Agentic AIEnterprise AI

Digital.ai Release 26.3 Puts AI Agents Through Enterprise Approval Workflows

Digital.ai is extending its release management software so developers can initiate software deliveries from AI coding tools while keeping company approval requirements in place. In its November 18 announcement, the company said Release 26.3 brings agent orchestration, a Model Context Protocol (MCP) server, and an AI release assistant together within governed delivery workflows.

The software coordinates the steps between development and production, including automated testing, security checks, and human decisions. Digital.ai says participating agents operate under the policies, approvals, and audit requirements applied to other release activities. Where a workflow requires a person to approve a change, that approval remains necessary.

The update addresses a practical issue for platform engineering and release teams: how to let AI tools participate in delivery without losing control over who can change a workflow, start a release, or authorize its progress.

Digital.ai Release 26.3 connects AI coding tools to release workflows, with permission controls, human approvals, and assistance for pipeline investigation.

Release controls reach into AI coding tools

The Release MCP Server connects AI development tools with release workflows. Digital.ai names Cursor, Claude Code, and Codex among the tools developers can use to initiate and analyze releases. The connection brings release operations into the environment where developers are already working.

The company’s MCP plugin documentation specifies what that access can involve. Read operations include retrieving release details, examining templates, checking variables, and reviewing activity logs. When write access is enabled, available tools include creating a release from a template, starting it, changing tasks, and managing gate conditions.

Access follows the permissions of the user whose personal access token authenticates the connection. An agent can reach only the resources that user is authorized to access. Administrators can disable the MCP server or enable a read-only mode that hides tools capable of changing or running release objects.

Those settings matter when configuring an AI client. The documentation says the bundled MCP server is enabled by default and supports both read and write tools unless read-only mode is selected. The scope of an agent’s activity therefore depends on the user’s permissions and the server configuration.

AI assistance for release investigation

Release Assistant is aimed at the teams responsible for maintaining delivery workflows. Digital.ai describes it as a way to analyze pipelines, investigate failures, and identify workflow improvements.

The on-premises documentation provides a more specific description of AI Assistant, the name used in the technical guides. It supports natural-language questions about release status, overdue tasks, blockers, templates, and audit activity. Users can also inspect text-based attachments, such as logs and configuration files, within their existing permissions.

The documented on-premises assistant is read-only. It can inspect existing integration configurations, but it cannot test connections or create configurations. These boundaries distinguish conversational investigation from the write operations available to separately connected MCP clients.

Release notes also describe a task troubleshooting function launched from a failed task’s “Explain error” button. A Reporting Agent can answer questions with charts and tables, but that component is explicitly labeled a technology preview.

Availability builds on earlier capabilities

Digital.ai said Release 26.3 and the three capabilities were available at the time of the November announcement. Its Release Premium product page presents the combined offering as a Premium package and directs existing customers to confirm their current entitlements with the company. It also identifies agent orchestration as an existing foundation that the MCP connection and Release Assistant extend.

The documentation shows that parts of the technology preceded the announcement. July 2026 SaaS release notes already listed a built-in MCP server. The November news therefore brings together capabilities with different release histories, rather than marking the first availability of every underlying component.

For customers upgrading existing installations, the 26.3 notes identify another transition: the standalone Release MCP Server is scheduled to reach its sunset date on December 31, 2026. Customers using that component are instructed to migrate to the built-in server supplied through the Release MCP Tools plugin.

Digital.ai also offers a choice between its model infrastructure and enterprise-approved models. For the on-premises assistant, however, the release notes specify a validated set of Anthropic and OpenAI models, making the supported-model list part of deployment planning.

ERP News Editorial Team
+ posts

We cover ERP and enterprise technology for the people who select, implement, and run these systems.

Pitch a guest article or work with our editorial team on a news story, customer case study, or executive interview.

Editorial Calendar · Pitch a Guest Article · Commission a Story

Advertising Options · ERP News Magazine · ERPNews.com

Editorial Inquiries · [email protected]

Copyright & Usage · Our original editorial content is produced by the ERP News Editorial Team, with expertise in ERP and enterprise technology.

© ERP News. All rights reserved. Except as permitted by law, reproduction, republication, and redistribution require prior written permission or a licence from ERP News, along with clear attribution to ERP News as the source.

ERP News Content Licensing · License full-text ERP News articles for your internal research tools, delivered through a private JSON feed. [email protected]

Shares: