The agents shipped with this release, with their settings and what each may change. Costs shown in the admin are estimates; see Invocations.

Two columns need explaining:

  • Triggers: Entity event agents run when a record is created or updated, one record per run. Scheduled agents run on a timetable and look back over a period.
  • Can change: the fields the agent is allowed to write, and only when its install is in an action mode that permits writing. An agent with nothing proposes and reports, but never edits content.

Content

Blog Post Proofreader

Proofreads a single blog post for spelling, grammar and clarity, and records each suggested correction as a recommendation against the post.

  • Triggers: Entity event · Modes: Read-only, Accept to action
  • Reads: blogs, blog posts
  • Can change: blog post name and body text
  • Parameters: variety (default us), spelling and usage variety to enforce; includeStyleSuggestions (default on), also suggest clarity and concision improvements, not just errors

Content Freshness Audit

Flags pages and blog posts that look stale, not updated in a long time, or containing dated references, and recommends a review or refresh.

  • Triggers: Scheduled, Entity event · Modes: Read-only, Accept to action
  • Reads: pages, blog posts
  • Can change: nothing
  • Parameters: staleAfterMonths (default 12), content not updated within this many months is a freshness candidate; flagDatedReferences (default on), also flag content referencing a past year or an outdated fact, even if recently edited

FAQ Suggester

Reads a page and proposes frequently-asked-question entries it would answer, recording each suggested question and answer as a recommendation.

  • Triggers: Entity event · Modes: Read-only, Accept to action
  • Reads: pages, FAQ pages, FAQ questions
  • Can change: nothing
  • Parameters: maxQuestions (default 5), cap on entries proposed from one page; answerLength (default short)

Product Description Enhancer

Reviews a single product's description for clarity, completeness and persuasive quality, and proposes an improved version.

  • Triggers: Entity event · Modes: Read-only, Accept to action
  • Reads: products, product categories
  • Can change: product description text
  • Parameters: tone (default neutral); maxWords (default 120), soft upper bound on the proposed description

SEO

Page Meta Audit

Audits a page's SEO title and meta description against length and quality targets, then proposes improved values.

  • Triggers: Entity event, Scheduled · Modes: Read-only, Accept to action
  • Reads: pages
  • Can change: page SEO title and meta description
  • Parameters: targetTitleLength (default 60); targetDescriptionLength (default 155); brandSuffix (default empty), optional brand string appended to titles, e.g. " | Acme"; tone (default neutral)

Heading Structure Audit

Audits a page's heading hierarchy (single H1, no skipped levels, descriptive headings) and recommends fixes where the structure hurts accessibility or SEO.

  • Triggers: Entity event, Scheduled · Modes: Read-only, Accept to action
  • Reads: pages
  • Can change: nothing
  • Parameters: requireSingleH1 (default on), flag pages with zero or more than one H1

Image Alt Text Audit

Finds images on a page missing alt text or with unhelpful alt text, and proposes concise descriptive alternatives.

  • Triggers: Entity event, Scheduled · Modes: Read-only, Accept to action
  • Reads: pages, images
  • Can change: image alt text
  • Parameters: maxAltLength (default 125); treatDecorativeAsEmpty (default on), propose empty alt text for purely decorative images rather than a description

Internal Link Opportunities

Reads a page and suggests relevant internal links to other pages or blog posts on the same site, proposing each with anchor text and target.

  • Triggers: Entity event, Scheduled · Modes: Read-only, Accept to action
  • Reads: pages, blog posts
  • Can change: nothing
  • Parameters: maxSuggestions (default 5); includeBlogPosts (default on), consider blog posts as link targets, not just pages

Ecommerce

Every ecommerce agent is read-only and cannot be given write access.

Sales Report

On a schedule, summarises recent sales into a concise report. When the install has an outbound chat channel configured, the report is delivered there. This is how a daily sales summary is posted to Slack.

  • Triggers: Scheduled · Modes: Read-only
  • Reads: orders, invoices, invoice transactions
  • Parameters: periodDays (default 1), how far back the report covers; topProductsCount (default 5); currencyLabel (default $); tone (default neutral)

Abandoned Cart Insights

On a schedule, reviews recently abandoned carts and summarises patterns (common drop-off products, value bands and likely friction) into a short report for the store team.

  • Triggers: Scheduled · Modes: Read-only
  • Reads: orders, schema
  • Parameters: periodDays (default 7); currencyLabel (default $)

Order Anomaly Triage

Inspects a single order for anomalies (unusual totals, mismatched details, rapid repeat orders) and produces a short plain-language assessment. Triage only; for decisions and escalation use Order Fraud & Refund Decision.

  • Triggers: Entity event · Modes: Read-only
  • Reads: orders, people
  • Parameters: largeOrderAmount (default 500), totals at or above this (store base currency) are called out as notable

Order Fraud & Refund Decision

Assesses a single order for fraud risk and refund legitimacy, escalates ambiguous or high-value cases to a human, and records its conclusion as a recommendation. This is the agent most likely to escalate; configure its escalation target carefully before enabling it.

  • Triggers: Entity event · Modes: Read-only
  • Reads: orders, invoices, invoice transactions, people
  • Parameters: autoDecideUnderAmount (default 100), orders at or below this total may be decided without escalating; escalateOnAddressMismatch (default on), escalate when billing and shipping differ materially, regardless of total; riskTolerance (default balanced): cautious escalates more, lenient escalates less