Illustrative workflow · reviewed 22 Oct 2026

Use a coding assistant on an existing repo

One small change to an existing codebase, reviewed and tested by a developer before it is merged.

You provide: A repository you may share with the tool, a passing test suite, and one small, well-defined task.
You get: A change that passes your tests and your review. Nothing is merged on the tool's word.
Tools: One per step: 3 tools, chosen below from the options we researched.
Integrations: None documented between these steps. All 2 handoffs are manual.
Review checkpoints: 3 steps need a person to check the output, marked "Check by hand".
Status: Illustrative. Built from vendor documentation, not from a test. Methodology

Step 1 Choose a safe repository and task
Pick a repository whose code you are allowed to send to a third party, with tests that pass today, and a task small enough to review in one sitting, such as adding a test for a documented bug.
Check by hand. Confirm your employer's or client's rules on sharing code, and read the vendor's data terms for the plan you will use. We have not verified any vendor's data handling.
Manual handoff: Open the repository in the assistant's editor or plugin. No integration between these tools is documented in the sources we read, so moving the file or text between them is manual.
Step 2 Ask the assistant for the change
Describe the task and the files involved, and ask it to explain what it plans to change before it edits.
- GitHub Copilot · shown first, not ranked
  Inline completions and chat in VS Code, JetBrains and GitHub, with agent mode in several editors. The free plan lists 2,000 completions and 50 chat requests a month.
  Free access, price and licence terms
  - Free access: Free plan. 2,000 inline completions and 50 chat requests a month; limited agent usage
  - Paid entry: From $10/mo, billing frequency not stated
  - Commercial terms: Not verified
  GitHub Copilot overview
- Amazon Q Developer
  Agentic coding in IDE plugins and the CLI. The free tier lists 50 agentic requests a month and no IP indemnity.
  Free access, price and licence terms
  - Free access: Free plan. 50 agentic requests a month; 1,000 lines of code a month for Java upgrades; no admin dashboard; no IP indemnity
  - Paid entry: From $19/user/mo (currency not named), billed monthly
  - Commercial terms: The pricing page lists IP indemnity as included on Pro and not included on the Free Tier.
  Amazon Q Developer overview
- Cursor
  An AI coding editor with agent and chat features. The free plan lists limited agent and chat requests, without exact counts.
  Free access, price and licence terms
  - Free access: Free plan. Limited agent and Chat requests
  - Paid entry: From $8/mo (currency not named), billed monthly
  - Commercial terms: Not verified
  Cursor overview
Check by hand. Read the whole diff. Check that it stays within the task and touches no unrelated file.
Manual handoff: The change stays in your editor and repository for review. Nothing is transferred between tools.
Step 3 Run your tests and review
Run the project's own tests, then review the change as you would a colleague's pull request.
Check by hand. Do not merge if tests fail, the diff is larger than the task or you cannot explain a change. Record how many requests you used and how long the review took.

Your selection

GitHub Copilot, Amazon Q Developer, Cursor

Compare these tools View saved tools

Comparison shows at most three tools. Tools doing different jobs are warned about on the comparison page.

Before you rely on this route

This workflow is illustrative. It is built from vendor documentation, not from a test.
Free tiers meter requests. Heavier use may use paid credits or hit a limit; check each plan page.
Code ownership and indemnity terms differ by plan. Amazon Q Developer's free tier lists no IP indemnity; Copilot's page lists none on individual plans. This is not legal advice.
The comparison page suggests a fair way to compare two assistants on the same task.

Guides for this workflow

Coding assistant for an existing repo: What to compare before you point an AI coding tool at a real codebase: usage metering, data and IP terms, editor support and agent behaviour.
Test an AI tool before buying annually: A two-week pilot method for deciding whether to pay for an AI tool, with trial and refund terms several vendors state and a scorecard you can reuse.

All workflows · Search by task