# HyPeMa — Web App Guide > The whole product in one page. About 10 minutes to read end-to-end. --- ## Table of Contents - [Quickstart](#quickstart) - [Chrome Extension](#chrome-extension) - [How HyPeMa writes emails](#how-hypema-writes-emails) - [AI Settings](#ai-settings) - [AI Company Research](#ai-company-research) - [Campaigns & History](#campaigns--history) - [Billing & Usage](#billing--usage) - [API Keys](#api-keys) - [FAQ & Troubleshooting](#faq--troubleshooting) --- ## Quickstart HyPeMa turns cold outreach into three steps, start to reply. The whole loop takes about ten minutes the first time. ``` Find Leads --> Draft Emails --> Review & Send (AI, CSV, (Personalize (Create Gmail manual) with AI) drafts) ``` **1. Find leads** — Describe your target criteria once (industry, role, region, company size) and AI returns real people with verified emails. Or upload a CSV (columns auto-map). Or add leads by hand. Leads live in groups so you can reuse them. **2. Draft emails** — Pick a group, write one sentence about what you're pitching ("AI bookkeeping for restaurants, book a short call"), preview three sample emails, add feedback if you want them tighter, and hit **Draft Emails**. The AI writes a unique email per lead using what it found about their company. **3. Review & send** — The Chrome extension places every email as a draft in your real Gmail. Edit any one inline, then click **Send All via Gmail** to fire them off. You still hit Send yourself — HyPeMa never sends on your behalf, so your sender reputation stays intact. ### Lead status flow ``` new --> generated --> drafted --> sent ``` - **new** — lead added, no email yet - **generated** — AI has written a personalized email - **drafted** — email placed as a Gmail draft - **sent** — email sent from Gmail After sending, come back to your Workspace and click **Sync**. The extension scans your Gmail for replies and rolls them up into each campaign's reply rate. > During onboarding you entered your company website URL (or details manually). That context is what the AI uses to personalize every email. You can update it anytime in **Settings → General**. Watch the full flow: https://www.youtube.com/watch?v=LjnyHAijGFM --- ## Chrome Extension The extension does one job: turn HyPeMa's generated emails into Gmail drafts. It never sends. It never reads your inbox — except the one scoped reply-sync scan you trigger from the dashboard, which matches thread subjects to sent leads. ### Architecture ``` HyPeMa Web App --> Chrome Extension --> Gmail (Your Drafts) postMessage opens compose ``` ### Install Get it from the [Chrome Web Store](https://chromewebstore.google.com/detail/hypema-%E2%80%94-precise-email-ou/apnbcdldkcamjajnlmcifgeabhemonpf). Keep the HyPeMa web app open in a tab for the extension to work — they communicate via the browser. ### Why drafts, not direct send? Emails come from your real Gmail, so they inherit your sender reputation. No warmup, no spam-folder roulette. And you review every email before it goes out. ### Popup says "Not Open"? Refresh both the HyPeMa tab and the extension popup. They need to be in sync for the bridge to connect. Watch the extension in action: https://www.youtube.com/watch?v=CfUwpzov-Ow --- ## How HyPeMa writes emails HyPeMa doesn't ghostwrite. It *structures*. Every email the AI produces follows six rules distilled from Lavender, Outbound Squad, and Gong's 28M-email dataset — the research on what actually earns replies. **1. Short & scannable** — Subject: 1–4 words, lowercase. Body: 50–100 words in 3–5 short lines. Short words, short sentences. **2. About them, not you** — The opening references one specific detail from the research about their company or role. Never fabricated — every claim traces back to what the research actually found. **3. One proof, no features** — Exactly one sentence of proof from your own company. No feature lists, no ROI claims like "2× reply rate" — buyers filter those out. **4. One easy question** — The email ends with one binary, easy-to-answer question. No calendar links, no "book a 15-minute call" — that's for the follow-up, not the first touch. **5. Peer-to-peer tone** — Informal. No "hope this finds you well", no "I came across your profile", no hype. If a line sounds like it's being sold, it gets rewritten. **6. Discipline** — Product