Capabilities
Every GetMint MCP tool, grouped by job, with credit and safety annotations
The single /mcp endpoint exposes 22 tools (13 read-only Insights, 9 Content Studio for the brief → article flow), each proxying one REST endpoint. Annotations: read-only never changes state or spends credits; write changes state without spending credits; consumes credits deducts from your balance on success.
Discovery
Find the domain and topic to work on. IDs flow forward: list_domains → domainId, list_topics → topicId.
| Tool | Use it for | |
|---|---|---|
list_domains | List every domain in your org — the entry point that yields the domainId all other tools require. | read-only |
get_domain | Full domain details: brand attributes, aliases, tracked markets/languages, owned sources. | read-only |
list_topics | Every tracked topic attached to a domain (competitors live on topics, not domains). | read-only |
get_topic | A single topic's configuration — prompts, LLM models, markets. | read-only |
get_topic_summary | Quick headline for a topic: your brand's average visibility and competition score over the last 30 days, with a trend. Brand only. | read-only |
Visibility & competition
| Tool | Use it for | |
|---|---|---|
get_topic_visibility | Your brand's visibility score vs each tracked competitor, ranked highest-first, with a per-model breakdown. | read-only |
get_topic_mentions | Share-of-voice leaderboard: brands ranked by how often LLMs name them on the topic, over the last 6 months. | read-only |
get_topic_competition | Your brand's competition score on each of the topic's 10 most recent analyses. | read-only |
Brand perception
| Tool | Use it for | |
|---|---|---|
get_domain_sentiment_aggregated | "How are LLMs describing my brand?" — aggregated sentiment scores, quotes, trend, and a per-model breakdown over a date range. | read-only |
get_domain_alignment_aggregated | How closely LLM depictions match your brand's stated attributes, aggregated over a date range with a per-model breakdown. | read-only |
Citations & sources
| Tool | Use it for | |
|---|---|---|
get_domain_explorer_aggregated | What your brand is known for in LLM answers — top keywords, sources, and entities (compact summary). | read-only |
list_explorer_citations | The exact source URLs LLMs cite, row by row — filter by topic name, cited domain, or model. | read-only |
Shopping
| Tool | Use it for | |
|---|---|---|
get_topic_shopping_overview | Shopping surface only: carousel rate, product presence and position, competitor brands, merchants and top products for one topic, fed by the prompts of the topic assigned to shopping. | read-only |
Brief lifecycle
| Tool | Use it for | |
|---|---|---|
list_briefs | Paginated list of briefs for a domain — start here to discover briefIds. | read-only |
get_brief | Full brief details including research, strategy, and title suggestions. | read-only |
create_brief | Start the research + strategy pipeline for a new brief. Returns a briefId to poll. | consumes credits |
get_brief_progress | Poll pipeline progress — currentStep, totalSteps, progress %, and any errorMessage. | read-only |
Document
| Tool | Use it for | |
|---|---|---|
get_document | The structured outline — metadata, keywords, strategy, and sections with topics/instructions. | read-only |
finalize_document | Lock the outline. Required before generation. | write |
Generation
| Tool | Use it for | |
|---|---|---|
trigger_content_generation | Enqueue article generation for a finalized brief. Returns 202; poll for the result. | consumes credits |
get_generation_status | Poll generation progress — current phase, progress %, and any errors. | read-only |
get_article | Fetch the finished article — markdown, per-section breakdown, metadata, and review notes. | read-only |
Behavior and safety
- Insights tools are read-only and idempotent — call and re-call them freely; no state change, no credits.
- Every tool declares its behavior up front.
tools/listreturns the four standard MCP hints —readOnlyHint,destructiveHint,idempotentHint,openWorldHint— on all 22 tools, so your agent can decide how to treat a call before making it. Two are worth wiring into your retry logic:create_briefandtrigger_content_generationare the only tools that reportidempotentHint: false, so a timed-out call must be polled, never blindly repeated — a retry buys a second brief and a second credit charge. The exception istrigger_content_generationcalled with anidempotencyKey: a retry carrying the same key returns the first call's result and charges once.create_briefis also the only tool withopenWorldHint: true, because its research step reads the live web; everything else stays inside your workspace. - Two tools spend credits.
create_briefdeductsBRIEF_GENERATION;trigger_content_generationdeductsCONTENT_GENERATIONscaled by word count. Confirm both with the user first. - Autopilot chains generation.
create_briefwithautopilotMode: trueauto-selects the first title, initializes the document, and triggers generation — spending generation credits on top of the brief. Leave it off for human review. - Finalize before you generate.
trigger_content_generationrejects an unfinalized brief; the order is alwaysget_document→finalize_document→trigger_content_generation. - Poll, don't spin.
create_briefandtrigger_content_generationrun on workers — pollget_brief_progress/get_generation_statusat 3–5s intervals (polling shares the 200 req/min budget). - Articles can come back flagged.
get_articlereturnsstatus: needs_reviewwhen it fails a compliance check; still readable, but a human should resolve the gaps before publishing.
Known limitations
Some tools trade completeness for a fast, agent-sized response. For full control over windows and payloads, use the REST API — the Cookbook has recipes.
list_explorer_citations is hard-capped to a rolling one-month window (today minus one month → now); the citation set is huge and a wider scan would time out. Page with page/pageSize; for arbitrary ranges, query GET /domains/{id}/explorer/citations over REST.
get_topic_visibility and get_topic_mentions take startDate, endDate and models; omit them for the last 6 months over every model. Each call returns totals for that window, with no time series and no rolling average: for month-by-month figures, call once per month with that month's startDate/endDate. get_topic_visibility reports the period its scores cover in dateRange.
get_topic_visibility also returns a trimmed payload: visibilityScore, scoreVariation (percentage points), competitors[{ name, visibilityScore }] (ranked highest-first), modelBreakdown, topMentions (capped at 10), totalAnswers (answers the scores are computed from). For the full report, use GET /domains/{id}/topics/{topicId}/visibility/aggregated over REST.
get_topic_summary averages your brand's visibility and competition score over a fixed last-30-days window; it returns no competitor figures and no sentiment. get_topic_competition returns your brand's competition score on the topic's 10 most recent analyses — older analyses are out of reach, so it is not a monthly history.
get_domain_explorer_aggregated caps topKeywords and topSources at 25 each and drops webSearchResults. The REST endpoint returns the uncapped lists and raw web-search results.
get_domain_sentiment_aggregated and get_domain_alignment_aggregated take inclusive fromDate/toDate (translated to startDate/endDate); omit for the last 6 months. The topic tools — get_topic_visibility and get_topic_mentions — take startDate/endDate directly, plus models.
A tool rejects any argument it does not declare, and the error names the arguments it does accept. An agent that tries a month argument, or a date on a tool without date filtering such as get_topic_summary, gets an explicit error instead of unfiltered figures presented as filtered.
Inline, the citations block is 93-99% of those two responses, so both tools paginate it: citations.items is one page (default 20, max 100), with totalItems/hasMore for the remainder. Move through it with citationsPage, or pass citationsPageSize: 0 for the scores alone. The REST endpoints still return every row.