IQ-imposed (to prevent runaway loops) and per-connector (Shopify rate limits, Google Ads quota, Slack message limits). The agent respects both. If a rate limit hits, the run is paused and resumed at the next quota refresh. Does the agent have access to all my data by default? The agent has read access to the same data layer the workspace user has. If your role on Vortex IQ has restricted scope (for example, you cannot see finance KPIs), App Builder agents you build inherit that scope. Can two apps share data? Yes, through Vortex Memory. One app writes a result to Memory, another app reads it. This is how multi-step workflows span days or weeks.