Lexidesk

Reads the full context of one person, for example to prepare a consultation: the profile, every conversation with case details, custom analysis values, booked appointment, marketing attribution, and the feedback a firm user left, plus the full transcripts of the newest conversations. Pass the clientIds array of one person from search_clients. maxLatestConversationTranscripts controls how many of the newest conversations carry the full transcript; transcriptFormat "compact" halves their tokens. Use includeOnlyFields or excludeFields to keep the conversation objects concise. Fields with a null or empty-list value are absent from the output, except id fields. An absent field means null or empty.

get_client

Explains how to use the Lexidesk MCP well: the recommended call order, how to keep responses concise, where the field definitions are, and how write tools behave. Read it once at the start of a work session. You are currently using the MCP definition that was updated at 2026-10-24.

get_mcp_usage_instructions

Defines the output fields and enum values of the Lexidesk MCP tools: conversation outcomes and reasons, lead qualification, retainer lifecycle and progress states, appointment and attribution objects, and more. Use it when a field or value is unclear.

get_output_field_definitions

Reads the business configuration of one AI assistant (agent): identity, active state, channels, phone numbers, languages, appointment offers, transfer destinations, lead qualification criteria, custom variables, and the knowledge base content. Use it to inspect how the assistant is set up, or to read the current knowledge base before update_knowledge_base. The configuration matches what the dashboard shows; prompt texts are not part of this view. An absent field means null or empty.

get_agent

Returns the change history of the Lexidesk MCP tools, the live updatedAt date, and the current tool names for your role. Your cached definition was updated at 2026-10-24: when the returned updatedAt is newer, or when toolNames differ from the tools you see, your client holds a stale definition — tell the user to refresh the Lexidesk connector or start a new chat. Also call this tool when any tool call fails with "unknown tool".

get_mcp_changelog

Lists the custom analysis fields configured for one AI assistant (agent), so you know which keys can appear in the customAnalysis object of conversations and what they mean. Fields come from two sources: the AGENT configuration and the OUTBOUND_CALL_INSTRUCTION configurations; on a field-name clash the instruction value wins. For the fields that retainer preparation extracts, call get_fields_possible_in_retainer_resolved_variables instead.

get_fields_possible_in_conversation_custom_analysis

Lists the variable fields each retainer variant of one AI assistant (agent) can resolve, so you know which keys can appear in a retainer's resolvedVariables and what they mean. A retainer variant is one kind of retainer agreement the firm sends. Fields with origin RETAINER_CUSTOM_ANALYSIS are extracted by AI from the full chronological client context during retainer preparation; the other origins are resolved by code from the agent, client, conversation, or variant configuration. The resolved values land on the retainer record, not on conversations; for conversation-level fields call get_fields_possible_in_conversation_custom_analysis instead.

get_fields_possible_in_retainer_resolved_variables

Reads one intake conversation in full: the complete message transcript in order, the AI case summary, case details, custom analysis values, the booked appointment, marketing attribution, outcome, lead qualification, diagnostics (message counts, duration, who ended it), and the feedback a firm user left on the conversation. Use after search_conversations to inspect a specific enquiry. Voice conversations include a note when a call recording exists. Use includeOnlyFields or excludeFields to keep the response concise, and transcriptFormat "compact" or transcriptHead / transcriptTail to shorten the transcript. Fields with a null or empty-list value are absent from the output, except id fields. An absent field means null or empty.

get_conversation

Sends a feature wish to the Lexidesk team. Use it when the user asks for data or an action that no tool provides, or wants a product capability that does not exist. Set isFeatureForMcp to true when the wish is about this MCP connection, and to false when it is about the Lexidesk product. Confirm the wording with the user before you send it.

