DALIA MCP

search_contacts

Search contacts by name, email or phone, by custom attributes (attribution: utm_source, utm_medium, utm_campaign, ad_name, city…), by labels, by creation period and by whether they have an open conversation. Newest first, with next_cursor and the contact url. 'Cuántos vinieron de X' is attribution: prefer the attribution report for totals and use this to list the people.

get_contact

Full profile of ONE contact: channel identities, custom attributes (attribution), labels, internal notes and their conversations. Use it before acting on a person.

create_contact

Create a contact. Requires name plus at least one of phone_number or email. Phone/email must be unique per workspace.

update_contact

Partially update a contact: only the provided fields change. Pass an empty string to clear a field.

list_labels

List the workspace labels (tags) available for contacts.

No parameters.

create_label

Create a label (tag). Titles are unique per workspace. Requires label management permission.

tag_contact

Attach a label to a contact, by label_id or by label_title (case-insensitive exact match). Idempotent.

untag_contact

Remove a label from a contact, by label_id or label_title.

list_conversations

List inbox conversations (WhatsApp / Instagram / Messenger / web), newest activity first, with contact, channel, last message preview and the chat url. Filter by status, inbox, assignee (name), unassigned_only, priority, team, label, unread_only and created/updated date ranges. Paginate with next_cursor. Only shows what the caller may see: an agent restricted to their own conversations gets only theirs.

get_conversation_messages

Read the messages of a conversation, newest first, with a cursor for older pages.

assign_conversation

Assign a conversation to an agent by name, email or id. Requires assignment permission.

unassign_conversation

Remove the assignee from a conversation.

resolve_conversation

Mark a conversation as resolved (fires resolution automations).

reopen_conversation

Reopen a resolved/snoozed conversation.

set_conversation_priority

Set priority: urgent, high, medium, low or none.

snooze_conversation

Snooze a conversation until an ISO timestamp in the future.

tag_conversation

Attach a label to a conversation by label_id or label_title (fires label automations). Idempotent.

untag_conversation

Remove a label from a conversation by label_id or label_title.

get_conversation_detail

Full context of ONE conversation: status, priority, assignee, contact, labels and the most recent messages. Use it to summarise or to decide an action on a specific chat; for account-wide counts use the report tools instead.

list_inboxes

List workspace inboxes (channels) with their ids for filtering conversations.

No parameters.

list_agents

List workspace agents (for assignment).

No parameters.

list_teams

List workspace teams.

No parameters.

crm_list_pipelines

List CRM pipelines with their stages (name, type open/won/lost, order). Use this first to get pipeline and stage ids/names for the other CRM tools.

crm_list_leads

List CRM leads, newest first. Filter by pipeline/stage (name or id), status (new/qualified/unqualified/converted), owner (name, email or id) or unassigned_only, contact, source (substring) and creation date range. Paginate with next_cursor; each lead carries its dashboard url.

crm_get_lead

Fetch a single lead by id, including attribution (UTM) and custom attributes.

crm_create_lead

Create a lead in a pipeline (by name or id). Lands in the given stage or the first one. Attribution: pass utm_source/utm_medium/utm_campaign, or a legacy source label (whatsapp, instagram, facebook, web, email, llamada) that maps to standard UTM. Referral-type sources require attribution_detail (who referred).

crm_update_lead

Partially update a lead (name, contact fields, value, owner, status, attribution). Manual UTM edits are origin-tracked and the automatic value is snapshotted once. Owner changes notify the new owner by email.

crm_move_lead_stage

Move a lead to another stage of its pipeline, by stage name or id (board-drag parity: won-type stages set status qualified, lost-type set unqualified; the stage change is logged on the timeline).

crm_mark_lead_lost

Mark a lead as lost (status unqualified, moved to the lost stage when the pipeline has one, reason logged on the timeline). Converted leads cannot be marked lost.

crm_reopen_lead

Reopen a lost lead: status back to new, moved to the first open stage.

crm_convert_lead

Convert a lead into an opportunity in a target pipeline (by name or id). Enforces the conversion minimums (name + client + business context), links or creates the contact, carries attribution, interest items and human activities over.

crm_list_opportunities

List CRM opportunities (deals), newest first, filterable by pipeline, stage, status (open/won/lost), owner or contact.

crm_get_opportunity

Fetch a single opportunity by id, including amount, close dates and attribution.

