Azent

Azent is an AI coding agent that lives directly inside Azure DevOps. Instead of switching between tools, you trigger it from the work items and pull requests you already use — by @mentioning the bot in a comment or clicking an action button — and it does the work for you: writing code, opening branches and PRs, reviewing changes, and more, all without leaving the browser.

What it does

• Implement work items — describe a feature or change in a work item, trigger Implement with Azent, and the agent creates a branch, writes the code, and opens a pull request linked back to the work item.
• Fix bugs — point it at a bug ticket and let it diagnose and patch the issue.
• Address PR feedback — Fix comments with Azent has the agent resolve review comments and push updates to the same branch.
• Review pull requests — Review with Azent adds inline comments and a summary review on a PR.
• Manage work & docs — break down work items, answer questions in tickets, and edit wiki content.
• Verify its own work — the agent can run builds and tests through Azure Pipelines before handing the result back.
• Anything else - Azent does whatever user asks him to do in the mention

Visibility & control

Azent is driven entirely through Azure DevOps comments (@mention) and optional UI buttons from its Azure DevOps extension. Every run is observable: a dashboard shows status (queued, running, completed, failed) with a streaming event log, the agent posts status comments in DevOps. Access is restricted to members of an allowed Azure DevOps group, and the agent runs under a dedicated bot identity.

Flexible models & deployment

Azent supports both Claude (Anthropic) and Codex (OpenAI) as the underlying coding agent. It can run in two ways:

• SaaS — use the hosted deployment; you only configure your workspace (ADO org, token, agent/model settings), with no infrastructure to manage.
• Self-hosted — deploy the entire stack into your own Azure subscription via the provided Bicep template, with the option to keep all data inside Azure by running the LLM in Azure AI Foundry.

Configuration is split between per-project settings and per-workspace settings (agent provider, model, connection type, skills, and workflow), managed from a web dashboard that also handles run history, members, invitations, and project integrations.

In short, Azent brings autonomous, AI-driven software development into the Azure DevOps workflow — turning tickets, comments, and PRs into finished, reviewed code.