FAQ
Frequently asked questions
Everything you need to know about Specter — organized by topic.
General
What is Specter?
Specter is an AI-powered interview co-pilot built as a desktop application by Tattva Mind. It silently monitors your screen during job interviews, detects questions the moment they appear, and streams spoken-ready answers in real time. It's cross-platform (Mac, Windows, Linux), supports multiple LLM providers, and includes a built-in code runner.
Who is Specter for?
Specter is designed for anyone preparing for or currently in job interviews — software engineers, data scientists, product managers, behavioral round candidates, and anyone facing technical assessments. It works equally well for fresh graduates, experienced professionals, and career changers.
What platforms does Specter support?
Specter supports macOS (Apple Silicon arm64 + Intel x64, DMG and ZIP), Windows (NSIS installer + Portable EXE), and Linux (AppImage + .deb for Debian/Ubuntu). All platforms receive the full feature set.
Does Specter require an account or login?
No. Specter requires no account to use on the Free tier. Simply download, configure your API key, and you're ready. An account is only required if you subscribe to the Pro plan for billing purposes.
Features
What is Race Mode?
Race Mode fires two LLM providers simultaneously for every detected question — for example, Gemini 2.0 Flash and GPT-4o at the same moment. Specter streams whichever provider returns its first token first. This eliminates the worst-case latency of any single provider and reduces 95th-percentile time-to-first-token by approximately 40% compared to single-provider configurations.
How does multi-monitor scanning work?
Specter uses OS-level screen capture APIs to capture all connected displays simultaneously. Each monitor is scanned independently in parallel. Questions appearing on any screen — including a secondary display showing the video call — are detected instantly, without any manual switching or focus management.
What is Hold Mode?
Hold Mode queues detected questions without immediately streaming answers. This is designed for in-person and hybrid interviews where you need to control when you review AI assistance. Queued questions are available via keyboard shortcut. Hold Mode can be toggled on and off mid-session.
How does the code runner work?
Specter's built-in code runner uses the Piston API for sandboxed code execution. It supports Python, JavaScript, TypeScript, Java, C++, Go, and Rust. The "From Answer" button extracts code from the AI's response into the runner automatically. You run the code with a keyboard shortcut and see output inline. No external IDE required.
What are the 6 interview role modes?
Software Engineer (technical depth, code quality), HR/Behavioral (STAR-method answers, leadership narratives), Coding/DSA (algorithm breakdowns with complexity analysis), System Design (scalable architecture, trade-offs, capacity estimation), Data Science (ML model selection, statistical reasoning), and Product Manager (RICE/ICE frameworks, user empathy, go-to-market).
What does Session Debrief do?
After each session, Session Debrief reviews all detected question-answer pairs and generates an AI-scored report (1-10 per question category) with specific improvement suggestions. It evaluates answer quality, response timing, and your use of refinement actions. Reports are stored locally and exportable.
What is the Practice Mode?
Practice Mode provides 20+ curated interview questions per role (SWE, Behavioral, DSA, System Design, Data Science, PM). Use it to simulate live interview conditions with AI assistance, identify your weak areas, and build muscle memory for using Specter effectively before your actual interview.
Can I customize the AI prompt?
Yes. Specter supports custom system prompts. In Settings → Prompts, you can modify the base prompt for any role mode or create a fully custom prompt. This is useful if you want the AI to focus on specific company values, use particular frameworks, or adapt to your communication style.
Privacy & Security
Is Specter really invisible to screen share software?
Yes. Specter uses OS-level window exclusion APIs — the same APIs used by video streaming apps to protect DRM content. When you share your screen in Zoom, Google Meet, Teams, Webex, or any other software, Specter's window is excluded from the capture. The interviewer sees a blank area where the window sits. This is not a hack — it's a standard OS capability.
Does Specter store or transmit my interview data?
No. All processing happens on your device. Questions detected, AI answers, session history, and API keys are stored in local files on your machine. Tattva Mind has no server-side access to this data. The only external communication is direct API calls from your device to your chosen LLM provider (Gemini, OpenAI, or Anthropic).
Are my API keys safe?
Your API keys are stored locally in Specter's settings file on your device. They are never transmitted to Tattva Mind's servers. Specter uses them only to make direct API calls from your machine to the LLM provider. We recommend storing your keys as environment variables or in a secrets manager for maximum security.
Is Specter GDPR compliant?
Yes. Because Specter processes all data locally on your device and does not send interview content to Tattva Mind servers, there is no personal data processing that requires GDPR compliance on our part. The only data subject to GDPR considerations is any data you transmit to LLM providers (Gemini/OpenAI/Anthropic) — each of which has their own GDPR compliance documentation.
Does Specter have any telemetry?
Specter has no telemetry by default. We do not collect usage analytics, error reports, or behavioral data unless you explicitly opt in. The opt-in crash reporting feature (if enabled) sends only anonymized crash logs — no interview content, no API keys, no personal data.
Billing & Plans
What is included in the Free tier?
The Free tier includes all Specter features — multi-monitor scan, Race Mode, content protection, multi-panel workspaces, code runner, audio capture, all 6 role modes, answer refinement, 300-entry history, practice mode, and session debrief. The only requirement is bringing your own API key (BYOK). There are no feature restrictions on the Free tier.
What does the Pro plan include?
$19/month (or $15/month annually). Pro includes everything in Free plus: $20 in bundled API credits per month (no need to configure your own key for typical usage), priority support with a 24-hour response SLA, early access to new features, advanced session analytics, and bulk history export in JSON and CSV.
How much does API usage cost on the Free tier?
With your own Gemini API key, the standard free tier offers generous monthly limits. At paid Gemini pricing, a typical 45-minute interview session costs approximately $0.05-$0.15. GPT-4o and Claude are more expensive per token. Full interview prep across many sessions typically costs $2-8/month in API fees on the Free tier.
Can I cancel the Pro plan anytime?
Yes. Cancel anytime from Settings → Subscription. Your Pro features remain active until the end of the current billing period. You'll revert to the Free tier automatically — all your data remains intact.
Technical
What are the minimum system requirements?
macOS 12+ (Monterey), Windows 10 64-bit+, or Ubuntu 20.04+ / Debian 11+. Minimum 4 GB RAM (8 GB recommended). 500 MB disk space. Internet connection required for LLM API calls. No GPU or local model required.
Why does Specter ask for screen recording permission on macOS?
Screen recording permission is required for multi-monitor question detection. Specter uses the macOS CGWindowListCreateImage API to capture screen content. This permission is requested once during first launch. Specter only captures your screen during active sessions — never in the background.
Does Specter work with virtual interview platforms?
Yes. Specter detects questions via screen capture and audio capture — it works regardless of which platform the interviewer uses. Compatible platforms include: HackerRank, CoderPad, LeetCode, Codility, CodeSignal, Karat, interviewing.io, Zoom, Google Meet, Teams, and all browser-based interview tools.
What LLM providers does Specter support?
Specter currently supports Gemini 2.0 Flash (Google), GPT-4o (OpenAI), and Claude 3.5 Sonnet (Anthropic). You can configure a primary provider, a secondary provider for Race Mode, and a fallback provider. Additional providers (Mistral, Llama via Groq) are planned for a future release.
Still have questions? Email us and we typically respond within 24 hours.