How to connect Brand24 MCP with DeepSeek Harness

Returns events (anomalies, mention spikes, viral moments) from ALL projects in the account, mixed chronologically. Use this for "summary" or "overview" requests across everything the user monitors. For anomalies on a SINGLE named brand, call brand24_project_events instead. Events returned by this tool span every project - attribute each event strictly by its own project_name field and treat events from other projects as unrelated to the user's question. "summary my projects" / "overview" / "what's happening" → this tool. "list my projects" → brand24_get_projects. "anomalies for [brand X]" / "events for Nike" → brand24_project_events. Args: - response_format (string, optional): Output format - "json" or "markdown" (default: "markdown") Returns: { "scope": "all_projects", "note": "Events span ALL projects; use project_name to attribute each row.", "events": [ { "anomaly_date": string, "project_id": number, "project_name": string, "description": string, "peak_mentions": number, // Total mentions on the event day - ONE value per event, NOT per anomaly type "peak_reach": number, // Total reach on the event day - ONE value per event, NOT per anomaly type "is_hashtag_anomaly": boolean, // Categorical label for the same event "is_user_discussion_anomaly": boolean, // Categorical label for the same event "is_emotional_anomaly": boolean, // Categorical label for the same event "is_news_anomaly": boolean, // Categorical label for the same event "is_social_media_anomaly": boolean, // Categorical label for the same event "is_non_social_media_anomaly": boolean // Categorical label for the same event } ] } Important: peak_mentions and peak_reach are event-day totals for ONE event. The is_*_anomaly flags are categorical labels that may all be true for the same event. Keep one row per event even when multiple flags are true - splitting by anomaly type duplicates the metrics. Usage Examples: - Use when: "Summary of my projects" → params: {} - Use when: "What's happening with my projects?" → params: {} - Use when: "Overview of my projects" → params: {} - Use when: "Give me a summary across all my brands" → params: {} - Use when: "What significant events happened across all my Brand24 projects?" → params: {} - Use when: "Show me all recent anomalies across monitored brands" → params: {} - Use when: "Any viral moments in any of my projects recently?" → params: {} - Use when: "Overview of all important brand events this month" → params: {} - Don't use when: Only want events for ONE project → use brand24_project_eventsThe tool returns a complete markdown answer (header, key/value pairs, and an events table grouped by project). RELAY that markdown to the user verbatim — do NOT rephrase, reorder, or summarize the table into prose. You MAY prepend at most ONE short sentence of net-new insight (a 'why', anomaly, or comparison not in the table) ONLY if you genuinely have something to add. Never replace the table with a textual summary.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in the specific project name from the most notable event). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"). If the same follow-up has been asked and not acted on, pick a different one rather than omitting. How to pick (read this — pick by the dominant event in the list): - One project clearly drives the biggest / most recent event → brand24_project_events on THAT project to dig in. - The biggest event is hashtag-/social-/topic-driven (is_hashtag_anomaly OR is_social_media_anomaly true on the top event) → brand24_project_events still works, but consider that the next layer the user will want is topics — keep it to project_events here, the per-project tool will steer toward topics. - The biggest event is news-led (is_news_anomaly OR is_non_social_media_anomaly true) → brand24_get_mentions on that project around the event date. - Volume / sentiment context matters more than the spike itself → brand24_project_stats for that project. - User has been collecting multiple datapoints / mentioned a deliverable → brand24_pdf_report_builder. Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_project_events — drill into the events of ONE specific project from the list (using its project_name). Sample question: "Want me to drill into the events for [project name]?" - brand24_project_stats — daily mention / sentiment / reach context for the project tied to an event. Sample question: "Want me to pull the daily stats for [project name] around that spike?" - brand24_get_mentions — read the actual mentions around an event date. Sample question: "Want me to fetch the actual mentions around that event for [project name]?" - brand24_pdf_report_builder — polished, downloadable PDF report covering one of the affected projects. Sample question: "Want me to build a downloadable PDF report covering this?"

