Pythagora opens by saying it is deprecated, and its licence is Apache-2.0 in one place and ISC in another
Generate automated tests for your Node.js app via LLMs without developers having to write a single line of code.
At a glance
- What is it?
- Pythagora generated Jest unit tests for Node.js functions by parsing the call graph with AST and sending the result to a Pythagora server that called GPT-4. The interesting parts are the ones the front page does not smooth over: a deprecation notice in the first heading, a licence mismatch, test tooling shipped as runtime dependencies, and a privacy note that ends at OpenAI's policy.
- Who is it for?
- Pythagora is a historical package rather than one to adopt. The repository's own first heading says it is deprecated in favour of GPT Pilot, the last commit is dated 2026-06-16, and the project publishes no releases at all, so the version you can obtain is whatever the registry currently serves.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 152 days ago.
- What is it written in?
- Mainly JavaScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- Pythagora-io/pythagora1,820 ★ · -2
| Repository | 2026-09-10 | 2026-11-15 |
|---|---|---|
| Pythagora-io/pythagora | 1.8k | 1.8k |
-2 Stars in the last 66 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The first heading on the front page says the repository is deprecated
The top of the README is a single line: this repo is deprecated, and the work has moved to a different repository, GPT Pilot. Everything below that line, the quickstart, the options list, the configuration section, the accuracy notes and the FAQ, is documentation for a package that is no longer the active project. GitHub's own archive flag is off, and the branch was pushed to on 2026-06-16, so the deprecation is stated in prose rather than enforced by the platform. There are also no GitHub releases for the repository at all. The practical consequence is that this is the shape of a project that has been handed off: useful as a record of how the approach worked, and a poor basis for a new dependency.
The licence is Apache-2.0 in the metadata and ISC in the manifest
The two licence statements do not match. Repository metadata reports the project under Apache-2.0, and a LICENSE file sits at the root. The package manifest declares its own licence field as ISC. Nothing in either file is referenced from the README, so a reader has to notice the discrepancy rather than being told about it. This matters more than it might for a testing tool, because generated tests end up committed inside someone else's repository and travel with that repository's history. If you are auditing a project that used Pythagora, check which licence the code that was actually installed was published under, and treat the root LICENSE file and the registry record as two separate claims until you have read both.
The package describes integration tests, the front page documents unit tests
The manifest's description says the package generates integration tests by recording server activity without writing a single line of code. The README says the opposite, opening by stating that the details on the front page are for generating unit tests and pointing to a separate document under src/docs/integration for the integration path. Both paths exist, and the dependencies support the integration story: a network interception library is a direct requirement, which is what recording server activity needs. What the front page never does is document that path. The unit path is the documented one, and it is a different product from the one the registry description advertises.
Your function source is shipped to a Pythagora server before GPT-4 sees it
The mechanism is three steps. First the target function is located, then every function called from inside it is found by parsing the abstract syntax tree, and then the target plus its callers are sent to the Pythagora server, which generates the unit tests with GPT-4. The server side is not a black box: it is described as open source in a separate repository, and the prompts live in a prompts folder in that repository. That means the quality of what you get back is a function of someone else's prompt file, at a version you did not pin. The privacy note states that no code is stored on their servers, and then immediately says the code is being sent to GPT and therefore to OpenAI, linking OpenAI's policy. That is the whole disclosure.
Jest and the coverage tool are runtime dependencies, not dev dependencies
The dependency list is arranged in a way that tells you what the package thinks it is. jest and the coverage wrapper sit in dependencies rather than in devDependencies, so anyone who installs Pythagora into a project also installs a test runner and a coverage tool into that project's dependency tree. handlebars is there for templating the generated test files, axios for the server calls, and a terminal UI library provides the progress display during a run. One dependency is a separate package under the vendor's own scope, an AST-oriented code processing module at an early version. The keyword list, meanwhile, reaches well past what the tool produces: MERN, MEAN, mongoose, mongodb, redis and express appear there, while the output format is Jest only.
