# Trustpilot MCP server for support teams

> A Trustpilot MCP server for support teams: everyone can read reviews in Claude, but a reply is public, so only named roles may post one.

**TL;DR** A Trustpilot reply is public the moment it posts. Elaichi's Trustpilot connector lets a support team read and triage reviews in Claude or ChatGPT, while an admin keeps the reply and delete-reply tools on a smaller role. Each person connects their own Trustpilot business user, so the reply runs on the right token, and the audit log records who asked for it.

## What should a Trustpilot MCP server for support teams allow?

Yes, a support team can use Trustpilot in Claude without everyone posting replies. Connect Trustpilot in Elaichi with OAuth, add https://api.elaichi.ai/mcp in Claude, ChatGPT, Cursor or Microsoft Copilot, and hold the wider support team to seven named read tools with a restriction on their role. A smaller replier role gets one more tool, create_a_trustpilot_review_reply.

The reason is simple. Reading a review, private fields such as the customer's email and order reference included, stays inside the support team. A reply is the company speaking in public, and a customer or a journalist can quote it. MCP (Model Context Protocol) is the standard way an AI assistant calls tools in other apps, and the [Trustpilot connector](/connectors/trustpilot/) gives a client 83 tools, so the line between reading and speaking has to be drawn on purpose.

## What does Trustpilot require before its API works?

An API add-on and a business user. Trustpilot says the API module is open to Trustpilot for Business users who sign up to it, sold in yearly call tiers, with Tier 1 included on the Enterprise plan (checked November 2026, [Trustpilot API module overview](https://developers.trustpilot.com/api-module-overview/)).

Two kinds of credential then split the API in half. Public endpoints, such as latest reviews and likes, take an API key. The private endpoints take a business user OAuth token. Replying, deleting a reply, editing tags and the Find Reviewer request all sit in that private group (checked November 2026, [Trustpilot service reviews API](https://developers.trustpilot.com/service-reviews-api/)). Elaichi's Trustpilot connector signs in with OAuth, so a person's business user stands behind each connection.

## Whose business user sends a Trustpilot reply?

The one that connected. In Elaichi, every call through a connection uses the account that connection was made with, whoever is calling. A connection an admin shares with the whole support team sends every reply on that admin's token.

For a public reply that is rarely right. Have each agent who may reply connect their own business user, and keep those connections unshared. The reading team can share one connection, since a read is not attributed to anyone in public.

Over MCP, Elaichi does not pause a write to ask first. If you want a second pair of eyes on each reply, turn on confirmation in Claude or ChatGPT, or make it part of the team routine.

## Which Trustpilot tools should every support agent reach?

Seven reads, and nothing that speaks:

- trustpilot_business_units_list_reviews and trustpilot_service_reviews_list_latest for what came in
- get_single_trustpilot_service_review_by_id and trustpilot_service_reviews_get_private for one review in full
- trustpilot_business_units_list_private_reviews for private details across reviews
- list_all_trustpilot_review_tags for how the team already labels them
- trustpilot_consumers_list_reviews for a reviewer's history

In Elaichi, connected tools are never listed one by one, however few there are. An agent asking for this week's one-star reviews triggers a search, then a call to the latest-reviews tool.

## Which Trustpilot tools belong only to repliers, or to nobody?

The replier role gets the seven reads plus create_a_trustpilot_review_reply. Most teams let everyone draft with the model and let one replier post it.

Keep these off both roles unless someone clearly owns them. trustpilot_review_replies_bulk_delete removes replies in batches. The three tag tools, trustpilot_review_tags_set, trustpilot_review_tags_add and trustpilot_review_tags_remove, change how reviews are labeled, which may feed reports. trustpilot_invitations_create_email sends invitation emails that reach customers directly. delete_a_trustpilot_catalog_product_by_id deletes a catalog product.

Two more are easy to miss. trustpilot_private_product_reviews_create_conversation and create_a_trustpilot_conversation_comment start or continue a conversation with a reviewer. That is also the company speaking.