Goal: turn this data into actionable marketing and business insights — what is working, what is at risk, who is driving the conversation, and where to focus next. Prefer concrete recommendations over raw numbers. List all Brand24 projects in the account, enriched with 30-day mention counts. Use this tool for: - Any request to see, list, show or browse the user's projects ("show me my projects", "list my projects", "what projects do I have", etc.) - Any internal step where another tool needs a numeric project_id — pick the project_id off the returned list. Returns: Array of: { "project_id": number, "project_name": string, "totalMentions": number | null } Rendering rules: 1. The tool already returns the project list AND a follow-up sentence as markdown when the user calls it conversationally. RELAY that markdown to the user verbatim — do NOT rephrase, reorder, or replace the closing sentence with your own (e.g. do NOT write "Tell me which one you want to open"). 2. Do NOT repeat numbers, dates, rankings, or top-N items from the list beyond what the tool already produced. 3. You MAY prepend at most one short sentence of net-new insight (a 'why', anomaly, or comparison not on screen) ONLY if you genuinely have something to add. Follow-up rule (controls whether to keep the closing sentence "Which project do you want to explore? I can show stats, events, topics, sources, mentions, or build a PDF report."): - KEEP it when this is the user's FIRST message in the conversation AND they only asked to see/list projects without specifying a next action. - DROP it (strip the closing sentence from the relayed markdown) when the user has already asked a specific follow-up question, is mid-workflow, has already been shown this follow-up earlier in the conversation, or has already chosen a project or action. Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in the project name). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"), OR if the tool-emitted markdown already ends with the canonical "Which project do you want to explore?" sentence (per the KEEP rule above) — in that case do NOT add a second follow-up. If the same follow-up has already been asked and not acted on, pick a different one rather than omitting. How to pick (read this — match the question to what the user seems to be doing): - User's first ask was "show / list my projects" with no follow-up direction → the tool-emitted closing sentence already covers this; do NOT add a second follow-up. - User mentioned anomalies / spikes / "what's happening" earlier → brand24_project_events on the most-mentions project. - User mentioned trends / sentiment / reach earlier → brand24_project_stats on the most-mentions project. - User mentioned PDF / report / export / summary deliverable → brand24_pdf_report_builder. - User mentioned reading / browsing actual mentions → brand24_get_mentions. - User asked something open-ended like "what's interesting" → brand24_topics_overview on the most-mentions project (themes are the highest-signal first dive). Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_project_stats — daily mentions / sentiment / reach trends for one project. Sample question: "Want me to pull the daily mentions, sentiment and reach for one of these projects?" - brand24_project_events — anomalies / spikes for one specific project. Sample question: "Want me to check for spikes or anomalies on one of these projects?" - brand24_topics_overview — main themes / topics being discussed in one project. Sample question: "Want me to break down the main topics being discussed in one of these?" - brand24_get_mentions — the actual mentions feed (with filters) for one project. Sample question: "Want me to show the actual mentions feed for one of these projects?" - brand24_pdf_report_builder — polished, downloadable PDF report for a project. Sample question: "Want me to build a downloadable PDF report?"

