Capabilities and limitations

This page is a catalog of what Scout can do today and where they intentionally stop. Scout is built for safe, in-project iteration: every edit lands as a draft, nothing is published, and nothing leaves your project until you take action.

Capabilities and limits change frequently — check the release notes for the latest.

Scout picks up a few things from your environment automatically, so you can talk in shorthand and they still know what you mean:

  • The currently open entity: the composition, entry, or pattern you're viewing is the default target when you don't name one. Saying "translate this to German" or "add an SEO description" operates on what's in front of you. To target something else, name it explicitly — for example, "edit the homepage instead."
  • The current locale: Scout follows the locale you have selected in the editor and operates there unless you ask for another.
  • The active release context: if you're working inside a release, Scout edits in that release and links results to the right pending items. To edit a different release, open it first so it becomes the active context.
  • Attachments: images, .txt, .md (see Attachments).
  • URL fetching: paste a URL into the prompt box and Scout reads the page (see URL attachments).
  • Skills: your team's reusable instructions, loaded automatically or on demand (see Skills).
  • AI guidance: context and constraints on components, content types, parameters, and fields (see AI guidance).
  • External MCP servers: third-party tools you have connected (see Connect MCP servers).

Scout's capabilities span content modeling, day-to-day editing, batch work, search, releases, localization, optimization, SEO, and platform guidance.

Content modeling#

  • Create and update component definitions: introduce new components, edit parameters, and group fields.
  • Create and update content type definitions, including auto-generated data types so the new content type is usable end-to-end.
  • Suggest improvements to naming, structure, reusability, and field grouping.
  • Explain field and parameter types and recommend best practices.
  • Create compositions and attach them to project map nodes.
  • Create new project map nodes when you need a route for a new composition.
  • Create entries from content types or entry patterns.
  • Edit compositions, entries, and patterns: fields, parameters, slots.
  • Add component instances to compositions from natural language.
  • Duplicate or move component instances within a composition.
  • Add, remove, or reorder components in slots.
  • Explain which components are allowed in a given slot so you know what can go where.
  • Set data resources to draft content while you iterate.
  • Find entries by meaning when picking a value for a reference field.
  • Edit pattern instances: override pattern parameters and data resources, with reliable deep links to entry patterns.
  • Edit pattern definitions: manage slot sections, and configure data resources on the pattern itself (not just on individual pattern instances).
  • Edit content inside an existing edition, with locale filtering.
  • Batch edits across many entities in a single prompt: ask Scout to "translate all blog entries to German" or "add an SEO description to every product page" and they will work through the set in a single pass. For multi-entity changes, Review changes mode and the Pending edits dialog let you inspect each entity before applying.
  • Semantic search across compositions, entries, assets, and patterns: ask in natural language — for example, "entries about coffee machines" or "images about food" — and Scout ranks results by meaning.
  • Find specific text inside compositions, entries, and patterns: useful for things like "find every entry mentioning our old product name" or "which compositions reference this URL?".
  • Reference other content in prompts: for example, ask Scout to create a new landing page based on an existing one.
  • Search assets by label: asset labels are indexed, so Scout can find images and files by what they depict.
  • Use structured metadata filters as you always have.
  • Search across, list, and inspect releases: open, locked, queued, launching, launched.
  • Edit content inside the active release, with results linked to the right pending items. See How Scout works with your context for how the active release is determined.
  • List and explain locales in your project, and enable or disable them on a composition, entry, or pattern.
  • Switch the editor's current locale so you can work in the language you want.
  • Translate localizable content into one or more target locales at once.
  • Configure personalizations with audience-based variants on any component.
  • Set up A/B tests: variants, traffic splits, default variant.
  • View A/B test and personalization performance on demand: ask Scout for current statistics on a running test, a personalization's variants, or aggregated optimization data, and they will pull the numbers.
  • Declare winners when a test is ready.
  • Manage enrichments on compositions or individual enrichment fields.
  • Full CRUD on optimization entities — signals, quirks, intents, audiences, aggregates, and enrichments. Common cases such as query-string signals can be created end-to-end without follow-up questions.
  • Run SEO analysis, fix issues, and set or update meta tags.
  • Generate Open Graph previews, surfaced near the top of the results so you don't have to dig for them.
  • Suggest accessibility and performance improvements.
  • Suggest SEO-friendly parameter values, and reconcile competing title sources (for example, page title versus SEO title) so you don't end up with two titles fighting each other.
  • Provide guidance on URL structure, including metadata and slugs.
  • View page traffic and conversion data when the Uniform Insights integration is set up.
  • Answer questions about Uniform concepts, features, and best practices.
  • Link to relevant documentation.
  • Help you navigate listing pages, the visual editor, structure panel, preview, and property panels.

All Scout edits land as drafts. Nothing is published, no workflows are advanced, and nothing leaves your project until you take action. Beyond that default, Scout is intentionally constrained in a few places. The limitations fall into three groups:

  • Security and human control: actions that affect what's live, how users reach pages, who has access, or how the project is configured stay manual. This includes publishing, administration tasks, and project configuration.
  • Outside Scout's scope today: some configurations need human supervision, or aren't yet a good fit for an AI assistant.
  • Legacy features: deprecated parameter and field types, and legacy AI prompts, aren't supported.

Actions that publish content, change URL routing, manage workflows or releases, grant access, administer the team, or change project configuration stay with a human:

  • Cannot publish content; all work stays in draft until you publish it yourself.
  • Cannot edit the project map beyond creating new nodes. Changes there can immediately break live URLs, so they're left to a human.
  • Cannot manage workflows or transition content between workflow stages.
  • Cannot change release settings such as locking or scheduling. Scout can edit content inside the release you currently have active, but will not write into a different release than that active context.
  • Cannot manage integrations.
  • Cannot manage project settings, team settings, user permissions, or roles.
  • Cannot handle billing, licensing, or sales questions.

Some features need a person in the loop, or aren't yet a good fit for an AI assistant due to complexity:

  • Cannot create, upload, edit, or delete assets, or create new assets from uploaded files. When you upload an asset yourself, Scout-powered helpers can still suggest a title, description, labels, and focal point as part of the upload flow.
  • Cannot edit Block type fields and parameters.
  • Cannot edit the criteria that decide when a conditional value applies. Scout can still edit the conditional value itself, and can enable allowConditionalValues on a field or parameter definition.
  • Cannot set visibility rules on component instances.
  • Cannot create or manage Loop component instances.
  • Cannot configure external data connections from data sources and data types.
  • Cannot create or manage data resources or work with dynamic tokens.
  • Cannot convert an existing component or composition into a pattern, or unlink a component pattern to make a local copy. Both operations are left to a human.
  • Cannot create new editions of compositions or entries. Scout can edit content within an existing edition, but the edition itself must be created by a person.
  • Cannot monitor results after creating an A/B test or a personalization. Use the analytics surfaces in Uniform to track performance over time, and come back to Scout when you're ready to declare a winner.
Last modified
Next
Skills