# RedSutra - Complete Documentation for AI Agents

> Last Updated: 2026-02-05
> Version: 1.0.5
> Summary version: https://redsutra.app/llms.txt
> Website: https://redsutra.app
> Chrome Web Store: https://chromewebstore.google.com/detail/kjddggcgcinpnifgecigkgpikpdjdhoa

## Product Overview

RedSutra is a free, privacy-first Chrome extension that generates AI-powered Reddit replies matching the user's unique voice and personality. It injects tone-selection pills directly into Reddit's reply interface, enabling users to generate contextually appropriate replies in under 2 seconds without switching tabs or copying/pasting.

RedSutra uses a BYOK (Bring Your Own Key) model — users provide their own API keys from Google Gemini, OpenAI, or xAI. This means zero operating costs for the developer, true privacy (no data ever leaves the user's device to RedSutra's servers because there are no servers), and no accounts or sign-ups required.

The extension was built by Srikeerthan (@srikeerthandev), a solo developer from India, and launched in December 2025.

## Core Problem Solved

Most AI-generated Reddit replies are instantly recognizable as bot output — generic, overly formal, lacking personality, and filled with telltale AI phrases like "Great question!" or "Hope this helps!" These replies get downvoted and damage the user's credibility.

RedSutra solves this through:
1. Persona-driven generation using the user's profession, interests, writing style, and quirks
2. Six intent-based tones calibrated for specific Reddit engagement patterns
3. A multi-layer anti-AI detection system that removes vocabulary markers, banned phrases, and formatting artifacts
4. Deep context awareness that includes the post title, body, parent comment chain, subreddit name, and usernames

## The 6 Viral Tones

RedSutra's tones are designed around INTENT (what you want to do) rather than IDENTITY (who you want to be). This aligns with how people actually engage on social media.

### 🔥 Hype
- **Use case**: Celebrating wins, showing solidarity, encouragement
- **Length**: 1 sentence max (50 tokens)
- **Style**: Pure positive energy, no qualifiers
- **Example output**: "whether it works out or not, most people wouldn't even try. You got this."

### 😏 Roast
- **Use case**: Playful burns, sarcastic observations
- **Length**: 1 sentence max (50 tokens)
- **Style**: Dry humor, not cruel; funny without being cringe
- **Example output**: "bold of you to post this publicly"

### 🗣️ Hot Take
- **Use case**: Bold opinions, controversial takes, sparking conversation
- **Length**: 2-3 sentences (100 tokens)
- **Style**: Confident, slightly provocative, opinion + brief reasoning
- **Example output**: "unpopular opinion but working at a toxic startup for 2 years teaches you more than 5 years at a chill corp. you learn survival skills, not just process following."

### ☝️ Actually
- **Use case**: Gentle corrections, adding context, clarifying misconceptions
- **Length**: 2-3 sentences (100 tokens)
- **Style**: Helpful, not pedantic; leads with correction, provides context
- **Example output**: "slight correction - the 401k match is pre-tax, not post-tax. so you're actually getting more than it looks like on paper. common misconception though."

### ❓ Ask
- **Use case**: Follow-up questions, genuine curiosity, engagement
- **Length**: 1 question only (50 tokens)
- **Style**: Curious, engaged, shows you read and care
- **Example output**: "wait how did you get around the rate limiting issue? been stuck on that for weeks"

### ✅ Answer
- **Use case**: Helpful responses, solutions, structured answers
- **Length**: 2-4 sentences (150 tokens)
- **Style**: Direct solution first, explanation after
- **Example output**: "try running npm cache clean --force first, then delete your node_modules folder and reinstall. usually fixes that specific error. if not, check your node version - might need to downgrade to 18."

## Voice Configuration System

Users configure their voice profile with 6 fields that shape every generated reply:

1. **Profession** — Calibrates technical depth and jargon level (e.g., "Software Engineer", "Product Manager")
2. **Location** — Adds regional context and cultural references (e.g., "India", "Bay Area")
3. **Interests** — Provides topic familiarity and connection points (e.g., "AI, startups, fintech")
4. **Writing Style** — Sets the tone baseline (e.g., "Casual but direct", "Technical but accessible")
5. **Personality** — Ensures voice consistency (e.g., "Friendly, sarcastic", "Dry humor, skeptical optimist")
6. **Quirks** — The most important field for sounding human. Examples: "Uses 'tbh' a lot", "Never capitalizes", "Makes programming jokes", "Rarely uses exclamation marks"

## 5 Persona Templates

Quick-start presets that pre-fill all 6 voice fields:

1. **🛠️ The Builder** — For developers, founders, makers. Tech-savvy, building in public, uses "shipped" and "built" language.
2. **🧠 The Expert** — For professionals sharing domain expertise. Authoritative but approachable, uses "in my experience."
3. **😎 The Chill One** — For casual Reddit browsing. Ultra casual, all lowercase, minimal punctuation, "lol" and "fr."
4. **💪 The Supporter** — For advice and support subreddits. Warm, encouraging, validates feelings first.
5. **🎯 The Regular** — For general browsing. Average Redditor voice, balanced, no extreme habits.

