Sourcegraph Cody has been around long enough that most developers have at least seen it in a VS Code marketplace search. The free plan is the part that gets overlooked. It is not a trial that expires after 14 days. You can install it, sign in with a GitHub account, and use AI chat and code completions inside your editor without pulling out a card.
That said, free comes with trade-offs. You get a capable assistant for individual work, not the full enterprise platform Sourcegraph sells to large engineering teams. The trick is knowing which parts of Cody the free tier covers well and where it will push you toward paid alternatives.
What Sourcegraph Cody (Free) Actually Gives You
Cody is an AI coding assistant that lives in your IDE. The free plan targets solo developers, students, and anyone evaluating AI tools before committing to a subscription. Core features include:
- Chat: Ask questions about code in your open files or project. You can request explanations, refactors, regex, SQL, shell commands, and more.
- Autocomplete: Inline suggestions as you type, including multi-line completions for functions and boilerplate.
- Commands: One-click prompts for common tasks like ‘Explain code,’ ‘Generate unit tests,’ and ‘Find code smells.’
- Local context: Cody can pull context from files you have open, selected code, and your local repository, so answers are less generic than a fresh ChatGPT tab.
Plan limits change from time to time. Sourcegraph has adjusted autocomplete and chat allowances since launch, so check the pricing page and your account dashboard for current caps. In practice, the free tier is enough for a side project, a coding bootcamp assignment, or a workday where you lean on autocomplete in bursts rather than all day.
Installing Cody Free in Your Editor
Setup is quick. The extension is available for VS Code, JetBrains IDEs like IntelliJ and PyCharm, and Neovim. A typical install looks like this:
- Open your editor’s extension marketplace and search for ‘Cody AI’ or ‘Sourcegraph Cody.’
- Install the extension from Sourcegraph.
- Sign in with GitHub, Google, or email when prompted.
- Open a project folder. Cody uses your workspace for context, so an empty window gives weak answers.
- Click the Cody icon in the sidebar, or use the keyboard shortcut, and ask your first question.
One small habit pays off immediately: open the files you want Cody to consider. If you ask about an authentication bug while only a CSS file is open, you will get a vague answer. Open the middleware, the route handler, and the failing test, then ask again.
Where the Free Plan Earns Its Place
Explaining unfamiliar code
You inherit a 2,000-line Rails service from a contractor who left no documentation. Highlight a method called reconcile_ledger_entries and ask Cody to walk through it step by step. Because it can see the surrounding file, the explanation often names actual variables and control flow instead of paraphrasing the obvious.
Writing tests without the blank-page problem
Ask Cody to generate pytest cases for a function that parses dates. It will usually produce a handful of tests, including edge cases like leap years and invalid time zones. The first draft is rarely perfect, but editing five generated tests is faster than writing them from scratch. I have used this for small utility functions and saved 20 to 30 minutes per function.
Refactoring bounded changes
Cody free handles small, well-scoped refactors well. Converting a callback-based function to async/await, renaming variables across a 60-line file, or extracting a helper method are good fits. Large architectural changes across 40 files are not. The context window and free usage limits make those requests frustrating.
Debugging stack traces
Paste an error and the relevant code into chat. Ask, ‘What are the three most likely causes, and how would I verify each?’ This works better than ‘Fix this’ because it forces Cody to reason. You still have to test the suggestions, but it shortens the hunt through logs and documentation.
The Limits You Will Hit on the Free Tier
Free does not mean unlimited. Expect friction in a few areas:
- Usage caps: Autocomplete and chat have monthly or fair-use limits. If you write code eight hours a day with autocomplete on, you may hit them before the month ends.
- Model access: Free users usually get a subset of models. Premium models may be reserved for paid plans, and model options can change without much notice.
- Context size: Cody can read local files, but it is not the same as Sourcegraph’s enterprise code search across hundreds of repositories. Cross-repo questions will fall flat.
- Team features: Shared prompts, analytics, SSO, and admin controls are enterprise territory. The free plan is for one developer at one keyboard.
- Support: You get community resources and documentation, not a dedicated success manager.
How to Stretch Sourcegraph Cody (Free) Further
Small changes in how you prompt make a bigger difference than any settings toggle.
- Name the files. Instead of ‘Why is login broken?’, say ‘Look at auth/session.ts and routes/login.ts. The session cookie is undefined after redirect. What should I check?’
- Give constraints. ‘Refactor this to use early returns, keep the public API the same, and do not add dependencies.’ Cody follows constraints surprisingly well.
- Use commands for repetitive work. The ‘Generate tests’ and ‘Explain code’ commands save typing when you are reviewing unfamiliar files.
- Keep chat threads focused. A long thread about database migrations will confuse a later question about CSS. Start a new chat for a new topic.
- Turn autocomplete off for prose files. Nothing burns usage faster than accepting suggestions in Markdown or JSON configs.
Free Cody vs. the Alternatives
GitHub Copilot has a free tier now, and Cursor offers limited free usage. Cody’s edge is its chat quality and Sourcegraph’s code intelligence background, even if the free plan does not include the full code search product. If you mostly want autocomplete, Copilot’s free tier is strong. If you want to ask questions about a local codebase and get explanations with real context, Cody free is competitive. Continue.dev is another option if you are comfortable wiring up your own API keys, but that is not truly free once you pay for tokens.
Privacy Notes Before You Paste That API Key
Free AI coding tools send code to model providers. That is how they work. Do not paste secrets, production credentials, or customer data into any chat window, including Cody. Sourcegraph publishes privacy documentation, and you can usually opt out of code snippet training in settings. If your employer has strict rules, check with them first. A free tool is not worth a security incident.
A 30-Minute Test Drive That Tells You If Cody Free Is Worth It
Install Cody in your main editor. Open a real project, not a toy repo. Pick one file you understand well and one file you do not. Ask Cody to explain the confusing file, then ask it to generate a test for a small function in the familiar file. Next, introduce a deliberate bug, run the test, and paste the failure into chat. Finally, check your usage dashboard to see how much of the free allowance you consumed.
If those four tasks felt useful, keep Cody free in your toolbox. If the caps annoyed you or the answers missed your stack, you have lost half an hour and learned something about how you want to work with AI. That is a better deal than another 14-day trial that turns into a forgotten subscription.