Free Open-Source Interview Assistants: 3 Real Options
Compare three available open-source interview assistants by license, model and audio costs, setup work, and capture limits before choosing a DIY tool.
Three free open-source interview assistants worth evaluating are Cheating Daddy, AI Interview Helper, and InterviewCopilot. They offer different levels of provider choice, audio support, and setup work. Choose a tool for private mock interviews or explicitly AI-permitted assistance, then test the complete workflow before depending on it.
Disclosure: GeekBye publishes this guide and sells a managed desktop assistant. We checked repository availability, licenses, and documentation on November 6, 2026. This is a source review, not a hands-on performance or security audit. We do not claim that any option guarantees interview success, runs entirely locally in every configuration, or stays hidden from every capture system.
What “free” and “open source” actually tell you
Free software access means the app or source can be obtained without a software payment. It does not establish that the model behind it is free.
Open source requires a license granting the relevant rights to use, modify and redistribute the software. MIT and GPL-3.0 are listed by the Open Source Initiative. Their obligations differ: keep the license and notices, and read the applicable terms before distributing a modified version.
Source-available means code can be read, but its license may impose restrictions inconsistent with open-source criteria. A GitHub page without a license is not a substitute for permission to reuse the code. A proprietary binary with a free plan is another category entirely.
Check the app license separately from model-weight licenses, dependencies, and API service terms. Open-source application code does not grant unrestricted rights to everything it connects to.
Three available projects, compared
The repositories below were public and not archived at the check. Recent pushes show activity, not a maintenance guarantee or a completed security review. Re-check the live issue tracker and releases before installation.
| Project | License and checked revision | Good reason to evaluate | Main practical limitation |
|---|---|---|---|
| Cheating Daddy | GPL-3.0; 3cccc36 |
A small Gemini-based screen/audio assistant | Own API key, dependency setup, experimental Linux support |
| AI Interview Helper | MIT; d692c39 |
Multiple answer providers and screenshot input | Local answer models do not establish local transcription |
| InterviewCopilot | MIT; 4a80d84 |
Documented local Whisper/Ollama default and practice mode | Beta; macOS captures microphone only |
1. Cheating Daddy: a Gemini-based starting point
Cheating Daddy documents screen and audio capture with a configurable overlay and profiles. Its setup asks for a Gemini API key, dependency installation, and a local app launch. The README recommends current macOS and Windows versions and calls Linux support experimental.
This is a reasonable starting point if you want to understand or adapt a relatively direct screen/audio workflow. It is less suitable if you need someone else to manage model changes or troubleshoot your operating system's audio routing.
We checked revision 3cccc36, including its GPL-3.0 license. The repository reported a push on October 21, 2026. Its documented Gemini model is not proof of current API entitlement: confirm the endpoint works with your key and inspect the provider's current terms and charges.
2. AI Interview Helper: more provider choice, more settings to verify
AI Interview Helper documents screenshot analysis, voice input, and several cloud answer providers plus Ollama. Its README describes speech recognition through cloud services. Selecting Ollama for answers therefore does not establish an entirely offline audio-to-answer pipeline.
The project also describes additional system-audio routing for Windows and macOS. That matters if your microphone works but the assistant cannot hear the other side of a practice call. Treat those as separate input paths.
We checked revision d692c39 and its MIT license; the repository reported a push on October 28, 2026. Evaluate it when provider flexibility matters. Verify vision support for your selected model and the speech provider's billing before testing screenshots and a full-length mock session.
3. InterviewCopilot: a local-first practice option with beta limits
InterviewCopilot documents local Whisper transcription, Ollama answers, locally stored history, and a practice mode. Optional cloud providers change that data path and can create separate API charges.
Its README is explicit about a significant limit: the macOS beta transcribes the microphone only, not the other side of the call, and its Mac builds have not been tested on a real Mac. The project also documents unsigned installers, no automatic updates, and substantial model storage and memory requirements. Those limits make a rehearsal essential.
We checked revision 4a80d84 and its MIT license; its reported push was October 30, 2026. Consider it for local practice if your hardware and platform fit. An overlay is not, by itself, evidence of screen-share exclusion.
Why Pluely and Glass are not recommendations here
Older lists can outlive the software they describe. A recently indexed Pluely README states that v1 and later ship as proprietary binaries, while earlier GPL code remains licensed in history. The live repository could not be verified during this check. Do not call the current product open source based on its older description, or assume a historical version is maintained.
Glass by Pickle's former repository URL also did not resolve in this check. That does not establish why it is unavailable. It does mean we cannot recommend it here as a currently available, licensed project. A similarly named fork needs its own provenance, license, and maintenance checks.
For the broader product comparison, see Natively, Cluely and Pluely, while using current license files for installation decisions.
Budget the whole pipeline before calling it free
Write down five costs for your chosen configuration:
- Software: the application license and any paid hosted features.
- Speech: transcription minutes, streaming sessions, and any separate audio-routing tools.
- Answers: model input and output tokens, including repeated conversation context.
- Images: screenshot frequency, image-token charges, and retries.
- Operations: hardware, disk space, electricity, installation time, and updates.
A free API allowance can help with evaluation, but quota, supported models and eligibility may change. Check the provider console and put a spending limit in place before a long practice session. Avoid extrapolating the cost of one short text answer to an hour of streaming audio and screenshots.
For local configurations, verify both transcription and answer generation. Run a controlled practice session with network access disabled after model downloads. A successful text reply alone does not establish that the complete voice workflow works offline.
Check capture behavior and data handling separately
Use a private test call with a second device to inspect the recipient's view. Repeat with the exact operating system, meeting app, window or display selection, and recording method you expect to use. Update changes can invalidate an earlier result; our screen-sharing checklist provides a repeatable approach.
Then trace where each input goes: microphone audio, system audio, screenshots, transcripts, prompts, and logs. Locally stored keys do not mean requests stay local. Review saved practice material and remove sensitive documents you no longer need.
If maintaining that pipeline is more work than you want, compare managed options in our LockedIn AI alternatives guide or Interview Coder alternatives guide. GeekBye is one option, with a free Starter allowance of 30 Listen minutes per month. Its AI requests go through a backend; it is not an open-source or all-local substitute.
Frequently asked questions
Which project should I try first?
Choose Cheating Daddy for a focused Gemini setup, AI Interview Helper for provider choice, or InterviewCopilot for its documented local-first practice workflow. Platform limitations can outweigh the feature list, especially on macOS.
Can I reuse or sell a modified version?
Read the license attached to that exact revision and preserve its required notices. Distribution obligations vary by license. Check dependencies and model terms separately; a project's README label alone is insufficient.
Can I use these tools during an actual interview?
Use them only when the interviewer or assessment rules explicitly permit the assistance and relevant capture. Private practice is the simplest evaluation setting. Software availability and capture behavior do not establish permission.