request_feature

Searches the intake conversations the Lexidesk AI assistant handled for the firm: phone calls, website voice and text chats, emails, SMS and WhatsApp threads. Every enquiry a potential or existing client made goes through these conversations, so this is the tool for questions like "what leads came in last week", "did anyone mention a car accident", or "show abandoned calls from yesterday". Each result carries the case summary fields, the custom analysis values, the booked appointment, the marketing attribution, and the feedback a firm user left on the conversation (null when there is none). Use hasFeedback: true to collect every conversation with feedback, for example to review what the AI assistant can do better. The message transcript is not in the results by default; put "transcript" into includeOnlyFields to read transcripts in bulk (transcriptFormat "compact" halves the tokens), or call get_conversation for one conversation. Pass conversationIds to read a known set of conversations. Use includeOnlyFields or excludeFields to keep the response concise. The EMAIL channel holds every email the firm inbox received, including newsletters and notifications: use excludeChannels: ["EMAIL"] when you measure intake quality. All filters combine with AND. Results only cover conversations the signed-in account may see. Fields with a null or empty-list value are absent from the output, except id fields. An absent field means null or empty.

search_conversations

Searches the firm's people directory: every lead, client, and other contact the Lexidesk AI assistant has spoken with, one row per person (merged across channels by phone number, email, or client record). Use this for questions about people rather than individual conversations: "which qualified leads have not booked an appointment", "who contacted us most recently", "find the client called Smith". Each row includes clientIds (pass them to get_client for the full context of the person), contact details, first and last contact dates, conversation count, the highest lead priority seen, the relationship with the firm, and recent qualification signals with conversationIds — pass one to get_conversation to read the enquiry. All filters combine with AND. Results only cover people the signed-in account may see.

search_clients

Searches the outbound campaigns of the firm, newest first. An outbound campaign calls a list of contacts with one AI assistant, with retries, a calling window, and a set of firm numbers to dial from (agentPhoneNumbers). Each result carries the campaign settings and state. Call get_outbound_campaign for the progress counts of one campaign, and search_outbound_campaign_schedules for its contacts. Results only cover campaigns of the AI assistants the signed-in account may see.

search_outbound_campaigns

Former name of search_outbound_campaigns. Call search_outbound_campaigns instead. Searches the outbound campaigns of the firm, newest first. An outbound campaign calls a list of contacts with one AI assistant, with retries, a calling window, and a set of firm numbers to dial from (agentPhoneNumbers). Each result carries the campaign settings and state. Call get_outbound_campaign for the progress counts of one campaign, and search_outbound_campaign_schedules for its contacts. Results only cover campaigns of the AI assistants the signed-in account may see.

search_outreach_tasks

Searches the retainer agreements the Lexidesk AI assistant prepared and sent for the firm, sorted by last update. Each result carries the dispatch state (lifecycle), the signing progress, the payment progress, the hold reason when a human must act, the retainer kind (variant), the client, and the originating conversationId. Use it for questions like "which retainers wait for approval", "what did the client Smith sign", or "how many retainers were paid this month". All filters combine with AND. Results only cover retainers the signed-in account may see.

search_retainers

Searches the contacts (schedules) of one outbound campaign: one row per person the campaign calls, with the contact state, the number of call tries, the last and next try time, the firm number the last try dialed from (lastAgentPhoneNumber), the latest conversationId, and the custom analysis of that conversation. Use it for questions like "who was reached", "which contacts are still waiting", or "which number did we call Smith from". Read a conversation in full with get_conversation. All filters combine with AND.

search_outbound_campaign_schedules

Former name of search_outbound_campaign_schedules. Call search_outbound_campaign_schedules instead. Searches the contacts (schedules) of one outbound campaign: one row per person the campaign calls, with the contact state, the number of call tries, the last and next try time, the firm number the last try dialed from (lastAgentPhoneNumber), the latest conversationId, and the custom analysis of that conversation. Use it for questions like "who was reached", "which contacts are still waiting", or "which number did we call Smith from". Read a conversation in full with get_conversation. All filters combine with AND.

search_outreach_task_schedules

Sends short usage feedback to the Lexidesk team: what the user tried to do with this MCP and whether it worked. Use it only when the user asks to send feedback or agrees to send it. Never send it unasked. Do not include client names or case details.

send_usage_feedback

Updates the knowledge base of an AI assistant (agent): the firm facts the assistant answers from, such as services, offices, hours, and fees. This is a write tool with replace semantics: contentMarkdown and contentForFollowUpMarkdown overwrite the full stored text. Read the current content through get_agent first, apply your edits to it, and send the complete new text. The update creates a new agent version, so the dashboard shows the change history.

update_knowledge_base