Platform
Solutions
Integrations
Case studies
Resources
Pricing Start free Request a demo Log in
Headless conversation analytics

Call quality scores in your BI tool and CRM. Next to your revenue.

Ender Turing's REST API returns the quality scores of your analyzed calls, chats and emails, point by point, with their topics and summaries, and an Ender, an automation rule, can push each new score or summary by webhook. Your integration writes them into your BI tool, warehouse or CRM, joined to your records by an ID both sides hold.

REST API and webhooksPublic guide for AI agents 4.8/5 on G2
How it works3 steps
  1. Create an account for the integration

    An administrator invites a user with the predefined Session API role, or a role of your own, and ticks System/Technical User. Your integration or AI agent sees what that role allows, no more. API access ↗

  2. Hand over the guide

    One public plain-text page tells an AI agent, or the engineer building your integration, how to send the token, filter conversations and get their scores and summaries. Guide for AI agents ↗

  3. Pull each day, or have it pushed

    On a schedule, load yesterday's conversations with their scores, topics and summaries and re-load recent weeks, since a reviewer may add or change a score days later. Or set an Ender, an automation rule, to send each new score or summary to an address you choose. Enders (automations) ↗

What reaches your systems

What Ender Turing's API and webhooks deliver.

Four kinds of data leave with each analyzed conversation, by two paths. Each item is in the guide for AI agents or the Help Center article linked on its card.

1

Quality scores, point by point

The API returns every quality score on a conversation, with its scorecard, its reviewer and the result of each point. Each conversation in a list carries its average score and how many scores it has, and lists filter by manual or automatic scoring.

2

The topics of each conversation

The API returns the topics found in each conversation, from your own topic catalog, so your reports can count them by account, agent or team. One request turns topic, agent and team IDs into names.

3

Summaries in your own format

The API returns each summary with its text and the name of the prompt that wrote it. Summary prompts in Gen AI Studio write plain text or structured JSON, so a summary can hold the fields your reports need; the plan comparison shows which plans include your own prompts.

4

The IDs that join it to your data

Each conversation carries its ID, start time, length, direction, type, agent and team, and the data that came with it from other systems, often a call ID your phone system also writes into a CRM or help-desk ticket. Which IDs arrive depends on how conversations reach Ender Turing. Your system can add its own data to a conversation through the API.

5

Pulled on your schedule

Your ETL job, a script or an AI agent lists conversations with any filter of the Conversations screen, one day or one week at a time. An exact count for the same filter, with the conversations' total length, shows whether the filter works and how many conversations each load should hold.

6

Pushed as each result is added

An Ender, one of Ender Turing's automation rules, calls your webhook when a conversation is added, when the system it came from updates it, or when a score or a summary is added. Its JSON body holds the conversation's details, agent and team, with the summaries, topics and scorecard scores it has by then, and a mapping of your own can reshape it.

What the API leaves out: Transcript text is not part of the published API, and the webhook's body has no transcript section. For what was said, your integration gets a conversation's summaries where they are set up, or an AI agent asks EnderGPT once an administrator has enabled it and its role may create chats. Which plans include API access is in the plan comparison.

Before you build

What your data team needs to know.

Hand these to whoever builds your integration. Each is in the guide for AI agents or the Help Center.

  • Test the join today, in a spreadsheet — Download XLSX on the Conversations screen gives one row per conversation: its topics, agent, team, CRM statuses, a link back, each summary in its own column, and the IDs your phone system or CRM sent with it, kept as text so long IDs stay whole. XLSX Expanded gives each top-level field of a JSON summary its own column. Each contact-center agent's scoring table downloads one row per review with each point's score, the reviewer, comments, the summary and topics.
  • Its own account — an administrator creates it with the Session API role or a role of your own. Keep its credentials in your secrets manager, sign in once and reuse the token until it expires. Deactivating the account ends its access.
  • The same scope as the web interface — every endpoint checks its own permission, and a role limited to one team, or to its own conversations, returns those and nothing else. If data is missing, check the role: a summary prompt set to selected roles shows its summaries to those roles.
  • Names for IDs — topic, agent and team names come in one request the guide describes. Scorecard and point names come with the scorecards, which need a role that may view them, made by copying Session API and adding that permission; the plan comparison shows which plans include custom roles.
  • History one window at a time — load one window per query, a day or a week with fewer than 10,000 conversations (split a busy day into hours), and page each window to the end, one request after another, not in parallel. Re-fetch recent weeks: a reviewer may add or change a score days later.
  • Webhooks to public addresses — check a delivery before it runs: Preview shows a real conversation's body and the full URL it goes to, with placeholders filled in, and Send Test delivers it to your endpoint right away. Give the Ender your endpoint's final, public address: private and internal addresses are refused, and redirects are not followed.
  • Name your client — identify your integration the way the guide shows. If Ender Turing cloud answers with a firewall page instead of data, support adds an exception for your integration from the Ray ID on that page; retrying returns the same page.