crm_create_opportunity

Create an opportunity (deal). Requires a linked contact, an expected close date (YYYY-MM-DD) and an owner — same rule as the UI.

crm_update_opportunity

Partially update an opportunity. Setting status=lost (or clearing the reason of a lost deal) validates the loss reason against the account catalog. Owner changes notify the new owner by email.

crm_move_opportunity_stage

Move an opportunity to another stage of its pipeline, by stage name or id. Won-type stages close the deal as won; lost-type stages require lost_reason and close it as lost; open stages reopen it. This is how deals are won, lost and reopened.

crm_mark_opportunity_outcome

Close an opportunity as 'won' or 'lost': sets the status and actual close date, moves it to the matching terminal stage and logs the outcome. Losing requires lost_reason from the account catalog (see crm_move_opportunity_stage). Reopen later with crm_update_opportunity status=open.

crm_list_activities

Activities of the CRM. With entity_type + entity_id: the timeline of one lead or opportunity (notes, calls, meetings, tasks, stage changes), newest first. Without entity_id: account-wide list, filterable by type, entity_type, owner (assigned agent), pending_only (open items with a due date, soonest first) and due date range: 'tareas pendientes', 'vencidas esta semana'.

crm_add_note

Add a note to the timeline of a lead or opportunity.

crm_schedule_activity

Schedule a call, meeting or task on a lead or opportunity, with a due date and optional assignee.

crm_complete_activity

Mark a scheduled activity (call/meeting/task) as done.

list_crm_properties

Lista propiedades inmobiliarias del catálogo del CRM (crm_properties). Pensado para cuentas inmobiliarias. Filtros: - name_search: busca por nombre/código de la propiedad o proyecto (búsqueda parcial). USA ESTE FILTRO cuando el usuario mencione el nombre de un proyecto o desarrollo (ej. "Proyecto Alfa", "Torres del Parque"). Las propiedades tienen nombres como "PROYECTO ALFA - Apartamentos". - city / sector (búsqueda parcial) - operation: la operación comercial. IMPORTANTE: en la DB puede estar como "Compra" (= venta desde la perspectiva del vendedor) o "Arriendo". Puedes pasar "venta", "compra", "arriendo" o "alquiler" — la tool normaliza automáticamente. - property_type: ej. "Apartamento", "Casa", "Casa Campestre", "Local Comercial", "Oficina", "Lote campestre", etc. Case-insensitive. - availability (parcial) - min_price / max_price (sobre list_price) - min_bedrooms / max_bedrooms - only_active (default true) - only_published (sólo published al sitio público) IMPORTANTE: cuando el usuario pregunta por un proyecto inmobiliario por nombre (ej. "ventas de Proyecto Alfa", "oportunidades de Proyecto Beta"), usa name_search con ese nombre. No uses sector ni city para buscar por nombre de proyecto. Cuando el usuario dice "vendemos" o "en venta", filtra por operation = "venta" (la tool lo normaliza a "Compra" internamente).

get_crm_property_detail

Detalle completo de una propiedad: imágenes y oportunidades que la incluyen como ítem.

list_crm_lost_reasons

Lista los motivos de pérdida configurados en la cuenta. Consúltalos antes de cerrar una oportunidad como perdida: el motivo es obligatorio y debe ser uno de estos.

No parameters.

get_crm_funnel

Devuelve el embudo (funnel) de un pipeline para un periodo dado: cuántas oportunidades hay en cada stage, monto agregado y tasa de conversión stage → siguiente. Úsalo cuando el usuario pregunte: - "cómo va el embudo / pipeline este mes" - "muestra el funnel de ventas" - "conversión por etapa" Si no se especifica pipeline_name, usa el pipeline marcado como is_default; si no hay default, devuelve un error claro y pide elegir. Cuenta SIEMPRE oportunidades creadas dentro del period (no las que ya estaban abiertas antes).

get_crm_conversion_report

Reporte de conversión: cuántas oportunidades se cerraron como won vs lost en un periodo, win rate y ciclo medio (días entre created_at y actual_close_date). Si se pasa compare_with_previous=true, devuelve también el periodo anterior comparable.

get_crm_agent_performance

Ranking de agentes (owners) por desempeño de oportunidades en un periodo. Para cada agente devuelve: oportunidades won / lost / open, monto ganado y win rate. Ordenado por monto ganado descendente.