## Allow rule or block rules for the Trustpilot roles?

An allow rule. Holding a role to reads takes either an allow rule naming those reads, plus allow rules for every other app the role uses, or block rules naming every non-read tool. The connector has 83 tools, and the non-read set is long and includes replies, tags, invitations, conversations and deletes. Seven names are easier to keep correct than dozens.

The catch is that once a role carries any allow rule, every connector and tool none of its allow rules names is blocked for that role. So the reader role also needs allow rules for the help desk and chat apps it uses. Custom roles and restrictions are part of Gold and Black, so a Personal workspace cannot do this split. In Elaichi, restriction targets are role or user only; the organization default is the absence of a rule, which means allow everything, so create a Support Reader and a Support Replier role and put each person on one. The change takes effect within about two minutes.

## How do you check the split, and what does the log keep?

Test both roles before launch. A reader should see recent reviews and be refused on the reply tool. A restricted tool is left out of the tools Elaichi offers a client and cannot be called. Search still names it, flagged restricted, with no schema. A replier should post one test reply on a review of your own.

The audit log keeps one entry for each connected-tool call that reaches execution (a call refused earlier writes none). Each entry names the person, the tool, the connection and the MCP client. When a customer asks who wrote a reply, that entry is the answer, because Trustpilot's reply call takes a business user token and not a name. An agent who needs one reply can file an access request, which an admin decides under Governance. [Governance in Elaichi](/blog/mcp-governance/) explains the layers.

## What if one person answers every Trustpilot review?

Then you do not need role rules. A founder or a single support lead who replies to every review can use one connection in one client and draft before posting. The split starts to matter once two or more people can ask an AI client to reply on the same account.

For the principle behind the reader and replier roles, [least privilege](https://en.wikipedia.org/wiki/Principle_of_least_privilege) says each user should reach only what their purpose needs (checked November 2026). A team that already runs a help desk in the same client can read [how support teams use Claude with Zendesk tickets](/blog/support-team-claude-zendesk-tickets/). Open the [Trustpilot connector page](/connectors/trustpilot/) for the full tool list, and compare [one tool against a whole app](/blog/per-tool-vs-per-app-restrictions/) when you write the rules.

## FAQ

### Can an AI assistant post replies to Trustpilot reviews?

Yes, through Trustpilot's reply endpoint. Trustpilot documents that posting a reply is a POST to a private reviews path for a review ID, authorized with a business user OAuth token (checked November 2026, https://developers.trustpilot.com/service-reviews-api/). In Elaichi the matching tool is create_a_trustpilot_review_reply. Restrict it to the roles that may speak for the brand.

### Does Trustpilot let me delete a reply through the API?

Trustpilot's service reviews API documents a delete on the same reply path, using the same business user token (checked November 2026, https://developers.trustpilot.com/service-reviews-api/). Elaichi's connector has trustpilot_review_replies_bulk_delete for removing replies in batches. Keep it off any role that only needs to read or draft.

### Do I need a paid Trustpilot plan to use the API?

Trustpilot says API access is an add-on module that Trustpilot for Business users can sign up to, sold in yearly call tiers, with Tier 1 included on its Enterprise plan (checked November 2026, https://developers.trustpilot.com/api-module-overview/). Check what your own account has before planning a rollout.

### Will Elaichi ask a person before a Trustpilot reply goes out?

No. A client holding the "Use your connected apps" grant can post a reply as soon as it calls the tool. Any confirmation comes from the client itself, so the role's restriction is the one check Elaichi makes before the reply goes out.

## Read next

- [Zendesk for support agents, minus bulk deletes](/blog/support-team-claude-zendesk-tickets/) — Zendesk for support agents in Claude: let them read and update tickets, block deletes and bulk sends, and give the lead one exception.
- [Restrict one AI tool or the whole app? Six cases](/blog/per-tool-vs-per-app-restrictions/) — Restrict one AI tool when a role needs part of an app, and block the whole app when it needs none of it. Six cases, and what new tools do to each rule.
- [MCP governance: who may do what, in which app](/blog/mcp-governance/) — MCP governance decides who connects which app, which tools each role calls, whose account a call uses, what is logged and how access ends.
