1. Cross Referencing Support in Smart Docs

An enhanced Cross Reference support has been provided for Smart Docs, allowing users to reference a numbered item, table, or image from one work item inside other work items in the same document. References can be inserted as a hyperlink or plain text, and clicking a source scroll to and highlights the referenced item. References carry through to HTML, Word, and PDF exports and Smart Report previews as clickable.

1. Copy URL for Trace Files

Added a Copy URL option for Trace files to make sharing and accessing files easier. Users can copy the Trace file URL from the file context menu or the overflow (⋮) menu when the file is open.

The copied URL can be pasted into another browser tab or shared with users who have the required access permissions.

1. Copy URL for Baseline Files

Added a Copy URL option for Baseline files to make sharing and accessing files easier. Users can copy the baseline file URL from the file context menu or the overflow (⋮) menu when the file is open.

The copied URL can be pasted into another browser tab or shared with users who have the required access permissions.

1. Signature Options for Smart Reports

Added flexible signature options for Smart Reports, allowing users to generate either Document Only or Document + Work Item Signatures. The option is available to approvers when generating a report from a Review, or from the Smart Doc version on which the Review was created.

1Jobs Tab – Centralized Agent Execution History

The Jobs tab now serves as the master execution history for all agents within a project, giving users a clear, filterable view of how agents are running over time. This makes it easier to monitor automation, identify failures, and understand where user input is needed.

• See every agent run as a job in a dedicated grid, with columns for Job ID, Agent name, trigger event, work item type, repository, status (In Progress, Success, Failed, Terminated), number of open tasks, and timestamp.
• Filter jobs using a flexible filter bar with options for agent name, trigger event, work item type, repository, and status, and a dedicated "Open Tasks" filter that shows only jobs waiting for user input.
• Refresh the grid to pick up new jobs that start while the user is on the page and use row-level checkboxes plus a bulk Delete action (with confirmation) to remove historical jobs that are no longer needed.
• The Usage view opens a dedicated Credits Usage dashboard that gives users a complete breakdown of how credits are consumed across all agents in a project over a selected time period, making it easy to track spend, pinpoint high-consumption agents, and understand execution patterns at both a summary and transaction level.
• The Credits Status option provides an at-a-glance summary of the organization's credit balance directly from the Jobs tab, so users can quickly check where they stand without opening the full Usage dashboard.
• Access a context menu for each job with options to view details in Job Explorer, terminate an in-progress job, or delete it, ensuring users can intervene when executions need to be stopped or cleaned up.

a. Job Explorer – Deep Dive into Individual Job Runs

For each job, the Job Explorer provides a rich, three-tab experience that surfaces execution details, agent responses, and where applicable interactive tasks that require user decisions. This helps users understand what happened, why, and what to do next.

• Use the Run Details tab to see the job's status, core information (Job ID, and for event-based runs, the associated work item type and trigger event), execution model, skills, execution timing (start, end, duration), and logs.
• Review logs that show high-level activity for manual and scheduled runs, and deeper, step-by-step processing for event-based runs, including configuration resolution, code execution, and input/output details that support troubleshooting and validation.
• Use the View Agent Response tab to see the agent's own narrative output, including its plan, findings, and any limitations or errors, presented in a human-readable format that explains what the agent attempted to do.
• For assisted autonomous agents that require approvals, use the Tasks tab to work through pending questions: see a highlighted message explaining that the agent is waiting for input, review each task in a card layout with status, question, and answer field, and progress through tasks using Next/Proceed buttons until execution can continue.
• Users can get help from an integrated Chat Assistant within Job Explorer to ask questions about the agent's execution details, get help to interpret tasks and guide users by providing appropriate responses to the questions.

2Agents Workspace – Viewing and Managing Agents

Users can now open the Agents tab in Agents4DevOps to see all configured agents in a single, filterable grid, making it easy to understand what automation is available in a project or across the organization. The grid clearly distinguishes between system and custom agents with collapsable sections, shows how each agent is configured, and allows users to quickly inspect or act on any agent.