Quickly compare popularity of brands or other entities (people names, sports clubs, etc.) on the Internet. This tool provides a quick comparison but is NOT as detailed or accurate as Brand24 Project Stats data. IMPORTANT: This tool should be used when you want to quickly compare how two or more brands or entities are popular in the Internet relative to each other. The data is less detailed and accurate than Brand24 Project Stats. When to use: - Quick comparison of popularity between brands or other entities (people names, sports clubs, anything discussed on the Internet) - When you need a fast overview without detailed metrics - As a backup option when there's no Brand24 project with the given brand/entity name in your account When NOT to use: - When you need detailed, accurate data → use brand24_project_stats instead - When you have a Brand24 project for the brand/entity → use brand24_project_stats for better accuracy - When you need historical trends, sentiment analysis, or event tracking → use brand24_project_stats or brand24_project_events Args: - brandNames (string[]): Array of brand/entity names to compare (1-10 items) Examples: ["Nike", "Adidas", "Puma"], ["Apple", "Microsoft", "Google"], ["Lionel Messi", "Cristiano Ronaldo"], ["Real Madrid", "Barcelona"] - countryCode (string, optional): ISO 3166-1 alpha-2 country code (applies to all brands) Examples: "US", "GB", "DE", "PL" - response_format (string, optional): Output format - "json" or "markdown" (default: "markdown") Returns: { "totalBrands": number, "successCount": number, "failureCount": number, "countryCode"?: string, // Present if countryCode was provided "results": [ { "brandName": string, "status": "success" | "error", "brandSize"?: number, // Present if status is "success" "error"?: string // Present if status is "error" } ] } Usage Examples: - Use when: "Quickly compare popularity of Nike, Adidas, and Puma" → params: { brandNames: ["Nike", "Adidas", "Puma"] } - Use when: "How popular are Apple and Microsoft compared to each other?" → params: { brandNames: ["Apple", "Microsoft"] } - Use when: "Compare popularity of two football players" → params: { brandNames: ["Lionel Messi", "Cristiano Ronaldo"] } - Use when: "Quick comparison of sports clubs" → params: { brandNames: ["Real Madrid", "Barcelona", "Manchester United"] } - Use when: "Compare popularity of brands in a specific country" → params: { brandNames: ["Apple", "Microsoft", "Google"], countryCode: "US" } - Don't use when: Need detailed data about project → use brand24_project_stats - Don't use when: More than 10 entities → split into multiple requestsThe tool returns a complete markdown answer (header, key/value pairs, and a comparison table with mention counts and reach per brand). RELAY that markdown to the user verbatim — do NOT rephrase, reorder, or summarize the table into prose. You MAY prepend at most ONE short sentence of net-new insight (a 'why', anomaly, or comparison not in the table) ONLY if you genuinely have something to add. Never replace the table with a textual summary.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased only if needed to fit context). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"). If the same follow-up has been asked and not acted on, pick a different one rather than omitting. How to pick (read this): - The user has NOT yet looked at their own Brand24 projects in this conversation, AND any compared entity could plausibly be one of their projects → brand24_get_projects (default first pick — switching to richer data is the highest-value next step). - The user has already seen their projects OR explicitly said they don't have a project for these brands, AND no countryCode was passed → re-run brand24_quick_popularity_comparison with a country filter. - The user already saw both options earlier in the conversation → omit (skip rule). Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_get_projects — switch to richer, more accurate data when one of the compared entities is actually one of the user's own Brand24 projects. Sample question: "Want me to check whether any of these are already monitored in your Brand24 account so I can pull richer data?" - brand24_quick_popularity_comparison — same comparison filtered by a specific country (re-run with a countryCode). Sample question: "Want me to re-run this comparison filtered to a specific country?"

