Unwrap

MCP App internal: list filterable fields (with allowed operators, value type, and priority) for the chart iframe's filter editor. Returns only structuredContent. LLMs SHOULD NOT call this — use discover_filter_schema instead. Hosts that honor _meta.ui.visibility hide this tool from the LLM tool catalog.

MCP App internal: re-fetch a chart's data when the user changes filters, date range, chart type, or breakdown_field inside the chart iframe. Pass the same breakdown_field contract as plot_chart (none = All Feedback; Group / leaf_group / source / segment names). Returns only structuredContent (no markdown body) so the LLM transcript isn't polluted by every interactive filter change. LLMs SHOULD NOT call this tool — use plot_chart instead. Hosts that honor _meta.ui.silent will hide this tool from the LLM tool catalog.

Plot charts and tables over time or by breakdown. Use this when the user is asking about an issue or group over time, when they want top groups or top issues shown visually, or when a visual comparison would make the answer clearer. Use exactly one input series by default. Multiple input series are allowed ONLY if the user explicitly requests separately defined series or no single-series breakdown can represent the requested chart. Comparisons and multiple visible lines/bars do not by themselves require multiple input series. Use filter + breakdown_field: scope with series filters (e.g. Group = Notifications OR Group = Messages), then set breakdown_field to split that population (Group, leaf_group, source, or a segment name like Sentiment). Use breakdown_field=none (All Feedback) for a single volume line under the filters. When using multiple input series, explain which exception applies in explanation. Choose the chart type based on whether the user wants a time trend, an aggregate ranking, or a share comparison. After calling it, still return written analysis alongside the chart. Use search_segments or list_field_values before filtering or selecting numeric segments.

Query compact raw or aggregate feedback data without building a chart. Use fields for raw values, or group_by with a metric for counts, sums, averages, NPS, or CSAT. Prefer one grouped query over one query per value. Use plot_chart instead for trends or visualizations.

Read the full content of one or more Unwrap documentation articles and return them as markdown. Only URLs returned by `search_documentation` are allowed; any other URL is rejected without being fetched. Use this to ground answers to product, how-to, or integration-setup questions in the official docs.

Create a named sample of feedback entries. Use a representative random sample for prevalence, or semantic_query plus instruction for targeted examples. Slug names must be snake_case, 3-50 characters, lowercase letters, numbers, underscores only. MCP output includes sample_reference_id at the end — pass it to analyze_entry_text, classify_feedback_entries, or Sample Reference filters on later calls.

Get the top / biggest / most important groups by magnitude — use this for 'top issues', 'main pain points', or 'where should we focus' questions, optionally sorted by a numeric segment. To find feedback ABOUT a specific topic by semantic relevance, use `search_groups` instead. Leave `query` empty to rank purely by volume; pass a `query` only to narrow the ranking toward a topic. Pair with `plot_chart` when a visual summary would help.

Search Unwrap's official product documentation (the public help center) and return a ranked list of matching article titles and URLs. Use this when the user asks how to do something in Unwrap, how a feature works, or how to set up or configure an integration. Pair with `read_documentation_article` to read an article's content before answering.

Find feedback ABOUT a topic — taxonomy groups ranked by semantic relevance to `query`. For the biggest issues by magnitude ('top issues', 'where should we focus'), use `search_actionable_groups` instead. If you filter, discover fields and values first instead of guessing them. For Sample Reference predicates use integer sample_reference_id from sample_feedback_entries or classify_feedback_entries output (see discover_filter_schema).

Search for relevant segment fields and values by semantic similarity. Always call discover_filter_schema first to discover available fields before using this tool.