• View all agents in a grid with key details such as name, configuration including the agent mode and its type, project name, trigger event and enabled status.
• Use rich filtering to narrow agents by name, mode, type, projects, trigger event and enabled status, with multi-select dropdowns.
• Interact with a contextual menu per agent to view details, clone an existing configuration, edit agent, execute assisted autonomous agents, or permanently delete agents with confirmation safeguards.

a. Creating Execution Agents from Instructions, Code, or Templates

Execution agents can now be created through multiple flexible paths, allowing teams to either describe the behavior they want, upload existing code or configuration, or reuse prebuilt templates. This reduces setup time and ensures that automation can be tailored to each project's needs.

• Use the "Write Instructions" method: specify a code generation model, select one or more projects, define work-item-based triggers and work item types, and provide natural-language instructions. The system then generates agent code, name, description, and tags that can be reviewed and refined before creation.
• In the "Upload config or code" mode, upload existing source code or configuration files (such as MatCal or suspect link configurations) and let the system drive trigger events, work item types, and code details automatically, with the ability to adjust them prior to creating the agent.
• In the "Library" mode, browse a curated library of execution agent templates, open any template to review its description and tags, and then "Use" it to create a new agent that can be customized before being added to the grid.
• Access version management for skills, allowing users to see previous versions, preview older definitions, and restore them when needed, with clear version labels, timestamps, and creator or updater information.
• Users can get help from an integrated Chat Assistant available on each configuration tab to ask questions about the current agent, and to update code according to requirements.

b. Creating Assisted Autonomous Agents

The product now supports creating assisted autonomous agents through a guided configuration experience that balances automation with human oversight. Users can define how and when these agents run, whether they require approvals, and how they should orchestrate work.

• Start from the Agents tab using the "New Agent" button, choose "Assisted Autonomous Agent" as the mode.
• Choose from multiple creation modes: 'Create with AI' (describe the goal and let the system draft detailed agent instructions), 'Create manually' (define name, description, and instructions), or starting from 'Library' templates that can be tailored to your needs.
• Configure the scope of each agent by selecting one or more projects (including an "ALL" option), choosing an execution model from available LLMs, and optionally require approval so that the agent pauses and creates tasks when user input or clarification is needed.
• Select from Manual, Schedule, or Event-based agent types, with tailored options such as once, daily, weekly, monthly, or yearly schedules, or a rich set of DevOps event triggers spanning builds, code, pipelines, and work items.
• For event-based agents, define exactly which events and work item types or repositories should trigger execution, with support for multi-select options so agents can react to the right changes across code, pipelines, and work items.
• Use structured configuration tabs to manage behavior:
  • Trigger tab for all timing and event settings (including Required Approval, projects, and execution model)
  • Details tab for name, description, and instructions
  • Tools tab to search, filter, and add skills or use AI-driven skill suggestions, as well as manage tool calls by enable/disable them from provided toggles for the specified agent
  • Notifications tab to turn notifications on/off, configure subject/body, include logs and agent responses, and select recipients
• Access version management for agents, allowing users to see previous versions, preview older definitions, and restore them when needed, with clear version labels, timestamps, and creator or updater information.
• Users can get help from an integrated Chat Assistant available on each configuration tab to ask questions about the current agent, and receive guidance when choosing tools or adjusting settings.

3Skills Tab – Viewing and Managing Skills

A dedicated Skills tab now provides a centralized view of all skills defined in the organization, helping teams understand and govern reusable capabilities that agents can use. The grid clearly distinguishes between system and custom skills and shows where each skill is used.

