AIlon

Create a project, optionally with multiple AI-generated audiences in one step! Unless the user has already specified a different approach, check `get_playbook()` for a matching task playbook first and follow it — these define how to responsibly derive each audience's `prompt`. Each audience's `prompt` MUST reference only features confirmed to exist in AIlon — verify via `list_features_of_category` or `search_features` before writing it. Do not invent or assume trait names; an unverified prompt can silently produce an audience that doesn't match intent, with no error to signal the mismatch. Combine the feature with an AND/OR/NOT-logic in the prompt. Omit `organisation_id` for a private project owned by the authenticated user. Set `organisation_id` to create a project shared with your organisation (requires organisation create-projects permission). Pass `audiences` as a list of `{name, prompt}` objects to generate several audiences immediately. The call blocks until every audience has reached a terminal status (Berechnet or a failure status) — no separate polling is needed to wait for calculation to finish. Omit `audiences` to create an empty project and add audiences later with `generate_audience` (same prompt requirements apply there). This task can take up to one minute. Make sure the audience prompt does not contain PII like names oder adresses.

Create a new audience in the given project by generating an audience definition from `prompt`, then block until calculation finishes. Returns the audience's `name`, `description`, `definition_text`, `status`, `segments`. The call blocks until the audience reaches a terminal status — `'Berechnet'` (calculated) or `'Berechnung fehlgeschlagen.'` (failed) — no separate polling is needed to wait for calculation. `description` is populated with the original generation `prompt` once calculation finishes. `definition_text` is the final, computed Boolean definition AIlon actually used — the authoritative ground truth for what the audience contains, distinct from `description`'s input prompt. `prompt` MUST reference only features confirmed to exist in AIlon — verify via `list_features_of_category` or `search_features` before calling this tool. Do not invent or assume trait names; an unverified prompt can silently produce an audience that doesn't match intent, with no error to signal the mismatch. Combine features with AND/OR/NOT logic in the prompt. Note: this fails on Example Projects (`public: true`) — new audiences can only be added to the user's own projects. Make sure the prompt does not contain PII.

Return feature-category metrics for 1–20 audiences in one call, always in comparison form (even for a single audience). `feature_category_id` comes from `list_feature_categories`; `audience_ids` come from `list_audiences` or `create_project`/`generate_audience`. `sort_audience_id` is always required — it must be one of `audience_ids` and drives sorting, filtering, and row limits. Each feature row includes `feature.display_name` and `feature.display_name_category` so results are readable without a separate feature lookup. Defaults to the top 50 features by `index_value` (descending). Pass `search_keywords` to restrict results to matching features (same fuzzy + semantic search as `search_features`) — when combined with `feature_ids`, only features matching both are returned. Use `min_mean_audience` or `min_index_value` to drop low-signal features before applying `limit`. Set `limit` up to 500 for broader views. Charges credits for non-example projects; audiences in example projects (`public: true`) are always free. The response's `billing` field reports `credits_charged`, `balance_before`, `balance_after`, and `min_balance` for the call — check `whoami` beforehand if you need to reason about affordability, since there is no separate credit-preview tool.

Workflow slug to load in full, e.g. `analyze_project` or `create_audiences_no_data_brief`. Omit to list available workflow summaries — always prefer this over guessing a slug, since the set of playbooks can change.

List audiences across all projects visible to the authenticated user — this is global, not scoped to one project. Use `project_ids` and/or `audience_ids` to filter (optional; when both are given, results are the union of both filters). Each item includes `project_id`, segment IDs, `definition_text` (the final computed Boolean definition), and `potential_*` (reach) fields, so no separate per-audience lookup tool is needed. Use `list_projects` first to find `project_id` values if filtering by project. Paginate with `offset` when `count` is greater than `limit`.

List feature categories (widgets) available for this API version, including their base credit cost per audience per call (`credit_cost`). Use returned `feature_category_id` values as the `feature_category_id` parameter in `list_features_of_category` and `get_insight_data`. Actual charges for a given call are reported in `get_insight_data`'s `billing` response field, not here.

List demographic or behavioural features that make up a feature category. No `limit` parameter — response size scales with the category's tier: Basic categories have ~14 features, Extended categories ~2,900 combined across 15 categories (roughly 100–450 each), but Plus-tier categories are far larger — "Marken" alone returns 2,061 features (~200,000 characters) in one response. For Plus-tier categories, use `search_features` instead.

List dashboard projects visible to the authenticated user. Start here to discover ``project_id`` values for audience and insight tools. Paginate with ``offset`` when ``count`` is greater than ``limit``. Set ``public=true`` for example projects only, ``public=false`` for non-example projects.

Search features globally across all categories by name or meaning (fuzzy + semantic). Pass one concept per keyword, e.g. ``["Fitness", "Sport", "active"]``. A feature may appear in multiple categories. Does not charge credits or load metrics.

Return the authenticated AIlon user and account status behind this MCP session: `user_id`, `username`, `email`, `organisation_id`, `organisation_name`, `credit_balance`, `min_balance`, `mcp_resource_url`. Use `organisation_id` as the `organisation_id` parameter in `create_project` to create a project shared with the organisation instead of a private one. Use `credit_balance`/`min_balance` to reason about affordability `get_insight_data` calls — there is no separate credit-preview tool.