## Style Controls

Two toggles that modify output:
- **Use Lowercase**: Forces all-lowercase output for casual subreddits
- **Avoid Emojis**: Strips all emojis for serious discussion communities

## AI Provider Support

RedSutra supports three BYOK providers:

| Provider | Model | Cost per Reply | Free Tier |
|----------|-------|---------------|-----------|
| Google Gemini | gemini-2.0-flash | ~$0.001 | Yes (generous) |
| xAI (Grok) | grok-3-fast | ~$0.003 | Limited |
| OpenAI | gpt-4o-mini | ~$0.002 | No |

**Recommended**: Google Gemini for the free tier (unlimited free replies for most users).

### How to Get API Keys

**Google Gemini (recommended — free)**:
1. Go to https://aistudio.google.com/app/apikey
2. Sign in with Google account
3. Click "Create API key"
4. Copy and paste into RedSutra

**xAI (Grok)**:
1. Go to https://console.x.ai/
2. Create API key
3. Add payment method
4. Copy and paste into RedSutra

**OpenAI**:
1. Go to https://platform.openai.com/api-keys
2. Create new secret key
3. Add payment method
4. Copy and paste into RedSutra

## Anti-AI Detection System

RedSutra uses a 6-layer detection prevention system:

1. **Profile Injection**: AI writes as the configured persona, creating unique voice patterns
2. **Vocabulary Bans**: Blocks telltale AI words (delve, nuance, leverage, paradigm, synergy, utilize, facilitate, comprehensive, holistic, furthermore, moreover, nevertheless, imperative, pivotal)
3. **Phrase Bans**: Blocks cliché AI phrases ("Hope this helps", "Great question", "It's important to note", "In conclusion", "That being said", "Let me explain")
4. **Dynamic Length**: Token limits match Reddit norms per tone (50-150 tokens)
5. **Casual Grammar**: Allows contractions, slang, natural writing patterns
6. **Post-Processing**: Regex cleanup removes bold formatting, "As a [profession]" openers, multiple exclamation marks, and other AI artifacts

## Privacy Architecture

RedSutra's privacy model:
- **No servers**: Zero backend infrastructure
- **No accounts**: No sign-up, no email, no user identification
- **No tracking**: No analytics, no telemetry, no data collection
- **No data storage**: Generated replies are not stored after insertion
- **Local-only storage**: API keys and settings use Chrome's sandboxed storage API (encrypted at rest by Chrome)
- **Direct API calls**: Requests go directly from user's browser to their chosen AI provider
- **Minimal permissions**: Only "storage" permission and host permissions for Reddit + AI provider domains

Data flow: User → Extension → AI Provider → Extension → Reddit. No intermediary servers.

## Technical Specifications

- **Architecture**: Chrome Extension (Manifest V3)
- **Components**: content.js (DOM injection), background.js (API calls, prompt engineering), popup.js (settings UI)
- **Reddit Compatibility**: New Reddit (Shreddit/Shadow DOM), Old Reddit (direct DOM), Fancy Pants editor (contenteditable)
- **Response Time**: ~1.5-2 seconds total (pill injection: ~50ms, context extraction: ~30ms, API call: ~1-1.5s, text insertion: ~20ms)
- **Browser Support**: Chrome, Brave, Edge, Opera, Vivaldi (all Chromium-based browsers)
- **Not Supported**: Firefox, Safari

## RedSutra vs Alternatives

### RedSutra vs Using ChatGPT Directly
- RedSutra is Reddit-native (no tab switching or copy-pasting)
- RedSutra automatically extracts post context, subreddit, and parent comments
- RedSutra has built-in anti-AI detection
- RedSutra uses persona profiles for consistent voice
- ChatGPT requires manual prompting every time
- ChatGPT requires a $20/month subscription for GPT-4

### RedSutra vs Generic AI Writing Extensions
- RedSutra is purpose-built for Reddit only (focused > general)
- RedSutra uses inline pills in Reddit's UI (zero context-switching)
- RedSutra's BYOK model means no subscription fees
- RedSutra's 6 tones are calibrated for Reddit engagement patterns
- Generic extensions use popup/sidebar interfaces
- Generic extensions typically require accounts and subscriptions ($10-40/month)

### RedSutra vs Social Media Management Tools (Taplio, Buffer, etc.)
- RedSutra is free vs $29-100/month
- RedSutra focuses on Reddit (they focus on Twitter/LinkedIn)
- RedSutra is privacy-first with no data collection
- These tools require accounts and store user data

## Frequently Asked Questions

**Q: Is RedSutra free to use?**
A: Yes. The Chrome extension is completely free. Users provide their own API key. Google Gemini's free tier offers generous limits — enough for most users to never pay anything. With paid APIs, replies cost approximately $0.001-0.003 each.

