Skip to content

Sales tools by domain ​

The Sales-Management MCP can expose one focused Sales Platform domain or the combined tool set. Use this map to pick the smallest domain for the task; open the full reference only when you need exact parameters, result shapes, pagination, or write semantics.

Use this surface from Claude or Codex during configuration and prototyping. After a workflow is promoted, the Agents routine reaches the same domain through the Sales connector or tool enabled on its executing agent. See the routine-first operating model.

TaskDomainFull reference
Configure sales assistants, instructions, skills, available tools, and test chatssales_assistantsAssistants
Page through clients, manage leads, context, notes, tasks, attachments, labels, and pipeline stagessales_clientsClients, leads, and pipelines
Page through chats and retrieve bounded message historysales_conversationsConversations
Inspect or configure messaging and voice channelssales_channelsChannels
Configure follow-ups, preview an audience, send a one-off message, or place an outbound callsales_outreachOutreach and follow-ups
Manage FAQ knowledge and product integrationssales_knowledgeKnowledge Base
Inspect connected third-party integrations and obtain configuration linkssales_integrationsIntegrations
Inspect account, subscription, token, and usage summariessales_accountAccount and usage
List orders, retrieve one order, and aggregate order statisticssales_ordersOrders
List calendar events, staff, services, and aggregate statisticssales_calendarCalendar
Inspect and map Meta lead formssales_automationsAutomations
Save conversations as eval scenarios and run instruction comparisonssales_evalsTesting and evals

Pagination belongs to the caller ​

List and scan tools expose bounded pages with filters, limits, offsets or cursors where documented. For a large dataset, the routine or skill script should request a page, process it, advance the pagination state, and stop at an explicit total or completion condition. Do not treat an export-style name as permission to request every client or chat in one server call.

For example, an archive routine can page through chats, call get_chat_messages with a per-chat message limit, append normalized rows to its working CSV, and publish the final artifact only after all pages are complete. The script owns the total-record loop; each MCP call stays bounded and retryable.

Effects and testing ​

Read operations touch real connected systems and are appropriate for capability checks. Messages, calls, deletions, and live record changes are external effects: test them on a controlled record or recipient, require stable identifiers and idempotency where possible, and keep automation disabled until the effect and recovery path are accepted.

Complete reference for agent clients ​

Every tool, parameter, default, result, and error is documented in Sales tools reference. Its canonical raw Markdown is available at https://docs.mychatbot.app/agents/sales-tools-reference.md.

See also ​