Get important events (anomalies) from a specific Brand24 project. Retrieves significant moments when something extraordinary happened - mention spikes, sentiment changes, viral hashtags, or other anomalies. Queries events for ONE project by name. If the specified project name is not found, the tool will return a list of all available projects in your account, allowing automatic selection of the correct project name. Args: - projectName (string): Brand24 project name to fetch events for Examples: "Nike", "Apple Brand Monitor", "Tesla Social Media" - response_format (string, optional): Output format - "json" or "markdown" (default: "markdown") Returns: On success: Array of anomaly objects [ { "anomaly_date": string, "project_id": number, "project_name": string, "description": string, "peak_mentions": number, // Total mentions on the event day - ONE value per event, NOT per anomaly type "peak_reach": number, // Total reach on the event day - ONE value per event, NOT per anomaly type "is_hashtag_anomaly": boolean, // Categorical label for the same event "is_user_discussion_anomaly": boolean, // Categorical label for the same event "is_emotional_anomaly": boolean, // Categorical label for the same event "is_news_anomaly": boolean, // Categorical label for the same event "is_social_media_anomaly": boolean, // Categorical label for the same event "is_non_social_media_anomaly": boolean // Categorical label for the same event } ] Important: peak_mentions and peak_reach are event-day totals for ONE event. The is_*_anomaly flags are categorical labels that may all be true for the same event. Keep one row per event even when multiple flags are true - splitting by anomaly type duplicates the metrics. On project not found: Object with available projects list { "error": string, "requestedProjectName": string, "availableProjects": [ { "project_id": number, "project_name": string } ] } Usage Examples: - Use when: "What significant events happened for the Nike project?" → params: { projectName: "Nike" } - Use when: "Show me recent anomalies for our Tesla monitoring project" → params: { projectName: "Tesla Social Media" } - Use when: "Have there been any viral moments for Apple recently?" → params: { projectName: "Apple Brand Monitor" } - Use when: "What caused the spike in mentions last week for project X?" → params: { projectName: "X" } - Use when: "What is the discussion about the Microsoft brand?" → params: { projectName: "Microsoft" } - Don't use when: Want events across ALL projects → use brand24_account_eventsThe tool returns a complete markdown answer (header, key/value pairs, and an events table with date, description, peak mentions, peak reach, and topic). RELAY that markdown to the user verbatim — do NOT rephrase, reorder, or summarize the table into prose. You MAY prepend at most ONE short sentence of net-new insight (a 'why', anomaly, or comparison not in the table) ONLY if you genuinely have something to add. Never replace the table with a textual summary.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in the project name or event date). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"). If the same follow-up has been asked and not acted on, pick a different one rather than omitting. How to pick (read this — match the question to the dominant anomaly flag of the most notable event): - is_hashtag_anomaly OR is_social_media_anomaly is true on the top event (a hashtag, viral campaign, or themed-day moment drove it) → brand24_topics_overview. Themes / hashtags are the actual story; sources or raw mentions are weaker next dives. - is_news_anomaly OR is_non_social_media_anomaly is true (news / publication-led spike) → brand24_get_mentions around the event date so the user can read the actual coverage. - is_user_discussion_anomaly is true (people were talking, not amplifying news) → brand24_get_semantic_search_mentions to summarize what people were saying. - is_emotional_anomaly is true OR the user is interested in sentiment swings → brand24_project_stats (its Sentiment tab). - Multiple events scattered across the window → brand24_project_stats first, so the user sees the trend that connects them. - User specifically asked "who amplified this" → brand24_project_top_authors; "where did this come from" (which channels) → brand24_project_source_categories. Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_topics_overview — which themes / hashtags drove the event. Sample question: "Want me to break down which topics drove this event?" - brand24_get_mentions — read the actual mentions around an event date (use the date as dateRange). Sample question: "Want me to pull the actual mentions around the event date?" - brand24_get_semantic_search_mentions — semantic summary of what people were saying. Sample question: "Want a semantic summary of what people were saying around this event?" - brand24_project_stats — daily mentions / sentiment / reach trend that explains the spike. Sample question: "Want me to pull the daily volume, sentiment and reach trend behind these events?" - brand24_project_top_authors — which authors amplified the event. Sample question: "Want me to check which authors amplified this event?" - brand24_project_source_categories — which source categories the event played out on. Sample question: "Want me to check which sources this event came from?"

