Skip to content

How can our sales team use Chorus calls in Claude?

Chorus calls in Claude help a sales team prep deal reviews. Scope the Chorus token, restrict exports and deletes, and give reps a summary-only tool.

Uday Gajavalli 7 min read
A sales enablement lead asking an AI chat client for a call recap, with a Chorus call summary and its action items beside the chat

Why an enablement team wants Chorus calls in Claude

Emily Carter runs sales enablement for a team of account executives. Every Monday she preps two deal reviews and one coaching session. She needs last week's calls on two open deals, what the buyer pushed back on, and which rep let the pricing question slide. All of it sits in Chorus, one recording at a time. Chorus calls in Claude put those lookups in the chat where she already has the deal notes and the forecast.

MCP (Model Context Protocol) is the standard way an AI assistant calls tools in other apps. ZoomInfo documents its own MCP server at https://mcp.zoominfo.com/mcp (docs.gtm.ai, checked November 2026). Its tool list covers company and contact search, enrichment, intent and audiences, plus research agents that blend in CRM and conversation data (docs.gtm.ai, checked November 2026). That list names no tool for listing Chorus calls or reading one transcript. No MCP server is documented in Chorus's API docs (api-docs.chorus.ai, checked November 2026).

Elaichi's chorus connector calls the Chorus API with one API token. It carries call reads, media downloads, exports, coaching writes and deletes. Recorded calls hold customer speech, names and email addresses. So the real work is choosing which of those tools each role gets.

Whose recordings does a Chorus API token see?

A Chorus API token sees what its own user sees, and nothing more. Chorus creates tokens per user, on the Personal Settings page, and only for users whose role is allowed API access in roles and permissions settings (api-docs.chorus.ai, checked November 2026). During the early access period, the same docs say to ask your Chorus customer success manager for the token.

Three Chorus controls carry through to the API (api-docs.chorus.ai, checked November 2026). Recordings marked private are not returned. If data access control is on, the token user's settings apply, so reading every recording needs a user who can see every recording. Role permissions on actions in the Chorus app also apply to the same actions through the API.

That makes the token's owner the first decision. Do not create it from an admin who sees the whole company. Create it from an enablement user whose Chorus access covers the teams Emily coaches, and no more. A call marked private stays out of every answer Claude gives.

What does one recorded call carry?

One call carries far more personal data than a deal review needs. In Chorus's API reference, a fetched conversation includes the account, the deal, the owner, the participants, a summary and action items (api-docs.chorus.ai, checked November 2026). Its recording object holds the utterances, each with the speaker's name and the words they said. The engagement list's sample response shows each participant's name, email, title and company.

Elaichi's chorus connector reaches that data through these tools:

Kind Example tools For the enablement role
Call lists list_all_chorus_engagements, list_all_chorus_conversations Allow
One full call, transcript included get_single_chorus_conversation_by_id Allow
People and coaching data list_all_chorus_users, list_all_chorus_teams, list_all_chorus_scorecards, list_all_chorus_moments Allow
Coaching writes create_a_chorus_moment, chorus_playlists_create_moment Allow
Audio and video get_single_chorus_media_by_id Restrict
Exports chorus_conversations_export, chorus_reports_export, chorus_scorecards_export Restrict
Live-call and CRM writes chorus_conversations_join, chorus_sales_qualifications_writeback_crm, create_a_chorus_upload Restrict
Deletes delete_a_chorus_conversation_by_id, chorus_engagements_bulk_delete, chorus_conversations_bulk Restrict

Chorus's API reference has no separate transcript endpoint (api-docs.chorus.ai, checked November 2026). The transcript sits inside the conversation, which get_single_chorus_conversation_by_id returns. So that one tool decides who reads customer speech word for word.

Which Chorus reads are really exports?

The three export tools are the trap. Elaichi's catalog tiers chorus_conversations_export, chorus_reports_export and chorus_scorecards_export as reads, though each one exports data from Chorus. The conversation export emails download links to call audio and video. The report export emails a CSV link that stays valid for 7 days. Once the links reach an inbox, Elaichi's rules no longer apply to them.

So do not build the allow rule by picking every tool labeled read. Name the reads you mean. Then add block rules on the exports and on get_single_chorus_media_by_id, which downloads the recording itself.

Some writes need the same care. chorus_conversations_bulk is tiered write, yet its body accepts a list of conversation ids to delete and a force-delete flag. chorus_conversations_join joins a live call from an invitation link. create_a_chorus_login takes an email and a password. None of these belong in a coaching workflow.

Writing the allow and block rules for enablement

One allow rule on the enablement role names the Chorus reads and the two moment tools, and nothing else from Chorus. In Elaichi, restriction targets are role or user only; the organization default is the absence of a rule, which means allow everything. So the enablement team gets its own role, and each member holds that one role.

The same rule, or a second allow rule, must also name every other app the role uses, as whole connectors. In Elaichi, once a role carries any allow rule, every connector and tool none of its allow rules names is blocked for that role. A CRM or Slack connector left off the list is lost to the role. Do not name the Chorus connector whole and some of its tools in one rule, because Elaichi refuses that overlap.

Then add block rules on the exports, the media download, the join, upload and CRM write-back tools, create_a_chorus_login and every delete. Within a layer, a block always beats an allow. If someone later widens the allow rule to all of Chorus, those tools stay restricted. A restriction change takes effect within about two minutes. Per-tool versus per-app restrictions compares the two shapes.

