Top 10 Customer Support & CRM MCP Servers for AI Agents (2026)

jira_search(jql, limit?: number, fields?)

Search issues using JQL syntax with field filtering.

jira_get_issue(issue_key, expand?, comment_limit?: number)

Retrieve issue details, SLA status, and recent comments.

jira_create_customer_request(service_desk_id, request_type_id, request_field_values: object)

Submit a new customer service desk request.

jira_add_comment(issue_key, comment, is_internal?: boolean)

Add a public reply or internal work note to a ticket.

Limitations

• Exposing all 98 tools without filtering toolsets consumes excessive tool definition tokens.
• Attachments and issue images return base64 payloads that increase token usage.

search_help_scout(query)

Find available Help Scout operations matching support intents.

describe_help_scout(names: array)

Retrieve exact JSON input schemas for selected operations.

read_help_scout(name, arguments: object)

Execute read queries across conversations, customers, tags, or reports.

write_help_scout(name, arguments: object, confirmedForConversationId?: number, dryRun?: boolean)

Execute mutations with conversation ID matching and optional dry-run preview.

Limitations

• Gateway pattern requires multiple tool call turns to search, describe, and execute unfamiliar operations.
• Requires OAuth client credentials and acts as a shared application identity across all sessions.

run_soql_query(query, targetOrg?)

Execute SOQL queries across Cases, Contacts, Accounts, and custom objects.

list_all_orgs(all?: boolean)

List authenticated Salesforce scratch orgs, sandboxes, and production environments.

list_tools

Query dynamically registered tools and active toolsets.

enable_tools(tools: array)

Dynamically enable specific Salesforce tooling operations on demand.

Limitations

• Focuses primarily on developer tooling and SOQL rather than high-level CRM case workflow helpers.
• Requires pre-authenticated org credentials managed via the Salesforce CLI.

search_tickets(query)

Search support tickets by query filters, email, or status.

get_ticket_conversation(ticket_id: number)

Fetch all incoming messages, outgoing replies, and internal notes on a ticket.

create_ticket_reply(ticket_id: number, body)

Post a customer-visible reply to an existing support conversation.

create_ticket_note(ticket_id: number, body, private?: boolean)

Attach an internal collaboration note visible only to agents.

Limitations

• Does not implement proactive rate limit throttling on heavily throttled search endpoints.
• Returns full un-truncated conversation bodies that consume context on long email chains.

get-deals(status?, stageId?: number, limit?: number, cursor?)

Retrieve deals with stage filters, value ranges, and cursor pagination.

get-deal-notes(dealId: number, limit?: number)

Fetch customer notes and communication logs for a specific deal.

search-all(term, itemTypes?: array)

Search across deals, persons, organizations, leads, and files.

move-deal(dealId: number, targetStageId: number, expectedCurrentStageId?: number, execute?: boolean)

Move a deal to a new stage with optimistic concurrency checks and preview mode.

Limitations

• Read tools are intentionally scoped to core pipeline entities and exclude custom deal metadata creation.
• Requires global npm installation or Docker execution as it does not publish an npx binary.

list_people(search?, limit?: number, filter?: object, companyId?)

List and filter customer contacts with relational depth controls.

create_person(firstName, lastName, email?, companyId?)

Create a new person record with associated company links.

list_companies(search?, limit?: number, filter?: object)

Query company records, revenue metrics, and ideal customer profile data.

create_opportunity(name, stage, amount?: number, companyId?)

Create a new sales opportunity attached to an account.

Limitations

• Workspace API keys grant administrative read and write access without granular per-user role restrictions.
• Batch creation operations are capped at 60 records and are not executed in atomic transactions.

list_incidents(limit?: number, state?, assigned_to?, query?)

Query active IT incidents by priority, assignment group, or state.

create_incident(short_description, urgency?, impact?, caller_id?)

Create a new incident record in ServiceNow.

add_comment(sys_id, comment, is_work_note?: boolean)

Append a public comment or internal work note to an incident.

approve_change(change_sys_id, comment?)

Submit an approval vote on an active change request.

Limitations

• Workflow script tools require custom Script Execution API endpoint configuration on production instances.
• Pydantic string error formatting loses structured diagnostic context during Table API validation errors.

get_tickets(page?: number, per_page?: number, sort_by?)

List recent tickets with pagination parameters.

get_ticket(ticket_id: number)

Retrieve full details and custom fields for a specific ticket.

get_ticket_comments(ticket_id: number)

Fetch the chronological discussion history on a support ticket.

create_ticket_comment(ticket_id: number, comment, public?: boolean)

Add a public customer reply or internal note to a Zendesk ticket.

Limitations

• Lacks a dedicated search endpoint, requiring sequential pagination over ticket lists.
• Does not include built-in automatic backoff when encountering Zendesk API rate limits.

conversations_awaiting_reply(max_pages?: number)

Retrieve live chat conversations prioritized by longest customer wait time.

get_rich_context(session_id)

Fetch a complete customer dossier combining messages, profile details, and session metadata.

send_message(session_id, content, type?)

Send a live chat response or internal note to a visitor.

assign_conversation(session_id, operator_id)

Route a live conversation to a specific support operator.

Limitations

• Requires building from source as it is not packaged on npm or PyPI.
• Destructive actions such as conversation deletion are enabled without separate confirmation gating.

hubspot_search_data(query, limit?: number)

Perform semantic vector searches across cached customer contacts and conversation threads.

hubspot_get_tickets(criteria?: object, limit?: number, max_retries?: number)

Retrieve support tickets with customizable filter criteria and retry controls.

hubspot_get_contact(contact_id)

Fetch comprehensive properties and activity history for a specific contact.

hubspot_create_company(name, domain?, properties?: object)

Create a company record with automated domain duplicate prevention.

Limitations

• Relies on a fixed container storage path for FAISS embedding indexes, making Docker execution necessary.
• Downloads local transformer embedding models on first startup, introducing initial container launch latency.