Niches · Customer support & helpdesk
Read a customer's history before acting
Four or more companies' docs show a full fit.
What counts as done
The conversation thread or a filtered list of matching tickets, with only the fields asked for.
Measured, for our own tests: in at most two calls and under 4,000 tokens per ten results. No docs we read state these figures, so they don't count against a listing here.
How agents ask for it: what has this customer asked us before
, find tickets like this one
, pull the conversation for ticket 4821
, search tickets about login failures this month
Does the job 6
The docs show it meets the test above, and nothing we read says otherwise.
| Listing | Why |
|---|---|
| Front API + MCP BB FrontApp, Inc. | search_conversations takes full text plus inbox, teammate, team, tags, status and date filters; read_conversation pages the timeline with a limit; a read-only scope exists. Core search costs 40 per cent of the company limit. source |
| Intercom API + MCP BB Intercom, Inc. (with Intercom R&D Unlimited Company in Ireland) | MCP search takes a query DSL with free text, and search_conversations filters by state, assignee and timing with cursor paging; REST conversation fetch has display_as plaintext. Permissions are granted per app. source |
| Kustomer D Kustomer, LLC | POST /v1/customers/search takes and/or criteria with queryContext conversation, date ranges and sort, and fields limits returned attributes to up to 20. Keys are limited to named org.permission roles. source |
| LiveAgent D Quality Unit, s. r. o. | search_tickets takes a full-text query plus status, department and agent filters, returns 10 to 50 compact tickets per page with a cursor, and get_ticket_messages pages 20 messages at a time. source |
| Plain API + MCP B Not Just Tickets Ltd | searchThreads runs full-text search over titles, message contents and customer names with a ThreadsFilter, and GraphQL returns only the fields selected. Machine-user keys carry named permissions. source |
| Zoho Desk B Zoho Corporation Private Limited | Search Tickets takes _all for full text plus status, priority, assignee, tag, contact and created or modified ranges, and a fields parameter trims each record. OAuth scopes can be read-only. source |
Does the job, with a limit 15
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| Chatway MCP C Poptin Ltd | search-conversations, get-conversation and list-conversation-messages with cursor paging cover history, and inboxes and permissions are chosen at connect, but the Free plan keeps 30 days of history and field selection isn't documented. source |
| Chatwoot API D Chatwoot Inc. | Conversations List takes q (message text), status, labels, team and inbox, and Contact Conversations gives history, but there is no field selection and user tokens carry the full role. source |
| Crisp API + MCP D Crisp IM SAS | List Conversations takes search_query by text, segment or filter, and MCP has search_conversations, but pages are 20 to 50 whole conversations and the MCP token holds the website tier with no scopes; REST plugin tokens can be scoped. source |
| Dixa D Dixa ApS | POST /v1/search/conversations searches by text, filter or both, and end-user conversation lists exist, but records come whole. acc-31 hit a 404 on the MCP activity-log tool, fixed in July 2026. source |
| Drag E DragApp.com Ltd | search_threads searches across boards and filter_threads narrows by label, assignee or status, but the API key has full workspace access with no read-only option and field selection isn't documented. source |
| FreeScout F FreeScout | List Conversations filters by mailbox, folder, status, state, assignee, customer email and subject text, but not body text; embed defaults to threads, making records large, and keys have no scopes. source |
| Freshdesk API + MCP D Freshworks Inc. | Filter Tickets queries fields (status, priority, tag, agent, dates) but not body text, returns 30 a page for at most 10 pages, and excludes archived tickets; acc-18 calls the built-in search weak. source |
| Gladly C Gladly Software, Inc. | Customer profiles are found by email, phone or external id and their conversations listed, but there is no full-text conversation search, and lists stop at 100 conversations with no next page. Tokens have no scopes. source |
| Gorgias API + MCP C Gorgias Inc. | MCP search_tickets matches subjects, messages and customer details, but filtering by status or assignee needs a saved view, REST /api/search does not cover tickets, and the MCP has no read-only mode. source |
| Help Scout API + MCP B Help Scout PBC | List and Search Conversations take an advanced query over subject, body, email, tag, assignee and dates, and the MCP server is read-only by design, but threads are embedded whole and OAuth tokens have no scopes. source |
| HelpDesk D Text, Inc. | GET /v1/tickets takes query plus filters for status, requester, tags, priority, teams and date ranges with cursor paging, but no field selection, and a token inherits its creator's role. source |
| Missive API + MCP C Heliom Inc. | MCP Read conversations searches across accounts under a read-only grant, but REST List conversations must be filtered by mailbox, label or team and has no text query; REST tokens are personal with full access. source |
| Pylon API + MCP C Pylon Labs, Inc. | POST /issues/search takes fuzzy search_text intersected with a filter and cursor paging, and MCP has search_issues, but no field selection is documented and tokens have a role, not scopes. source |
| Zammad E Zammad GmbH | Endpoint, global and condition-based search run over Elasticsearch with paging, and a token can be narrowed to ticket.agent, but records have no field selection beyond expand off. source |
| Zendesk Support API B Zendesk, Inc. | GET /api/v2/search has a query syntax over status, dates, requester and tags, and OAuth scopes allow read-only, but records come whole with no field selection; acc-13 measured about 9KB a record. source |
Doesn't do this job 1
Listings in the category that do something else, with the reason.
| Listing | Why |
|---|---|
| Landbot MCP server E Landbot | No conversation or contact read tools are in the published tool list. source |
What stops agents finishing
- search returns whole records at kilobytes each with no field selection
- lists capped with no next page, such as 100 conversations
- MCP token holds the whole workspace with no read-only option
Around it
Before: chat.user-asks, support.webhook-on-new. After: support.draft-reply-for-review, support.summarise-and-report, issue-tracking.create.
Where this comes from
Weight 15 of 100 in Customer support & helpdesk: 7 of 42 accounts (acc-13, 14, 18, 31, 33, 34 and acc-12); the job most hosted MCP servers are limited to, and the one read-only servers from Help Scout and Intercom are built for
The same as data: /api/v1/use-cases/support.research-context.json, or the MCP tool get_use_case with id support.research-context. Other jobs in this category: /api/v1/niches?category=support.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.