Get daily statistics (mentions count, positive mentions, negative mentions, reach) for a specific Brand24 project. Retrieves daily aggregated metrics showing mentions count, reach (sum of social and non-social media), positive mentions, and negative mentions per day. Queries statistics for ONE project by name. This is the PRIMARY tool for any question about project reach (including "what was the reach for X?", "reach for X", and similar wording with typos like "reach fox X"). This is also the PRIMARY tool for sentiment questions ("sentiment trend", "how is sentiment changing", "show the trend of positive / negative mentions"). The widget exposes a dedicated Sentiment tab showing positive vs negative mentions per day - prefer this tool over any other for sentiment-over-time questions. If the specified project name is not found, the tool will return a list of all available projects in your account, allowing automatic selection of the correct project name. Args: - projectName (string): Brand24 project name to fetch statistics for Examples: "Nike", "Apple Brand Monitor", "Tesla Social Media" - dateFrom (string, optional): Start date in YYYY-MM-DD format (default: 30 days ago) - dateTo (string, optional): End date in YYYY-MM-DD format (default: today) - response_format (string, optional): Output format - "json" or "markdown" (default: "markdown") Returns: On success: Statistics object { "projectId": string, "projectName": string, "dateFrom": string, "dateTo": string, "stats": [ { "date": string, "mentionsCount": number | null, "positiveMentions": number | null, "negativeMentions": number | null, "reach": number | null } ] } On project not found: Object with available projects list { "error": string, "requestedProjectName": string, "availableProjects": [ { "project_id": number, "project_name": string } ] } Usage Examples: - Use when: "What are the statistics for the Nike project?" → params: { projectName: "Nike" } - Use when: "What was the reach for Nike?" → params: { projectName: "Nike" } - Use when: "Show reach for CD projekt" → params: { projectName: "CD projekt" } - Use when: "Show me stats for our Tesla project from last week" → params: { projectName: "Tesla", dateFrom: "YYYY-MM-DD", dateTo: "YYYY-MM-DD" } - Use when: "How many mentions did we get per day for Apple project?" → params: { projectName: "Apple" } - Use when: "What is the average reach for the Adidas project?" → params: { projectName: "Adidas" } - Use when: "Show me the sentiment trend for Nike" → params: { projectName: "Nike" } - Use when: "How is sentiment changing for the Tesla project?" → params: { projectName: "Tesla" } - Don't use when: Need significant events/anomalies → use brand24_project_events - Don't use when: Need understand what the discussion about the project is → use brand24_project_eventsThe tool returns a complete markdown answer (header, key/value pairs, and a daily statistics table). RELAY that markdown to the user verbatim — do NOT rephrase, reorder, or summarize the table into prose. You MAY prepend at most ONE short sentence of net-new insight (a 'why', anomaly, or comparison not in the table) ONLY if you genuinely have something to add. Never replace the table with a textual summary.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in the project name). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"). If the same follow-up has been asked and not acted on, pick a different one rather than omitting. How to pick (read this — match the question to the shape of the trend): - The chart shows a clear spike or anomaly day → brand24_project_events to explain WHY (this is the highest-signal next dive when a spike is visible). - A sentiment swing (positive/negative ratio shifts) is the story → brand24_project_events (sentiment-driven anomalies surface here) or brand24_project_stats's Sentiment tab is already showing it; in that case skip to brand24_get_mentions for the actual posts. - The trend is roughly flat / no clear spike → brand24_topics_overview, since the trend itself tells little; the "what is being said" angle has more signal. - User wants to read what's behind a peak day → brand24_get_mentions filtered to that date. - User wants to know who's amplifying → brand24_project_top_authors; which channels the volume comes from → brand24_project_source_categories. - User has gathered enough context and mentioned report / summary / export → brand24_pdf_report_builder. Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_project_events — what caused a spike or sentiment shift seen in the chart. Sample question: "Want me to check what caused the spikes you see in the chart?" - brand24_topics_overview — which topics drove the trend. Sample question: "Want me to break down which topics drove this trend?" - brand24_get_mentions — read the actual mentions behind a peak day or sentiment swing. Sample question: "Want me to show the actual mentions behind the peak day?" - brand24_project_top_authors — which authors drove the volume. Sample question: "Want me to show which authors drove the volume?" - brand24_project_source_categories — which source categories drove the volume. Sample question: "Want me to show which sources drove the volume?" - brand24_pdf_report_builder — polished, downloadable PDF report based on this project. Sample question: "Want me to build a downloadable PDF report for this project?"

