## The problem

Fully automated agent pipelines occasionally produce low-quality output. Teams need a lightweight review step without rebuilding the whole pipeline.

## How SuperTrove helps

```mermaid
flowchart TD
    A[Generation Agent] -->|bulk_add_items| B[SuperTrove: Drafts]
    B --> C[Human Reviews in UI]
    C -->|update_item status=approved| B
    C -->|update_item status=rejected| B
    D[Publishing Agent] -->|query_list status=approved| B
    D --> E[Publish to CMS / Shopify / Email]

Item shape

{
  "title": "Blog Post: Hydration Tips",
  "content": "...",
  "status": "draft",
  "generated_at": "2026-05-25",
  "reviewed_by": null
}

Why SuperTrove fits here

• The UI renders markdown and JSON cleanly for human review
• query_list filters by any field — agents only see what they need
• No custom database, no separate CMS — the review step lives where the data lives

## The problem

Coordinating work between multiple agents requires a shared, durable store that survives context resets. Email threads and chat messages don't cut it.

## How SuperTrove helps

```mermaid
flowchart LR
    A[Orchestrator Agent] -->|add_item status=pending| Q[SuperTrove: Task Queue]
    B[Worker Agent] -->|query_list status=pending| Q
    B -->|update_item status=done| Q
    C[Monitor Agent] -->|query_list status=failed| Q

Task item shape

{
  "task": "enrich_product",
  "payload": { "sku": "HFS-50ML" },
  "status": "pending",
  "assigned_to": null,
  "created_at": "2026-05-25"
}

Agent prompt

Poll your task queue at the start of every session:
query_list({ list_id: "<queue-id>", filter: { status: "pending" } })

After completing a task:
update_item({ list_id: "<queue-id>", item_id: "<id>", data: { ...task, status: "done", completed_at: "<date>" } })

## The problem

Raw Shopify product exports lack structured metadata for SEO, personalisation, and localisation. Manually writing this for 500+ SKUs is not viable.

## How SuperTrove helps

```mermaid
flowchart TD
    A([Shopify Export]) --> B[Research Agent]
    B --> C[Enrichment Agent]
    C -->|bulk_add_items| D[SuperTrove: Enriched Products]
    D --> E[Human Review]
    E -->|approved| F[Push back to Shopify]

Agent prompt

You are a product enrichment agent. For each product in the batch:
1. Research the ingredients and target audience
2. Write an SEO title (max 70 chars), meta description (max 160 chars), and PDP summary (max 300 words)
3. Store each result in SuperTrove using add_item with fields: sku, seo_title, meta_description, pdp_summary, tags
4. Flag any product where you're uncertain by adding uncertain: true

Result

The team reviews flagged items in SuperTrove, approves the rest, and a second agent pushes approved items back to Shopify metafields.

## The problem

Teams build internal wikis that agents can't easily read. PDFs, Notion pages, and Confluence docs require manual parsing.

## How SuperTrove helps

```mermaid
flowchart LR
    A[Team] -->|write structured items| B[SuperTrove Public List]
    B -->|visibility: linked| C[Public URL]
    D([Any Agent]) -->|get_public_list slug| B
    D --> E[Uses knowledge in response]

Setup (one-time)

1. Create a list, write your knowledge as structured items
2. Set visibility to Linked in the Visibility panel
3. Copy the slug from the public URL
4. Share with your team: get_public_list({ slug: "your-slug" })

Agent prompt for end users

Before answering questions about [topic], always pull the latest knowledge:
get_public_list({ slug: "<your-slug>" })
Use the returned items as your source of truth.

Result

One source of truth. Updated by the team, instantly available to every agent.

## The problem

AI agents have no persistent memory between sessions. When a research agent discovers something useful, that knowledge disappears when the context window closes.

## How SuperTrove helps

```mermaid
sequenceDiagram
    participant A1 as Agent 1 (Research)
    participant A2 as Agent 2 (Writer)
    participant ST as SuperTrove
    A1->>ST: add_item { finding, source, confidence }
    A1->>ST: add_item { finding, source, confidence }
    A2->>ST: query_list { filter: { confidence: "high" } }
    ST-->>A2: Relevant findings
    A2->>A2: Write content using findings

Agent prompt

After every research session, store your findings in SuperTrove:
add_item({
  list_id: "<research-list-id>",
  data: {
    topic: "<what you researched>",
    finding: "<key insight>",
    source: "<URL or document>",
    confidence: "high" | "medium" | "low",
    date: "<today's date>"
  }
})

Result

The next agent session starts by querying SuperTrove for prior research — no repeated work, no lost context.