Markup AI

List the checks this organization can run and the configurations available to apply. Use this to see which agents exist (name, one-line description, category), and the personas, brand voice profiles, and style guides available for markupai_run_agents config. Not needed before markupai_review — it applies configuration automatically. Returns a compact catalog: agents grouped by category and named config objects. No JSON schemas — inputs are always plain text.

markupai_capabilities_and_configs

Review content against the organization's configured quality checks in one call. Use this when the user wants content checked, reviewed, or cleaned up before publishing — brand voice drift, unapproved terminology, style-guide violations, hard-to-read sentences, AI-sounding phrasing, unsupported or risky claims, missing sourcing, and AI-search readiness (GEO/AEO: snippet readiness, answer-first headings, key takeaways, freshness). With no goal, the standard review set runs: style_agent, brand_voice, terminology, sentence_clarity, ai_voice_detector, generic_claims, source_authority. Pass goal to state what the user is trying to achieve, in plain language — the review composes the right set of checks for it and requests goal filtering from supported checks. Returns what was checked (agents run, how they were selected, profiles and guides applied, goal interpretation, content size), any configuration warnings, and the flagged issues per agent with locations, explanations, and suggested corrections — everything needed to apply fixes. When the style check runs, its result also includes quality scores.

markupai_review

Run an explicit, fixed set of agents. Use this only when the user or an integration names specific checks to run, or a predefined integration needs a deterministic set. For 'review this', 'clean this up', or any general quality or brand ask, use markupai_review instead — it selects the right checks automatically (from the goal when one is given). The organization's configured profiles and guides still apply unless overridden in config. goal here requests filtering from supported agents only; it never changes or reorders the agent set, and warnings report any filtering exemptions or failures. Valid agent names (for when specific checks are named) — persona (Simulates how a defined ICP persona reads content and flags where it would lose, confuse, or fail to resonate with that reader); brand_voice (Evaluates content against a brand voice profile); terminology (Checks text against organization terminology rules and suggests replacements); generic_claims (Identifies verifiable claims and promissory language in text and returns structured issues); source_authority (Identifies content lacking credibility and trust signals that AI models look for when deciding which sources to cite); freshness (Detects stale content, outdated references, and missing freshness signals using regex patterns and spaCy NER); link_integrity (Detects dead, mismatched, decayed, and off-topic hyperlinks and urls); focus_agent (Finds indirect body sentences in document sections); answer_authority (Detects Q&A header and answer authority issues); sentence_clarity (Identifies difficult sentences and suggests clearer rewrites); ai_voice_detector (Detects AI-generated content patterns using weighted issue detection across multiple subagents); style_agent (Analyzes text for style guideline compliance, checking for issues like passive voice, word choice, and consistency); snippet_readiness (Evaluates whether each section's opening snippet can stand alone if extracted by an AI system, with a concrete fix hint when it cannot); key_takeaways (Ensures that a short list of key takeaways from the text is present near the top of the document). Returns one result entry per agent that completed, inside the same evidence envelope as markupai_review; agents that fail appear as agent_failed warnings, not results entries.

markupai_run_agents