name mentioned once. Max one "we/our" sentence in the body. Signature is always "Kind regards, {your first name}" — same way you'd sign any other email. These six rules apply to every email — even on the free tier — because they're the common ground the best cold-email research keeps converging on. ### Secret Sauce — the proprietary layer Depending on the **Secret Sauce** mode you choose, the AI may additionally apply a set of proprietary techniques we don't publish: specific opener patterns, objection-preemption structures, and framing rules we've tuned from our own campaigns. That's what the extra tokens buy. The six rules above get you 80% of the way there; Secret Sauce is the last 20%. --- ## AI Settings Two controls shape every email: **Secret Sauce** (which copywriting framework the AI uses) and **Thinking Depth** (how much reasoning it spends). Set defaults in **Settings → General**, override per campaign on the Draft Emails page, or override per request via the API/MCP `effort` and `secretSauceVariant` parameters. ### Secret Sauce — three modes | Mode | Tokens/email | Best for | |------|--------------|----------| | **Reply Rate** | ~5–30 | Maximizes any response. Curiosity hooks, open loops, easy questions. Awareness, networking, research. | | **Positive Reply Rate** | ~8–51 | Maximizes interested replies. Outcome-framing, objection preemption, low-commitment CTA. Sales, partnerships, recruiting. | | **Cost-efficient** (`none`) | ~5–25 | Standard AI copywriting without Sauce. Lowest cost per email. High-volume campaigns. | Global default lives in Settings. Override per campaign in the Draft Emails Secret Sauce selector. Choose **Ask Every Time** at the tenant level to force a per-campaign decision. ### Thinking Depth — five effort levels Higher effort = deeper reasoning = better results = more tokens and longer wait. Each operation (Email, Lead Generation, Company Research, CSV Mapping) has its own tab in Settings. | Level | Email tokens | Email time | Lead gen time | Company research | |-------|--------------|------------|---------------|------------------| | **Low** | ~4–6 | ~6–7 sec | ~3 min | ~30 sec | | **Medium** | ~4–6 | ~6–9 sec | ~5 min | ~1 min | | **High** (default) | ~4–12 | ~7–21 sec | ~10 min | ~3 min | | **Extra High** | ~8–18 | ~17–39 sec | ~12 min | ~4 min | | **Max** | ~21–52 | ~45 sec–2 min | ~15 min | ~5 min | `Extra High` and `Max` are **Opus 4.7 only**. Email numbers measured 2026-04-19 via a 75-run benchmark across 5 profile-pairs × 3 secret-sauce variants × 5 effort levels. Lead gen / research numbers are interpolated from earlier measurements — they'll be refined when those operations get benchmarked directly. Under the hood, Thinking Depth controls Anthropic's extended-thinking effort. Higher levels let Claude reason for longer before writing. Those reasoning tokens are billed as output but not shown. API/MCP calls accept `effort` as a per-request override. ### Example — a real campaign **40-lead partnerships campaign on Pro.** You choose **Positive Reply Rate** (sales framing) and **Medium** Thinking Depth (balanced quality). - 40 leads × ~17 tokens/email = **~680 tokens** - That's ~14% of your 5,000 monthly Pro allowance. - Batch takes ~30 minutes total (45 sec × 40, with some parallelism). - ~4,320 tokens left for other campaigns this month. Drop to **Low** depth and the same batch runs in ~10 minutes for **~400 tokens** (~8% of monthly). Switch back to **Cost-efficient** Sauce and you're at ~200 tokens for the whole batch — at the cost of less opinionated copy. --- ## AI Company Research When you enter a company URL, AI uses real-time web search to find out what the company does, who runs it, what they ship, and who they sell to. That structured data is what every email references — it's the difference between "Hi Jane," and "Saw your Q2 launch post — congrats." ### Research flow 1. **Enter company URL** (e.g. `https://example.com`) 2. **AI web search** (30 sec – 5 min) 3. **Structured data** — Name, industry, CEO, product, customer profile, value prop 4. **Email personalization** — powers every outreach email ### Processing time 30 seconds to 5 minutes, depending on how much information is public. You'll see live status updates as the AI searches. Don't refresh the page — it runs in the background. --- ## Campaigns & History Every time you generate emails, HyPeMa auto-creates a **campaign** that groups the leads, emails, and sent/reply status. No manual setup. Campaigns are how you answer *"what happened after I sent?"