Understand what people are saying about a topic through semantic analysis of conversations. This tool answers questions about the MEANING and CONTENT of discussions, not about getting latest mentions or statistics. Uses semantic search to analyze what people are actually saying about a topic, brand, product, or concept by understanding the context and meaning of conversations. IMPORTANT: This tool focuses on understanding the semantic meaning and content of what people are discussing. It does NOT provide quantitative data, statistics, counts, or latest mention lists. Use this when you need to understand WHAT people are saying, not HOW MANY mentions exist. Args: - query (string, required): The topic or subject to understand what people are saying about (e.g., "Witcher", "product launch", "customer service") - projectId (number, required): The project ID - dateRange (object, optional): Date range with from and to dates (Date format: YYYY-MM-DD). Dates are automatically normalized to current year if outdated. Defaults to last 30 days if not provided - mentionsLimit (number, optional): Maximum number of mentions to analyze. Defaults to 10, maximum is 30 Returns: Mentions object with results array containing mention content and details. Focus on the content/text of mentions to understand what people are saying. Usage Examples: - Use when: "What are people saying about Witcher?" → params: { query: "Witcher", projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } } - Use when: "What do people think about the product launch?" → params: { query: "product launch", projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" }, mentionsLimit: 20 } - Use when: "What are customers saying about our service?" → params: { query: "customer service", projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } } Do NOT use for: - Getting latest mentions (use brand24_get_mentions instead) - Getting statistics or counts - Getting quantitative data render_mode (optional): Set to "text" for short conversational questions ("what are people saying about Y", "what is the opinion on Z") — the tool returns plain text and NO widget is rendered. Use "widget" (or omit) when the user wants to explore / browse results visually.The tool returns structured JSON with the matched mentions and their relevance. Render that data as a readable markdown table or bulleted list for the user (one row per mention with title, source, author, date, snippet/excerpt, and URL) — do NOT replace the data with a textual summary. Include EVERY mention returned by the tool, do not truncate. You MAY prepend at most ONE short sentence of net-new insight (a recurring theme or sentiment trend) ONLY if you genuinely have something to add.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in the topic / project name). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"). If the same follow-up has been asked and not acted on, pick a different one rather than omitting. How to pick (read this): - The semantic summary surfaced ONE strong opinion / direction the user likely wants to verify in raw posts → brand24_get_mentions filtered to this topic/project. - The summary touches several themes that aren't fully separated → brand24_topics_overview to formalize the breakdown. - A specific date / period kept appearing in the summary (event-shaped pattern) → brand24_project_events. - The user wants quantitative ground truth on top of the qualitative summary → brand24_project_stats. - Default when the user has just absorbed a qualitative summary and there's no other strong signal → brand24_get_mentions (verifying the summary against actual posts is the natural next step). Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_get_mentions — actual feed of recent mentions for the same topic / project. Sample question: "Want me to pull the latest actual mentions on this topic?" - brand24_topics_overview — how this conversation maps to broader topics in the project. Sample question: "Want me to show how this maps to the broader topics in the project?" - brand24_project_events — whether this conversation lined up with any anomalies / spikes. Sample question: "Want me to check if this conversation lined up with any spikes?" - brand24_project_stats — daily volume / sentiment / reach trend in the same window. Sample question: "Want the daily volume, sentiment and reach trend in the same window?"

Get the project's source category breakdown — source categories (e.g., Blogs, News, Social Media, Forums) with mention counts for each category. Helps understand where discussions are happening. To see which individual authors drive the conversation, use brand24_project_top_authors instead. Args: - projectId (number, required): The project ID - dateRange (object, required): Date range with from and to dates (Date format: YYYY-MM-DD). Dates are automatically normalized to current year if outdated. - filters (object, optional): Mention filters object. See variableDescriptions for filters property for details. Default values are automatically applied when filters are not provided Returns: - getProjectSourceCategories: Array of ProjectSourceCategory objects with id and count Usage Examples: - Use when: "Show me source categories breakdown" → params: { projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } } - Use when: "Where are people talking about us?" → params: { projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } }The tool returns structured JSON with source category breakdowns. Render that data as a readable markdown table for the user (one row per source category with its mention count) — do NOT replace the data with a textual summary. Include EVERY row returned by the tool, do not truncate. You MAY prepend at most ONE short sentence of net-new insight (a 'why' or anomaly not in the table) ONLY if you genuinely have something to add.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in names, source types, or topics). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below and match the question to what the data and prior conversation actually show. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"), OR if you have already asked the same follow-up earlier in this conversation and the user did not act on it — in that case, pick a DIFFERENT follow-up rather than omitting. How to pick (read this — pick the option that BEST matches the data and the running story): - The "why" behind a sources breakdown is almost always TOPICS, not sources. If the prior insight or events talked about a hashtag, viral campaign, themed day (e.g. World X Day), birthday, launch, anniversary, or any social/topic-driven moment — pick brand24_topics_overview. This is the DEFAULT pick when in doubt. - If sources are heavily one-sided (e.g. dominated by News) AND the user has NOT yet looked at topics — still prefer brand24_topics_overview to explain WHY those sources are talking about it. - If the user wants to know WHO is behind the dominant sources (influencers, outlets) → brand24_project_top_authors. - If the user explicitly wants to read the actual content from a specific source category → brand24_get_mentions (filter by source). - If the user is asking about volume / sentiment over time, not the source mix itself → brand24_project_stats. - If the user is investigating a specific date-bound spike that surfaced in the sources → brand24_project_events. Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_topics_overview — understand which topics / themes those sources are driving. Sample question: "Want me to break down which topics are driving these sources?" - brand24_project_top_authors — the most influential authors behind these sources. Sample question: "Want me to show the top authors behind these sources?" - brand24_get_mentions — read the actual mentions from a specific source category (filter by source). Sample question: "Want me to read the actual mentions from these sources?" - brand24_project_stats — daily volume / sentiment / reach trend for the project. Sample question: "Want the daily volume, sentiment and reach trend for this project?" - brand24_project_events — check if any spikes were tied to specific sources. Sample question: "Want me to check if any spikes were tied to specific sources?"