Emily connects the chorus connector once with her token and shares the connection at use with the enablement team only. A shared connection runs on its owner's credential, and Elaichi records which member made each call. Members then add https://api.elaichi.ai/mcp in Claude and sign in with their own account, as connecting Elaichi to Claude describes.

A summary-only view of Chorus calls for reps

Reps need the gist of a call, not the transcript or the buyer's email address. A synthetic tool gives them that. In Elaichi, a synthetic tool is a chain of 1 to 20 steps that the AI client sees as one tool. Each step calls a tool on a connection, and an optional output template shapes the final result.

Emily builds chorus_call_recap with one step: get_single_chorus_conversation_by_id, fed the conversation id the rep supplies. Its output template returns the call's deal, owner, summary and action items. It leaves out the utterances and the participants' email addresses. The model receives the recap, not the whole conversation object.

She adds the tool to a toolbox and shares that toolbox with the reps at use. She does not share the Chorus connection with them. A member who cannot use a connection reaches it only through stored toolboxes they can use that pin it. So the raw Chorus tools never appear in a rep's All tools toolbox, and the recap is their one path to Chorus.

Two cautions keep this honest. Do not put a block rule on get_single_chorus_conversation_by_id for the reps' role. Each synthetic step passes the restriction check for the person running it, so that block would stop the recap too. If the reps' role carries allow rules, they must name that tool. Second, the template limits fields, not calls. A rep can still ask for any conversation Emily's token can see. And Emily's synthetic tool cannot be transferred: if she leaves, offboarding deletes it and the reps' toolbox entry stops working. Give the recap to an owner who is staying.

Frozen parameters do little here. The Chorus list tools take only a page size and a cursor, so there is no team or owner filter to lock. Locking a wire's payee shows where frozen values fit.

How to check the setup before reps rely on it

Test from one account on each role. As an enablement member, ask Claude for last week's calls on one deal, and the read should run. Then search for "export" and "delete". 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.

As a rep, run the recap on a known call and confirm the answer has no transcript lines or email addresses. Then ask Claude for the full conversation. With no share on the connection, the rep has no tool that returns it.

Then read the audit trail. Elaichi writes one entry for each connected-tool call that reaches execution (a call refused earlier writes none). A direct read records the tool, the connection and the one path argument that names the object, as the target id, and nothing else about the arguments. For a call read, that target is the conversation id. A recap run writes one synthetic_tool.executed row with its status, duration and step count, never its input or output. What an AI audit log must capture lists what to check in those rows.

When someone needs a restricted tool, such as one export for a quarterly review, they file an access request. An admin approves it in the console, under Governance, then Access requests. The grant lifts that one tool for that one person, and the role is not edited.

When you do not need Elaichi for Chorus

If the team works inside Chorus all day, ZoomInfo's own product may cover the coaching half. ZoomInfo's conversation intelligence page says every call is recorded, transcribed and searchable, with pricing, competitor and commitment moments flagged (zoominfo.com, checked November 2026). It also says every call is scored against your framework, and that an agent can be asked what was promised. For account research from ZoomInfo data, ZoomInfo's MCP server signs each person in with their own ZoomInfo account. ZoomInfo's guide says agents can only do what that account allows (docs.gtm.ai, checked November 2026).

A single enablement lead testing a few queries with her own token may not need a control plane yet. When you can skip an MCP gateway covers that case.

Elaichi earns its place when the question crosses apps and roles. The call is in Chorus, the deal is in the CRM and the follow-up is in email. With Elaichi, one endpoint serves Claude, ChatGPT, Cursor and Microsoft Copilot alike. One set of rules says who reads transcripts, and one audit trail shows who asked. The per-rep CRM side is in HubSpot deals in Claude for a sales team. Browse the 900+ connectors for the rest of the stack.

FAQ

Frequently asked questions

Does Chorus have its own MCP server?

No MCP server is documented in Chorus's API docs at api-docs.chorus.ai (checked November 2026). ZoomInfo documents a ZoomInfo MCP server at https://mcp.zoominfo.com/mcp for its company and contact data and research agents, and its tool list names no tool for listing Chorus calls or reading a transcript. Elaichi's chorus connector calls the Chorus API instead.

Whose permissions does a Chorus API token carry?

The user's who generated it. Chorus's API docs say a token belongs to a registered Chorus user, that user's data access control settings apply when enabled, and their role permissions on actions apply to the API too. Recordings marked private are not returned (checked November 2026).

Can Claude read a Chorus call transcript?

Yes, through a connector. Chorus's Fetch a specific conversation response includes a recording object whose utterances carry the speaker's name and the words spoken (checked November 2026). In Elaichi's chorus connector that read is get_single_chorus_conversation_by_id. A synthetic tool with an output template can hand reps the summary and action items without the utterances.

Why restrict the Chorus export tools if they are reads?

Elaichi's catalog tiers chorus_conversations_export, chorus_reports_export and chorus_scorecards_export as reads, but each exports data from Chorus. The conversation export emails download links to call audio and video, and the report export emails a CSV link valid for 7 days. Name them in block rules rather than allowing every read.

Put AI to work in your own apps

14 days of Black, no credit card. Start with one app and one team.

When the trial ends, nothing is deleted. You move to the free Personal plan, and connections, roles and the audit log stay where they are.

Add one link to your AI app

The same link for everyone on your team.

https://api.elaichi.ai/mcp
  • ChatGPT
  • Claude
  • Cursor
  • Microsoft Copilot