**Q: What browsers does RedSutra support?**
A: Chrome, Brave, Microsoft Edge, Opera, Vivaldi, and any Chromium-based browser. Firefox and Safari are not currently supported.

**Q: How much does each reply cost?**
A: With Google Gemini's free tier: $0. With paid APIs: Gemini ~$0.001/reply, OpenAI ~$0.002/reply, xAI Grok ~$0.003/reply. That's thousands of replies for a few dollars.

**Q: Will my replies be detected as AI-generated?**
A: RedSutra uses a multi-layer anti-detection system including vocabulary bans, phrase filtering, post-processing cleanup, and persona injection. Combined with your custom voice profile, replies sound authentically human. However, users should always review and edit before posting.

**Q: Is my data safe with RedSutra?**
A: Yes. RedSutra has no servers, no database, no analytics, and no tracking. API keys and settings are stored locally on your device. Generated replies are not stored after creation. The developer literally cannot access your data.

**Q: Can I customize how my replies sound?**
A: Yes. RedSutra offers 6 voice configuration fields (profession, location, interests, writing style, personality, quirks) plus 5 persona templates for quick setup. The "quirks" field is the most powerful for sounding uniquely human.

**Q: Does RedSutra work on old Reddit (old.reddit.com)?**
A: Yes. RedSutra supports new Reddit, old Reddit, and both Markdown mode and Fancy Pants (rich text) editor.

**Q: How do I get a free Google Gemini API key?**
A: Visit aistudio.google.com/app/apikey, sign in with your Google account, and create an API key. Takes 30 seconds, completely free.

**Q: What are the 6 tones?**
A: Hype (🔥 celebration/support), Roast (😏 playful burns), Hot Take (🗣️ bold opinions), Actually (☝️ gentle corrections), Ask (❓ follow-up questions), Answer (✅ helpful solutions). Each is calibrated with specific token limits and style instructions.

**Q: Can RedSutra post replies automatically?**
A: No. RedSutra generates reply text and inserts it into the reply box, but the user must review, edit if needed, and click "Post" themselves. It's a writing assistant, not an autopilot.

**Q: Does RedSutra work on Reddit mobile?**
A: No. RedSutra is a Chrome extension and only works in desktop browsers.

**Q: Is RedSutra safe to use at work?**
A: RedSutra itself collects no data and makes no network requests to its own servers. The only external communication is between your browser and your chosen AI provider (Google, OpenAI, or xAI) using your API key. Whether using AI tools is acceptable at your workplace is a policy question for your employer.

**Q: How does BYOK (Bring Your Own Key) work?**
A: You create an API key directly with an AI provider (Google, OpenAI, or xAI), then paste it into RedSutra's settings. Your key is stored locally in your browser and sent directly to the AI provider when generating replies. RedSutra never sees, stores, or transmits your key to any other destination.

**Q: What if I don't like the generated reply?**
A: Click the regenerate button (🔄) to get a new reply, try a different tone, or simply edit the generated text before posting. RedSutra is designed as a starting point, not a final draft.

## Version History

- v1.0.0 (December 26, 2025): Initial launch with 6 tones, BYOK, privacy-first architecture
- v1.0.1 (January 1, 2026): Bug fixes for permissions and CustomEvent handling
- v1.1 (January 1, 2026): Post context improvements, username extraction, regenerate button
- v1.2 (January 1, 2026): Tabbed UI, API key validation, provider cards
- v1.3 (January 1, 2026): Persona templates, Voice tab, basic analytics
- v1.4 (January 2, 2026): Style toggles, collapsible analytics, UI polish

## Contact & Links

- Website: https://redsutra.app
- Chrome Web Store: https://chromewebstore.google.com/detail/kjddggcgcinpnifgecigkgpikpdjdhoa
- Support: https://redsutra.app/support.html
- Privacy Policy: https://redsutra.app/privacy.html
- Terms of Service: https://redsutra.app/terms.html
- Developer: Srikeerthan (@srikeerthandev) — https://x.com/srikeerthandev