CATEGORY · 9 TOOLS

AI coding tools help developers write, explain, test and refactor software. The category spans inline autocomplete inside an editor, chat panels that answer questions about a codebase, AI-first editors such as Cursor and Devin Desktop (formerly Windsurf), and terminal or cloud agents that take a task, change many files and open a pull request for review.

WHERE THESE 9 TOOLS KEEP DATA

  • AU hosting option2
  • Can run onshore1
  • No AU option published6

001 / OUR PICKS

The biggest shift since 2025 has been from suggestions to delegation. Command-line agents such as Claude Code and OpenAI Codex, GitHub's coding agent and background agents in Cursor can now run tests and iterate on their own. Consolidation also reshaped the market: Cognition, maker of Devin, acquired Windsurf in July 2025, and most tools now let you choose between models from several providers.

Australian users range from fintech and SaaS engineering teams to agencies, government digital units and solo developers maintaining client sites. For many teams the question has moved from whether to allow AI coding help to how to govern it: which repositories it may read, who reviews its pull requests and how secrets stay out of prompts.

ALSO IN THIS CATEGORY · 3

002 / HOW TO CHOOSE

  1. 01

    Code privacy and training settings

    Confirm in the contract that your code and prompts are not used for model training, how long they are retained and whether retention can be set to zero. Business and enterprise tiers usually offer these terms; individual plans often do not, so keep client code off personal accounts.

  2. 02

    Agent or autocomplete

    Autocomplete suits teams that want small, low-risk help while typing. Agents that edit many files and run commands suit well-tested codebases with good continuous integration, because tests are what catch their mistakes. Many teams use both, with agents limited to branches and pull requests.

  3. 03

    Editor and platform fit

    GitHub Copilot and JetBrains AI fit existing editors with little change. Cursor and Devin Desktop replace the editor itself, which some developers love and others resist. Terminal agents work with any editor. Check support for your languages, monorepo size and source control host, not only the demo.

  4. 04

    Admin and security controls

    Look for single sign-on, seat management, audit logs, the ability to exclude sensitive files and directories, and policy settings for which models and agent permissions are allowed. Ask how the tool handles secrets in environment files and whether it can be blocked from running destructive shell commands.

  5. 05

    Self-hosting and on-premise options

    Teams with strict client or government requirements should look at tools that can run inside their own cloud account or offline, such as self-hosted coding assistants or local models served through Ollama or LM Studio. Expect lower capability than the leading hosted models in exchange for control.

003 / FOR AUSTRALIAN BUSINESSES

Data that can't leave Australia? →

Source code is not usually personal information, but test data, logs and database dumps often are, so the Privacy Act 1988 applies when they reach an AI tool. GitHub Enterprise Cloud can host repositories in an Australian region, yet Copilot's own data residency covers only the US and EU, and most coding agents process prompts in the US. Government suppliers should check the Protective Security Policy Framework and agency rules before using hosted assistants on official code. Contracts with clients may restrict sending their code offshore at all. Seats are billed in US dollars, and local support is limited to business hours in US or European time zones for most vendors.

004 / QUESTIONS

Anything else is quicker to answer on a call. Thirty minutes, free, with the people who would build it.

Copilot is the simpler choice for teams already on GitHub who want help inside their current editor with central billing and policy controls. Cursor suits teams willing to switch editors for deeper multi-file editing and agent features. Run both on the same backlog tickets for two weeks and compare review effort, not just speed.

Thirty minutes, free. Tell us the job and we'll tell you which tool we'd use, and how we'd set it up.

Book a free call→