API and AI agents · Guide for AI agents ↗

FAQ

What teams ask first.

Does Ender Turing send quality scores, topics and call summaries into a BI tool, a data warehouse or a CRM through an API?

Yes. Through Ender Turing's REST API, your ETL job, a script or an AI agent lists conversations with the same filters as the Conversations screen and gets the quality scores on them that its role may see, point by point, with their topics and summaries. Your integration then writes them into your BI tool, data warehouse or CRM. Through webhooks, an Ender, one of Ender Turing's automation rules, pushes each new score or summary to an address you choose as it is added. Transcript text is not part of the published API.

Do we get each criterion's score, or just a total?

Each criterion's: on a scorecard, each criterion is a point. Every quality score comes with its scorecard, its reviewer and the result of each point, and each conversation in a list carries its average score and how many scores it has. Point names come from the scorecard, so your BI tool can average by point, agent, team, topic or account, over any period.

How do we join conversations to accounts, tickets or deals?

By an ID both sides hold, such as the call ID your phone system also writes into a CRM or help-desk ticket. Your integration lists conversations with the data that came with them from other systems and joins on that ID in your warehouse or BI tool. Which IDs a conversation carries depends on how it reaches Ender Turing, and one call recorded in parts can be several conversations. To find the call behind one ticket, search for its ID and keep only exact matches, upper and lower case included: the search also matches what was said and phone numbers, and finds only IDs in fields whose names end in _id. The Conversations screen's Excel download gives each such ID its own column, so you can test the join in a spreadsheet first.

Does Ender Turing push the data by webhook, or does our system pull it through the API?

Either. Your system pulls through the REST API on its own schedule. Or an Ender pushes: when a conversation is added, when the system it came from updates it, or when a score or a summary is added, the Ender sends the conversation's details, agent and team, with the summaries, topics and scorecard scores it has by then, but no transcript, to your address by POST or PUT. The address can carry a value from the conversation, such as a ticket ID, so the delivery reaches the matching record. Deliveries can be signed with a shared secret so your endpoint can verify them. They can use Basic or token authentication and retry up to 10 times.

How do we backfill past conversations and keep them current?

Load the history one day or one week at a time, keeping each window under 10,000 conversations, and page each window to the end. A window is easy to resume after a failure and to re-run. If an Ender pushes your data, Run Historical applies it to conversations that already exist, up to 20,000 per run. To stay current, re-fetch recent weeks through the API on every run, because a reviewer may add or change a score days after the conversation. An Ender also pushes each new score as it is added.

What do we need to build on our side?

An integration that pulls, or an address that takes the pushes. Your ETL job, a script or an AI agent calls the REST API on a schedule and writes the results into your warehouse, BI tool or CRM. Or an Ender's webhook sends each result by POST or PUT to an address you choose: a service of yours, or a system that updates records with PUT, with a mapping of your own that reshapes the body into the JSON that system expects. On Ender Turing's side, an administrator creates the integration's account in the web interface, and an Ender, set up there or through the API, handles the pushes.

Does the API return transcripts?

No. Transcript text is not part of the published API, and an Ender's webhook body carries none either. Each conversation comes with its details and the summaries, topics and quality scores it has, from your summary prompts, topic catalog, reviewers and AutoQA; a summary prompt can write the fields your reports need as JSON. If your warehouse needs the transcripts themselves, describe the use case to Ender Turing support: the right approach depends on your deployment.

Can an AI agent build and run the integration?

Yes, when it can send web requests. Point it to the public guide for AI agents and give it a token for an account you created for it: the guide tells it how to send the token, write filters, get scores and summaries and when to open the full API reference, and to change nothing you did not ask for. It sees what that account's role allows, no more. An integration that runs every day signs in with that account and reuses the token until it expires. With EnderGPT enabled by an administrator and allowed by the account's role, it can also ask EnderGPT about many conversations, such as why customers cancel, and schedule that question every day, week or month.

Security, hosting regions, sub-processors and response times are documented on the Data Security page; SOC 2 Type II audited, the latest report available under NDA.

More jobs for your AI agent: Ask your calls · Weekly quality review · ten platforms compared.

BI tool, warehouse and CRM

Put call scores, topics and summaries next to your revenue.

In 2026 an enterprise insurer and an enterprise bank moved their conversation analytics from Genesys to Ender Turing and kept Genesys as their telephony; they chose Ender Turing for analytics, automated insights and integration with their data warehouse, BI and CRM. Book a demo and bring the report you want to build, and the BI tool, warehouse or CRM it belongs in.