• View all skills in a grid distributed in 2 collatable categories — System and Custom — with columns for name, linked agents, and enabled status, so you can quickly see which skills are active and where they are applied.
• Filter skills by name, category, and linked agents using a rich filter bar with multi-select dropdowns and quick "clear" actions, making it easy to find the right skills in large environments.
• Use the "New Skill" button to create new custom skills and manage existing ones through a context menu that supports cloning, editing, and deleting, with system skills restricted to Customize and Clone to protect core functionality.
• Enable or disable skills via a toggle; when a skill is disabled, it remains visible but is not used during agent execution, even if it is configured on an agent, ensuring safe deactivation without losing configuration history.
• Access version management for skills, allowing users to see previous versions, preview older definitions, and restore them when needed, with clear version labels, timestamps, and creator or updater information.
• Users can get help from an integrated Chat Assistant available on each tab to ask questions about the skill, perform updates as well as get information about the artifacts attached as knowledge for the skill.

4Library – Reusing Pre-Created Agents and Skills

The Library tab introduces a curated catalogue of pre-created assisted autonomous agents, execution agents, and skills that can be discovered, filtered, and reused. This helps teams start from proven patterns instead of designing every agent or skill from scratch.

• Browse three library sub-tabs — Assisted Autonomous Agents, Execution Agents, and Skills — each with its own search and collapsable categories distribution.
• Search by keyword across names and descriptions; category labels with counts help you understand how many templates are available in each area.
• Open any library item in the assisted autonomous agent tab to see a detailed popup with fields such as name, library category, description, and tools (for assisted autonomous agents), along with a prominent "Use" action.
• Click "Use" to open the selected agent or skill in an editable form, with a clear notice that it is based on a predefined process template and should be validated for the current project, then create it so it appears in the main Agents or Skills grid.

5Configuration Agents4DevOps

a. Agent Connection Settings

Administrators can now configure how Agents4DevOps connects to backend services, including personal access tokens and base URLs. This ensures that agents operate securely and consistently across environments.

• Use the "Configure Agents4DevOps" section under the General tab in the Admin Panel to define agent connection settings, including one or more personal access tokens (PATs) and the base URL for the Copilot4DevOps service.
• Enter a primary PAT, validate it with a dedicated button, and see the associated PAT user to confirm that the correct identity is being used.
• Add up to ten PATs, including the "Additional PAT" field, so that if the primary PAT becomes invalid, the system can fall back to additional tokens without disrupting agent operations.
• Configure or override the default base URL (prepopulated with the hosted service address) when using a self-hosted or alternative deployment of Copilot4DevOps, ensuring agents connect to the correct backend.

b. Outreach Email Configuration

Administrators can now configure how Agents4DevOps sends outbound email notifications based on agent instructions, using either an SMTP server or the SendGrid Web API. This provides centralized, organization-wide control over agent email and ensures that outreach is sent from an approved, verified sender identity.

• Use the "Outreach Email Configuration" section under the General tab in the Admin Panel to enable or disable agent email sending for the entire organization.
• Select the preferred Email Provider — SMTP server or SendGrid (Web API) — so that email delivery aligns with the organization's existing mail infrastructure. Configuration fields update automatically to match the selected provider.
• When SMTP server is selected, define the full SMTP configuration including Server Address, Port, Use SSL/TLS, From Address, From Display Name, SMTP Username, and SMTP Password.
• When SendGrid (Web API) is selected, provide the SendGrid API Key along with the From Email (verified sender) and From Display Name. The API Key field can be left unchanged to retain the current key when updating other settings.
• Validate the configuration before going live by using "Send Test Email" to confirm that credentials and connectivity are working as expected, regardless of the selected provider.
• Create and maintain reusable email content through "Manage Templates" so that agent notifications remain consistent and on-brand across the organization.
• Download a complete Email Audit Log as a CSV file for traceability, compliance, and troubleshooting of all agent-sent email.

c. Daily Credit Consumption Alerts

Administrators can now monitor and govern daily credit usage by setting warning thresholds, designating notification recipients, and enabling automated safeguards. This helps organizations stay ahead of unexpected consumption and protect the monthly quota from being exhausted by a small number of high-usage agents.