* ### What the Campaign Detail page shows Four stat cards at the top — total leads, emails generated, emails sent, reply rate. Below that, a filterable list of every email in the campaign. | Tab | Shows | |-----|-------| | **All** | everything | | **Sent** | drafts sent via Gmail | | **Replied** | leads who responded | | **No Reply** | sent but silent | Click any row to expand — you see the full email that was sent, plus a preview of the reply if one came back. If a reply's there but the preview isn't, open Gmail for the full thread. ### Reply tracking — three things to know 1. **Gmail must be open in a tab** when you click Sync. The extension reads Gmail through the DOM, not via the Gmail API. 2. **Replies are matched by thread subject.** Gmail prepends "Re: " when someone responds — the extension matches those threads back to the emails you sent from HyPeMa. 3. **Sync is user-triggered.** HyPeMa doesn't poll Gmail in the background. Click the sync button when you want fresh numbers — reply rate updates on the Workspace and the Campaign Detail page. ### Re-running a campaign Open any campaign and click **Review & Send** to push emails to Gmail again — useful if drafts got cleared. Each lead's status (generated / drafted / sent) is preserved so you don't re-send duplicates. --- ## Billing & Usage ### Plans | Plan | Tokens | Price | Description | |------|--------|-------|-------------| | Free | 500 (one-time credit on signup) | €0 | Try it risk-free | | Pro (monthly) | 5,000/mo | €49/mo | For serious outreach campaigns | | Pro (yearly) | 5,000/mo | €441/yr (25% off) | Same Pro features, billed annually | Purchased token packs never expire and kick in after your monthly allowance runs out. ### Token costs by operation Costs vary by Secret Sauce mode and Thinking Depth. Numbers below are ranges at Low effort. | Operation | Cost | |-----------|------| | Company Research | 5–15 tokens | | Lead Generation | 10–30 tokens | | Email Generation | 2–10 tokens | | Managing Leads, Groups, Campaigns | Free | Only AI-powered operations cost tokens. Managing leads, groups, and settings is always free — tokens are only consumed when the AI does work for you. ### Check usage **Settings → Usage** — token consumption, daily chart, operation history. ### Upgrade **Settings → Billing** → pick a plan. --- ## API Keys Access HyPeMa programmatically — for custom integrations, automation, or connecting to other tools. ### Creating a key 1. Go to **Settings → API Keys** 2. Click **Create API Key** and name it 3. Copy the key — it's shown only once ### Security Keys start with `hpm_live_` and are hashed in the database. Never share your key — revoke immediately if compromised. ### Using a key ``` Authorization: Bearer hpm_live_your_key ``` Full API docs at [API Reference](/academy/api-reference). MCP setup at [MCP Integration](/academy/mcp). --- ## FAQ & Troubleshooting **Q: The Chrome extension shows "Not Open"** A: Keep the HyPeMa web app open in a browser tab. The extension talks to the web app via the browser. Refresh both and try again. **Q: I ran out of tokens** A: Upgrade to Pro in Settings → Billing for 5,000 tokens/month, or buy a token pack that never expires. The 500-token free tier is a one-time signup credit — it does not reset monthly. **Q: Email generation is taking very long** A: AI research with web search can take 1–5 minutes for complex companies — this is normal. If it's consistently slow, lower the Thinking Depth for Email in Settings → General (Max → Medium cuts generation from ~5 min to ~45 sec per email). **Q: I get "Rate limited" errors** A: HyPeMa limits concurrent AI operations to keep things stable. Wait for the current batch to finish before starting a new one. **Q: Can I use HyPeMa with Outlook instead of Gmail?** A: Gmail only for now (personal and Google Workspace). Outlook is on the roadmap. **Q: How accurate are the AI-generated leads?** A: Leads are found via real-time web search. Email addresses come from public websites or are constructed from company domain patterns, so some will bounce. Review before sending. Still stuck? Email **support@hypema.app**.