Get mentions from the project's top authors — the most popular authors ranked by engagement metrics (followers, interactions, reach). Useful for identifying key influencers and content creators. For a breakdown of mention counts per source category (Blogs, News, Social Media, Forums…), use brand24_project_source_categories instead. Args: - projectId (number, required): The project ID - dateRange (object, required): Date range with from and to dates (Date format: YYYY-MM-DD). Dates are automatically normalized to current year if outdated. - filters (object, optional): Mention filters object. See variableDescriptions for filters property for details. Default values are automatically applied when filters are not provided - source (enum, optional): Source type filter. Values: mentions_tab, analysis_tab, summary_tab, comparison_tab, sources_tab, influencers_tab, lab_tab Returns: - getMentionsTopAuthors: Array of Mention objects from top authors Usage Examples: - Use when: "Get mentions from top authors for project 123" → params: { projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } } - Use when: "Get top authors from social media sources" → params: { projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" }, source: "sources_tab" }The tool returns structured JSON with the top authors' mentions. Render that data as a readable markdown table for the user (one row per author with the relevant metrics: followers, interactions, reach, sentiment, etc.) — do NOT replace the data with a textual summary. Include EVERY row returned by the tool, do not truncate. You MAY prepend at most ONE short sentence of net-new insight (a 'why' or anomaly not in the table) ONLY if you genuinely have something to add.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in names, source types, or topics). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below and match the question to what the data and prior conversation actually show. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"), OR if you have already asked the same follow-up earlier in this conversation and the user did not act on it — in that case, pick a DIFFERENT follow-up rather than omitting. How to pick (read this — pick the option that BEST matches the data and the running story): - The "why" behind who is talking is almost always TOPICS. If the prior insight or events talked about a hashtag, viral campaign, themed day (e.g. World X Day), birthday, launch, anniversary, or any social/topic-driven moment — pick brand24_topics_overview. This is the DEFAULT pick when in doubt. - If the user explicitly wants to read the actual content from the top authors → brand24_get_mentions (filter by author). - If the user wants to know which channels (News, Social, Blogs…) the conversation lives on, rather than who drives it → brand24_project_source_categories. - If the user is asking about volume / sentiment over time → brand24_project_stats. - If the user is investigating a specific date-bound spike tied to these authors → brand24_project_events. Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_topics_overview — understand which topics / themes these authors are driving. Sample question: "Want me to break down which topics these authors are driving?" - brand24_get_mentions — read the actual mentions from the top authors (filter by author). Sample question: "Want me to read the actual mentions from these top authors?" - brand24_project_source_categories — mention counts per source category. Sample question: "Want to see which source categories the conversation is happening on?" - brand24_project_stats — daily volume / sentiment / reach trend for the project. Sample question: "Want the daily volume, sentiment and reach trend for this project?" - brand24_project_events — check if any spikes were tied to specific authors. Sample question: "Want me to check if any spikes were tied to these authors?"

