AI NewsDev toolsAnnouncement

Twenty v2.50 adds AI call summaries to its open Salesforce alternative

Twenty v2.50.0, released 21 October, adds AI call summaries for Microsoft Teams recordings and splits app permissions into workspace apps and user apps, in a 59,424-star open-source CRM positioned as an alternative to Salesforce.

AI News

Editorial2 min read

LinkedInX
GitHub social card for the twentyhq/twenty repository

Image: GitHub

Why it mattersA sales or operations team buying a CRM gets another option this week where meeting summaries, agents and workflow run on their own copy of the code, which is a different purchase from a seat on a closed platform.

Twenty, a four-year-old open-source CRM that bills itself as "the open alternative to Salesforce, designed for AI", released version 2.50.0 on 21 October. The repository is at 59,424 stars and 9,679 forks on GitHub this morning, and sits at number 17 on Trendshift's daily list of trending repositories as of 22 October. The release is tagged twenty/v2.50.0 and the full change list contains 131 merged pull requests, according to the release notes on GitHub.

The release's own notes name AI call summaries as the headline Teams integration feature for v1. The system takes a Microsoft Teams call recording that lands as an activity in Twenty, generates a transcript summary from it, and attaches that summary to the contact and deal it belongs to. A separate command generates summaries for Teams call recordings and calendar events on demand.

What else changed

The release renames application permissions as "Workspace Apps and User Apps", with the rules now enforced on the server. A workspace app is shared across the people in a workspace; a user app is a personal connection one person made. The split is a breaking change for a workspace that was mixing the two. Built-in account apps now show up in the Installed tab for easier discovery, and workspace admins can switch to a new REST metadata API format through the Lab.

Billing moves to per-seat workspace credits. The release notes name "Agent Engine" improvements too, with "better conversation management and reduced polling", and a campaign send preview that shows server-side blockers before a campaign is sent. The rest is interface work: field default icons, radio buttons that no longer shift on selection, stable grouped table rows, and column resizing that works in right-to-left layouts.

Where it fits

Twenty's position is a CRM you can run yourself, under the AGPLv3, with the agents and the chat built in. The project's own description names Salesforce as the comparison. The tech stack is TypeScript, NestJS, PostgreSQL, Redis, React and GraphQL. A workspace stores custom objects and fields through a code-first metadata model rather than a point-and-click configurator, which the project describes as its difference from the incumbent on the engineering side.

The repository was created on 1 December 2022 and the last push was 22 October 2026. The pull-request numbers in the v2.50 release notes run above 18,800, so the project is adding features at a steady pace. Trendshift's ranking says the repository is picking up stars today; the 59,424 figure above is the total since 2022.

For a buyer who wanted to try an open-source CRM but hit "designed for AI" as a tagline on every page, this release names the first AI feature a seller actually wants: a one-line summary of what a prospect said on the call, attached to the record, without the sales person doing it by hand. Whether the summary is accurate enough to act on is a question the buyer has to answer with their own Teams recordings, and nothing in the release notes claims it is.

Source

SourceGitHub

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project