Skip to content

Agentic Unit

Agentic Unit helps VWFS teams delegate software engineering work to an AI agent. An agent can inspect task context, work with approved repositories, make and verify changes, and publish a durable result for human review.

Agentic Unit is currently in beta. Users remain responsible for reviewing all generated changes before they are merged or deployed.

Ways To Use Agentic Unit

  • Web chat — select a team and agent, start or resume a conversation, and follow the agent's messages and tool activity.
  • Agentic Unit CLI (aca) — work from a terminal, select one or more repositories, attach a Jira ticket or local checkout, upload small files, resume jobs, and apply a published result locally. See Chat And CLI.
  • Jira Automation — dispatch a queued job using the environment-specific setup: v2 dispatch in Production and the legacy Jira webhook in Development. See Jira Automation.
  • GitHub pull request comments — request a review with the mention configured for the environment. See GitHub Pull Request Reviews.
  • Azure DevOps automation — dispatch work-item jobs with a scoped trigger token. See Azure DevOps Automation.
  • Schedules — run a team-owned job repeatedly or on demand. See Scheduled Jobs.

Connected Systems

Depending on the selected agent and its attached skills, Agentic Unit can work with:

  • VWFS Jira for issue context, comments, field changes, and workflow transitions.
  • VWFS GitHub for repository access, branches, commits, pushed work results, and pull request reviews.
  • Azure DevOps for work items, Azure Repos branches, and status callbacks.
  • Confluence and other tools when the selected agent has the corresponding skill and the integration is configured.

Access is always limited by the selected team, its configured integrations, and the selected agent's skills. A capability listed here is not automatically available to every agent.

What To Ask The Agent

Good tasks are specific, bounded, and verifiable. Include:

  • the ticket or business context;
  • the repository or component;
  • expected behavior or acceptance criteria;
  • tests or other checks to run;
  • constraints the agent must respect.

Suitable examples include fixing a bug, implementing a small feature, updating tests or documentation, reviewing a pull request, or investigating a failing build. Split large or ambiguous requests into smaller jobs.

Start Here

  1. Complete Onboarding.
  2. Read Core Concepts.
  3. Choose Chat And CLI or the relevant automation guide.
  4. Use Troubleshooting if a run does not start or cannot access its context.
  5. Check the FAQ for common workflow and safety questions.