The tool runs on itself and its generated tests are committed
The repository carries a directory of generated tests at its root, and the manifest wires two scripts to it. The test script runs jest over that directory with coverage, and a script named generate-unit-tests invokes the CLI against the project's own source folder. So the codebase is self-hosted: its tests were produced by the same tool it ships, they are versioned, and running the tool over the tree is a supported workflow rather than an accident. The accuracy note gives the shape of that output. Running the generator over one utility file produced 145 tests, of which 17 failed, and the README breaks the failures into 6 that were wrongly written tests and 11 that caught actual bugs in the source. Those generated tests are readable under a mirrored path in the output directory. Two larger examples are offered separately as forked public repositories: a Lodash fork reporting 1604 generated tests, 11 bugs found described as one edge case and ten bugs, and a four hour run, and a node-fs-extra fork reporting 98 tests, two bugs and thirty minutes.
Only Jest is supported, and the publish path is a shell script
Output is limited to one runner. To execute what was generated you run jest against the output directory, or against a path inside it, and the README states that other frameworks are not supported and invites a mail to the project's address if you want them. There is an integration test document, but there is no second runner to pair with it. The publish path is where the platform assumption shows: the script that publishes is a bash file, and the install hook runs a Node script after every install. The package floor is declared as Node 14 or newer, which is older than the versions some of the direct dependencies expect.
Windows users are told to use Git Bash
The four documented invocations differ only by what you name:
npx pythagora --unit-tests --func <FUNCTION_NAME>
npx pythagora --unit-tests --func <FUNCTION_NAME> --path ./path/to/file.js
npx pythagora --unit-tests --path ./path/to/file.js
npx pythagora --unit-tests --path ./path/to/folder/That covers one function, one function inside one file, every function in one file, and every function in a folder. A fifth command exists for a different job: npx pythagora --expand-unit-tests --path <PATH_TO_YOUR_TEST_SUITE> runs over a single test file or a folder of them to raise coverage on code that already has tests. The target function also has to be exported from its file, and the README illustrates that with a CommonJS module assigning a single function to module.exports. One explicit platform note appears in the quickstart: on Windows, every command should be run with Git Bash rather than PowerShell or anything similar.
Editorial conclusion
Pythagora is a historical package rather than one to adopt. The repository's own first heading says it is deprecated in favour of GPT Pilot, the last commit is dated 2026-06-16, and the project publishes no releases at all, so the version you can obtain is whatever the registry currently serves. If you are auditing an existing installation rather than adopting one, three things matter. Your source is shipped to a Pythagora server and on to OpenAI, and the repository's privacy note stops at the link to OpenAI's policy rather than making a promise of its own. The licence is stated as Apache-2.0 in the repository metadata and ISC in the package manifest, so settle that before you redistribute anything. And only Jest output is supported.
Frequently asked questions
Is there a Pythagora VS Code extension?
Yes. The front page links a Visual Studio Code extension published under the name PythagoraTechnologies.pythagora-vscode-extension on the marketplace. The repository itself is command line only, with the binary exposed as pythagora and all options given as flags.
How does Pythagora compare with Cursor or another AI coding tool?
The repository does not compare itself with any editor. What it documents is a single-purpose CLI that finds a function and its callers through AST parsing, sends them to a Pythagora server, and has GPT-4 write Jest tests for them, plus a VS Code extension for the same workflow. It is also marked deprecated in favour of GPT Pilot.
What happens to my code when I run Pythagora?
The target function and every function it calls are sent to the Pythagora server, which passes them on to GPT-4 at OpenAI. The privacy note says no code is stored on their own servers and points to OpenAI's privacy policy for the rest. The Pythagora server itself is open sourced in a separate repository, with its prompts in a prompts folder there.
Which test frameworks can Pythagora generate tests for?
Jest only. The generated suite is run with npx jest against the pythagora_tests directory, and the front page says other frameworks are not supported and asks to be contacted by mail if you need one.
Is the Pythagora repository still maintained?
Its own front page says the repository is deprecated and that work has moved to GPT Pilot. GitHub's archive flag is off, the branch was last pushed to on 2026-06-16, and there are no GitHub releases, so nothing has been tagged for distribution through the repository itself.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/pythagora-io-pythagora)Star history chart for your README
[](https://hysenlabs.com/projects/pythagora-io-pythagora)Was this useful?