get_crm_lost_reasons_breakdown

Razones de pérdida de oportunidades agrupadas, con conteo y monto perdido por razón. Útil para entender por qué se pierden deals y priorizar mejoras.

analyze_lead_response_times

Analiza tiempos de respuesta de leads del CRM: cruza leads → contactos → conversaciones → mensajes para calcular por cada lead: - Tiempo hasta primera respuesta humana (excluye bot) - Si el cliente respondió después del mensaje del asesor - Agente atribuido: usa el owner del lead CRM si existe; si no, usa el agente asignado a la conversación del contacto (assignee). Esto cubre cuentas donde la asignación se hace a nivel de conversación y no de lead. Devuelve métricas individuales + agregados (promedio, mediana, % contactados, % con respuesta del cliente, breakdown por agente). Úsalo SOLO cuando el usuario pregunte por leads del CRM ("tiempo de contacto de los leads", "qué leads respondieron después de la presentación"). Para los tiempos de Informes, "de dónde sale el promedio de un asesor" o "la conversación que más tardaron", usa analyze_agent_response_times. IMPORTANTE: NO llames esta tool una vez por agente. Llámala UNA SOLA VEZ sin owner_name y usa el campo by_agent del resultado. NO uses find_sla_breaches para esto.

get_overview_report

Workspace overview for a period: KPIs with previous-period comparison (conversations, contacts, messages, resolution, first-response times), daily timeseries, conversation breakdowns by status/channel/priority, and contact attribution by source. Default window: last 30 days.

get_conversations_report

Conversation operations for a period: incoming-volume heatmap (day × hour), first-response time distributions (human, business-hours, AI), AI agent stats and the current SLA queue of waiting conversations.

get_crm_report

Sales results for a period: funnel (conversations → leads → opportunities → won), revenue and average ticket, revenue per owner and per attribution origin, losses by reason, and the lead/opportunity attribution table per dimension.

get_agents_report

Per-agent performance for a period: assigned/resolved conversations, resolution rate and average first-response times. Pass agent (display name) to add their conversation-level first-response breakdown.

create_pipeline

Create a pipeline (kind: 'leads', 'opportunities' or 'mixed'). It is seeded with the default stage set, ready for cards. Requires workspace-setup permission.

update_pipeline_stages

Set the stage layout of a pipeline in one call: pass the FULL ordered list of stage names (or ids). Existing stages are reordered to match; names that do not exist yet are created as open-type stages. Terminal stages (won/lost) must be edited via the API.

create_team

Create a team, optionally with members (by name, email or id) and auto-assignment.

add_team_member

Add an agent (by name, email or id) to a team (by name or id).

remove_team_member

Remove an agent (by name, email or id) from a team (by name or id). The agent keeps their account access; only the team membership goes away.

edit_canned_response

Edita una respuesta predeterminada existente, buscándola por su código corto. Puede cambiar el código, el contenido o ambos. Cada edición queda registrada en el historial de la respuesta.

create_canned_response

Create a canned response (atajo) on an inbox (by name or id). short_code must be unique per inbox.

list_canned_responses

List canned responses, optionally filtered by inbox (name or id).

create_custom_attribute

Define a custom attribute for contacts or conversations. type: text, number, currency, percent, link, date, list (requires values) or checkbox.

list_custom_attributes

List the custom attribute definitions of the workspace.

No parameters.

update_inbox_settings

Rename an inbox (by name or id), toggle auto-assignment or turn the AI agent on/off. Channel credentials are never touched here.

get_ai_config

Read the AI agent configuration of an inbox (by name or id): temperature, memory, RAG settings, enabled tools, bypass labels, handover. The prompt is not readable here.

update_ai_config

Partially update the AI agent configuration of an inbox (by name or id). Only the keys you send change; every write re-runs the plan sanitizer. The prompt is not writable here.

create_knowledge_collection

Create a knowledge-base collection (embedding/chunking config is a fixed platform standard; plan limits apply).

add_knowledge_document

Add a document to the knowledge base from raw text (up to 100KB) or a public https URL (up to 2MB). It is chunked and embedded through the same pipeline as a dashboard upload; the returned status tells whether it is ready. Collection by name or id, optional.

search_knowledge

Semantic search over the account's knowledge base — the same RAG retrieval the AI agent uses (embeddings + similarity threshold).