• Use the "Daily Credit Consumption Alerts" section under the General tab in the Admin Panel to send a warning email whenever the organization's daily credit utilization exceeds a selected percentage.
• Set a Warning Threshold as a percentage of the monthly quota, so that alerts are triggered at a level appropriate to the organization's usage patterns.
• Specify one or more Notification Email Addresses to ensure the right stakeholders are informed as consumption approaches critical levels.
• Enable Agents4DevOps to automatically disable a defined number of the highest-consuming agents once today's credit utilization reaches a selected percentage of the monthly quota, providing an automated safeguard against runaway usage.
• Configure both the count of top consuming agents to disable and the utilization percentage that triggers the action, giving administrators fine-grained control over the automated response.

d. Agents4DevOps Recommended Models

Administrators can now configure which AI models power each Agents4DevOps module by defining a prioritized list of models per module. This gives organizations centralized control over model selection and allows lower-ranked models to serve as fallbacks, ensuring consistent behavior across all modules.

• Use the "Agents4DevOps Recommended Models" section under the General tab in the Admin Panel to configure AI models for each Agents4DevOps module from a single, structured table of Module Name and Ranked Models.
• Assign one or more ranked models to each supported module, including AgentCodeGenerationModel, AgentExecutionModel, AgentInstructionGenerationModel, AgentRegenerationModel, ChatAssistant, SkillGenerationModel, and SkillRegenerationModel.
• Order the models within each module to establish priority, so that the system uses the highest-ranked model first and falls back to subsequent models as needed without disrupting agent operations.
• Add or remove models per module directly from the table, with each selected model displayed as a removable entry for quick adjustment.

e. Configure Chat Assistant Model

Administrators can now select which AI model powers the Chat Assistant across the entire platform. This ensures a single, consistent conversational experience for all users, while remaining aligned with the models approved in the organization's recommended models configuration.

• Use the "Configure Chat Assistant Model" section under the General tab in the Admin Panel to configure the AI model used for the Chat Assistant across the platform.
• Select the desired Chat Assistant model from a dropdown of approved options, applying the choice consistently for all users.
• Choose only from models that are configured in the Agents4DevOps Recommended Models section, ensuring that Chat Assistant selection stays governed by the organization's central model configuration.

f. Global Tool Call

A new Global Tool Call section in the Admin Panel gives administrators centralized control over which tool calls are available to agents. This allows organizations to govern behavior, reduce risk, and enable only approved capabilities.

• Manage all tool calls from a structured, collapsible list under the General tab, grouped by tool and section, with each individual tool call displayed as "Name & Description" in a clear, readable format.
• Toggle entire tools section on or off, use section-level "Enable All / Disable All" actions, or control individual tool call via per-row toggles, with all settings persisted using the existing configuration pattern.
• Start from a secure, permissive default where all skills are enabled on first load, and then selectively disable capabilities that should not be used by agents, with runtime enforcement that blocks disabled tool calls and surfaces clear messages indicating that a tool call is disabled in Global Tool Call settings.

g. Global and Project-Level Permission Management

The release introduces a comprehensive permission model that lets administrators control what actions users can perform with agents and jobs, both globally and per project. This ensures that sensitive operations such as executing agents, deleting jobs, or managing tasks are only available to authorized roles.

• At the organization level, use the "Global Permissions" subsection in the Admin Panel to configure twelve key actions including: Manage Agent, Delete Agent, Execute Agent, Enable Agent, Customize Agent, Manage Skill, Delete Skill, Enable Skill, Customize Skill, Job Explorer, Delete Job, and Open Task — each with a simple Allow/Deny option.
• At the project level, open Project Settings → Agents4DevOps to see all teams and core Azure DevOps groups (Project Administrators, Build Administrators, Project Valid Users, Readers, Contributors) and configure the same twelve actions per team or group.
• For each action at project level, choose Allow, Deny, or Not set, with clear indication when a value is inherited from global settings (e.g., "Allow (inherited)" or "Deny (inherited)"), and the ability to revert to inheritance by selecting Not set.
• When an action is denied, the corresponding UI elements are disabled consistently.