TL;DR
KeelTest is an alpha VS Code extension that uses AI plus static analysis to generate and execute pytest suites inside a sandbox, flagging real bugs in source code. The tool offers a free plan with monthly credits, automated test fixing, and built-in dependency mocking.
What happened
KeelTest launched as a VS Code extension in alpha that generates pytest test suites from Python source files and runs them in a sandbox before delivering the tests to a developer's editor. The system combines static analysis (building an AST and examining control flow and boundaries) with an AI-driven, multi-step pipeline: it maps dependencies, mocks external services, generates tests, executes them, and attempts automated fixes when generated tests fail. When failures reflect issues in the original code, KeelTest flags those as source bugs and offers suggested fixes. The extension is distributed via the VS Code Marketplace and advertises a free tier that provides seven credits per month. KeelTest reports high pass-rate statistics across its internal benchmarks and claims to prioritize linted, linter-compliant output that fits existing project standards.
Why it matters
- Automating test generation and execution can speed coverage of edge cases that manual tests might miss.
- Flagging real source bugs during test generation turns failing tests into diagnostic signals rather than just generation errors.
- Built-in mocking and dependency mapping reduce the manual effort required to isolate units under test.
- Sandboxed execution and automatic formatting aim to deliver ready-to-review tests that comply with project linters.
Key facts
- KeelTest is available as a VS Code extension (alpha) and installs from the VS Code Marketplace.
- The extension currently targets Python and pytest (the site indicates Python/pytest support).
- Free plan: 7 credits per month; Starter and Pro paid tiers listed with higher monthly credits and extras.
- Credits map to file complexity: 1 credit covers a small file (≤15 functions); larger files use more credits.
- KeelTest runs generated tests in a sandbox and reports pass rates; marketing materials cite a 90% average pass rate and a 91% average pass rate in different sections.
- Site metrics shown include 847 test suites generated and 127 source bugs discovered (stats updated weekly).
- Features listed: deep static analysis, smart dependency mapping, self-healing test generation, linter-compliant output, and automatic mocking.
- KeelTest says it uses a KeelTest agent powered by OpenAI's models to produce tests and mocks.
- The product includes a triage rule: if a pass rate is below 70% due to generation issues, that generation does not consume a credit (per the FAQ).
- KeelTest provides a refund policy: full refund within 14 days if not satisfied (site claim).
What to watch next
- Whether KeelTest expands language support beyond Python/pytest (not confirmed in the source).
- How the product handles private code, secrets, and data during sandboxed execution and if additional enterprise controls are added (not confirmed in the source).
- How pass-rate claims and independent benchmark results hold up under third-party audits or larger-scale user testing (not confirmed in the source).
Quick glossary
- pytest: A popular Python testing framework used to write and run unit tests and test suites.
- Abstract Syntax Tree (AST): A structural representation of source code that breaks code down into nodes representing syntax elements, useful for static analysis.
- Static analysis: Examining source code without executing it to infer control flow, variable usage, and potential edge cases.
- Mocking: Replacing real external dependencies (like databases or APIs) with controlled fakes during tests to isolate units of code.
- Linting: Automatically checking and formatting code to follow style rules and best practices.
Reader FAQ
What happens if generated tests don't pass?
KeelTest attempts automated fixes for generation failures; if failures reflect real bugs in your code, it flags them and suggests fixes.
How does the credit system work?
Credits are tiered: small files (1–15 functions) cost 1 credit, medium files cost 2 credits, and credits reset monthly per plan.
Which AI model does KeelTest use?
KeelTest uses a specialized agent powered by OpenAI's models, according to the site.
Which languages are supported?
Currently supports Python with pytest (as stated on the source page).
Can I get a refund or change plans?
The site states a full refund within 14 days if unsatisfied and allows upgrades/downgrades at any time.
KeelCode Features Pricing FAQ About Blog Sign In Get Started VS Code Extension · Alpha AI Tests That Find Bugs Before Production Generate pytest suites that actually run – and…
Sources
Related posts
- Big tech feels stagnant — 2026 could be the year independents rise
- CES 2026’s Strangest Gadgets: From Superthin TVs to Tri-Folding Phones
- Clicks Communicator aims to fix what many modern smartphones get wrong