Skip to main content
App Builder turns a sentence into a running ecommerce app. This page walks the full pipeline from the merchant’s prompt through to a scheduled production app, covers each stage in detail, and explains the architectural shift from the V1 Agent Hub model that App Builder replaces.

The five-stage pipeline

1. Prompt   2. Plan   3. Tools   4. Sandbox   5. Production
   (you)    (agent)   (agent)    (agent+you)  (you+agent)
Each stage has a clear hand-off and a clear approval gate. The merchant is in control at every stage; the agent never silently ships something the merchant has not seen and approved.

Stage 1, prompt

The merchant types a description of what the app should do. Examples that work well:
  • “Every Tuesday at 9am, summarise last week’s Shopify sales by SKU and post the top 10 to the #ops Slack channel.”
  • “When a customer’s third order ships, send them a 15% off code from Klaviyo.”
  • “When checkout conversion drops 10% week over week on BigCommerce, open a Jira ticket assigned to the engineering lead.”
Examples that need refinement (the agent will ask follow-up questions):
  • “Make me a sales report.” (too vague: which platform, which period, which metrics, where to send it)
  • “Help with abandoned carts.” (which signal triggers it, which channel, what message, what discount)
The agent parses the description and either asks a clarifying question or moves to the plan stage.

Stage 2, plan

The agent generates a structured plan from the prompt. The plan has six parts:
Plan partWhat it captures
TriggerWhat starts the app: a schedule, a webhook, a Vortex Mind finding, a Nerve Centre threshold, a manual run.
Data sourcesWhich connected sources the app needs to read from (Shopify orders, BigCommerce products, Google Analytics sessions, Datadog metrics, Vortex Mind reports).
LogicThe conditional, computational, or filtering logic the agent will apply to the data.
OutputWhat the app produces: a Slack message, an email, a Jira ticket, a webhook call, a CSV export, an Ask Viq summary.
ScheduleWhen and how often the app runs.
Approval gatesWhether each run needs human approval before sending output, or runs autonomously.
The merchant reviews the plan, asks for edits, or approves it. Edits are conversational. (“Use the last 30 days instead of last 7.” “Add the conversion rate column.” “Send to a different channel.”) The agent re-generates the plan after each edit.

Stage 3, tools

Once the plan is approved, the agent picks the specific tools it needs from the App Builder toolkit. Tools fall into three buckets:
  1. Read tools. Connector adapters that fetch data from any source connected to Nerve Centre. The agent has read access to the same data layer that powers the dashboard. See Tools and integrations for the full read surface.
  2. Write tools. Notification channels (Slack, Microsoft Teams, email, SMS via Twilio), ticket systems (Jira, Linear, internal Kanban via Actions), platform-level write actions (post a Shopify product update, change a BigCommerce price, route an order in CloudHub).
  3. Reasoning tools. Ask Viq for plain-English Q&A on the merchant’s data, Vortex Mind report runners for triggering or pulling diagnostic findings, Vortex Memory for retrieving past run output.
The agent selects the minimum tools required for the workflow. The merchant sees the chosen tools in the plan view and can swap or add tools manually.

Stage 4, sandbox

The approved app deploys to a sandbox runtime. The sandbox is functionally identical to production: same code, same tools, same credentials. The difference is that sandbox runs route their write actions to a safe channel by default. A Slack message goes to a sandbox channel, an email goes to the merchant’s inbox, a Jira ticket goes to a sandbox project. This lets the merchant verify the actual output without polluting production systems. The merchant runs a test execution from the sandbox view. The output renders inline. If the output is wrong, the merchant goes back to the prompt or plan stage and edits. If the output is right, the merchant promotes the app to production.

Stage 5, production

The promoted app runs on its scheduled trigger or live event. Each run is logged to Vortex Memory with the input data, the agent’s reasoning trace, the tool calls it made, and the output it produced. Production runs are visible in Deployment and monitoring, where the merchant can pause the app, edit it, view error history, and roll back to a previous version.

The agent under the hood

App Builder agents are built on the same agent framework that powered V1 Agent Hub agents (the framework that runs ConnectorAgent, AskVIQ, Google Ads Performance Agent, and the rest of the legacy catalogue). The framework provides:
  • Tool calling. The agent picks tools from a registered toolkit and calls them with structured arguments.
  • Sub-agent delegation. Complex apps can break themselves into sub-agents (an “ask Ask Viq” sub-agent, a “post to Slack” sub-agent, a “filter the result” sub-agent), each with its own scoped toolkit.
  • Reasoning traces. Every decision the agent makes is logged so the merchant can see why the agent chose a particular tool or wrote a particular query.
  • Approval gates. Tool calls that have side effects (write actions) can be marked as gated, requiring human approval per run.
What is new in App Builder vs V1 Agent Hub:
CapabilityV1 Agent HubV2 App Builder
DiscoverabilityBrowse a catalogue of pre-built agentsDescribe what you want; agent assembles it
CustomisationLimited; engineering required for changesPlain-English edits at any stage
Tool availabilityPer-agent fixed toolkitFull toolkit available to every agent, agent picks
Cross-platformEach recipe was platform-boundApps span multiple platforms natively
ScheduleManual run in most casesScheduled, event-triggered, or on-demand
SandboxNot first-classFirst-class staging surface before production
Deployment surfaceRun from the catalogue UIOwned, editable production app
The legacy recipes still serve a purpose: they are starter templates. A merchant who wants a Shopify product SEO workflow can fork the V1 Product SEO Agent recipe as a starting point and edit it in App Builder. The starter saves the prompt-engineering time. The flexibility of App Builder lets the merchant take it anywhere from there.

The architectural shift, in one paragraph

V1 Agent Hub asked: “what pre-built agent do you want?” The catalogue dictated what was possible. The merchant was a consumer of the catalogue. V2 App Builder asks: “what do you want your app to do?” The merchant is the author. The agent is the implementer. The toolkit is the universe of possible actions, not a list of fixed recipes. This is the AI OS pattern at work: the platform is no longer a fixed feature set; it is an operating system on which the merchant can build.

Frequently asked questions

How long does the prompt to production loop take? For a simple app (one source, one schedule, one notification channel), under five minutes from the first prompt to a running production app. For a complex app (multiple sources, conditional logic, approval gates), 15 to 30 minutes including a few rounds of plan refinement. What happens if the agent picks the wrong tools? The merchant sees the chosen tools in the plan and can swap or remove any of them. If the agent has misunderstood the prompt entirely, the merchant edits the prompt and the agent regenerates the plan. There is no scenario where the wrong tools end up in production without the merchant approving the plan that contains them. Can the agent build apps that span multiple platforms? Yes. A single App Builder app can read from Shopify, cross-reference with Amazon inventory in CloudHub, ask Ask Viq for the conversion rate trend, and post the combined output to Slack. The agent picks tools from any connected source, regardless of platform. What happens if a connected source loses connection mid-run? The agent detects the failure and routes the error through the merchant-defined error handler (retry, alert, skip the run, route to a fallback channel). Default behaviour is to log the error to Vortex Memory and post an alert to the merchant. See Deployment and monitoring for the full error model. Can I see what the agent did on each run? Yes. Every production run produces a reasoning trace stored in Vortex Memory: the input data, the tools called, the agent’s chain of thought between calls, the output produced. You can replay any run to debug or audit. Does App Builder use my data to train its agents? No. App Builder agents run against your data per-run only. There is no training step that ingests your data into a shared model. Your prompts, plans, and run outputs stay scoped to your workspace.