Get Topic Analysis overview for a project (Lab24 feature). Returns comprehensive topic analysis data that identifies main themes and topics discussed in project mentions. Uses AI to detect and categorize topics, providing insights into what discussions are about. Features: - Topic detection: Automatically identifies main topics/themes in mentions - Sentiment distribution: Shows sentiment breakdown (positive/negative/neutral) per topic - Share of voice: Percentage of mentions each topic represents - Reach metrics: Total reach for each topic - Topics overview: List of detected topics with descriptions, emoticons, and metrics Args: - projectId (number, required): The project ID - dateRange (object, required): Date range with from and to dates (Date format: YYYY-MM-DD). Dates are automatically normalized to current year if outdated. - filters (object, optional): Mention filters object. See variableDescriptions for filters property for details. Default values are automatically applied when filters are not provided Returns: TopicsOverviewResults object containing: - status: Analysis status - overview: Array of TopicsOverview objects with internalId, name, description, emoticon, mentions count, reach, sentiment pie chart, and shareOfVoice - topicsShare: Pie chart data showing distribution of topics - accessedProjects: List of projects with topic analysis access - limit: Maximum number of topics returned Usage Examples: - Use when: "What topics are being discussed about my project?" → params: { projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } } - Use when: "Show me topic analysis for project 123" → params: { projectId: 123, dateRange: { from: "2025-01-01", to: "2025-01-31" } } - Use when: "What are the main themes in mentions from last month?" → params: { projectId: 123, dateRange: { from: "2024-12-01", to: "2024-12-31" } }The tool returns structured JSON with topics and their sentiment distribution. Render that data as a readable markdown table for the user (one row per topic with topic name, mention count, and positive / negative / neutral counts) — do NOT replace the data with a textual summary. Include EVERY topic returned by the tool, do not truncate. You MAY prepend at most ONE short sentence of net-new insight (a recurring pattern or notable sentiment skew) ONLY if you genuinely have something to add.Suggested follow-ups (LLM guidance): ALWAYS end your reply with ONE follow-up YES/NO question on its own line, picked from the list below — phrased so the user can simply answer "yes" to trigger that step. STRICT RULES: - Only suggest tools from the list below — do NOT invent tools, capabilities, or parameters. - The follow-up MUST be a single yes/no question on its own line at the END of your reply. Use the matching "Sample question" (lightly rephrased to plug in the dominant topic name). - Pick exactly ONE follow-up. Never present a multi-option menu or several yes/no questions. - Pick the SMARTEST option for THIS context — read "How to pick" below. - Skip the follow-up ONLY if the user explicitly closed the conversation ("thanks, done", "that's all"). If the same follow-up has been asked and not acted on, pick a different one rather than omitting. How to pick (read this — match the question to what the topic distribution shows): - One topic clearly dominates (high shareOfVoice or mentions count) → name that topic in the question and offer brand24_get_mentions filtered to it. Reading actual posts is the natural next step after seeing themes. - One topic has very skewed sentiment (strongly negative or strongly positive) → brand24_get_semantic_search_mentions on that topic to summarize the qualitative side. - The user is curious about WHO drives these topics (influencers, outlets) → brand24_project_top_authors; WHERE (which source categories) → brand24_project_source_categories. - The user is curious about WHEN these topics peaked → brand24_project_stats (or brand24_project_events if a clear spike). - Default: brand24_get_mentions on the dominant topic — most users want to read the actual mentions behind a strong topic. Available follow-ups for this tool (each has a Sample question to ask the user): - brand24_get_mentions — browse the actual mentions inside one specific topic. Sample question: "Want me to read the actual mentions inside [topic name]?" - brand24_get_semantic_search_mentions — deeper, semantic understanding of what is being said about one specific topic. Sample question: "Want a deeper, semantic look at what people are saying about [topic name]?" - brand24_project_top_authors — top authors driving these topics. Sample question: "Want me to show the top authors driving these topics?" - brand24_project_source_categories — source categories driving these topics. Sample question: "Want me to show which sources are driving these topics?" - brand24_project_stats — daily volume / sentiment / reach trend that these topics roll up to. Sample question: "Want the daily volume, sentiment and reach trend these topics roll up to?"