Introducing Skylarq Assistant: One Desktop App to Replace Your Entire Sales Stack

Today we are launching Skylarq Assistant — a macOS desktop app that replaces 15 sales tools with one AI agent that actually does things. It runs on your machine, controls your browser, and works while you sleep. Here is why we built it, what it does, and how to get started.

The Founding Story

In 2020, we built Homebase — a proptech platform helping people in Southeast Asia become property owners. We went through Y Combinator's W21 batch, raised $50 million, and scaled the company to 120 employees. During that stretch, we ran a full sales motion: inbound leads from paid and organic, outbound prospecting to property buyers and partners, enterprise sales cycles, partnership conversations, investor outreach. The whole thing.

At Homebase's peak, the founders used 15+ disconnected sales tools—Apollo for leads, Instantly for email, Granola for meeting transcription, HubSpot for CRM, and more—with no meaningful integration between them. This fragmentation required constant manual data entry and context switching that created zero revenue, consuming hours that should have gone to customer conversations.

To run it, we used Apollo to find leads. Instantly to send email sequences. Granola to record and transcribe meetings. Whisper Flow for voice notes. Calendly for scheduling. Zapier to stitch things together. HubSpot for the CRM. And half a dozen more tools on top of that. At peak, we were paying for 15 or more subscriptions that touched the sales workflow.

None of them talked to each other. Not in any meaningful way.

Apollo did not know what Instantly had sent. Instantly did not know what had happened on the last call. Granola produced transcripts that lived in a folder nobody read. Zapier automations broke and nobody noticed for days. HubSpot was perpetually out of date because updating it manually was something nobody had time for. The glue between all of these systems was a human — one of us, usually — doing data entry and context switching that created zero revenue.

The actual selling — the relationship-building, the discovery, the close — got squeezed into whatever time was left after the tool management overhead. That is backwards. Every hour spent copying data between Apollo and HubSpot is an hour not spent talking to a customer.

We sold Homebase and started Skylarq to solve this problem. Not with another point solution. With a single AI agent that owns the entire sales workflow from prospecting through follow-up, runs autonomously on your own machine, and does not require you to be the integration layer.

We spent 14 months building it. Today we are shipping it.

The Team

Building an AI agent that actually works — one that can reason about a prospect, navigate a browser, write a genuine message, and operate reliably on a schedule — requires a specific combination of product intuition, AI depth, and engineering execution. We have been deliberate about assembling that combination.

The Skylarq founding team includes Phillip An (CEO, ex-Homebase YC W21, $50M raised, Harvard MBA, Forbes 30 Under 30), Jack Zhu (CTO, former AI lead at Wish through its $14B IPO), and a founding engineer who previously served as CTO of a YC company. The team combines go-to-market experience with production-grade AI engineering.

Phillip An (that is me) is co-founder and CEO. Before Skylarq, I built Homebase through YC W21 to 120 employees and $50M raised. I am a Harvard Business School MS/MBA, a Schwarzman Scholar, a Forbes 30 Under 30, and a former McKinsey consultant. My background is at the intersection of go-to-market strategy and product — which is exactly where Skylarq lives.

Jack Zhu is co-founder and CTO. Jack led AI at Wish during its growth to a $14 billion IPO — one of the largest e-commerce companies in the world. He spent years building production AI systems at scale, which is a fundamentally different problem than building an AI demo. The gap between "impressive prototype" and "reliable system you can trust with your outreach" is enormous. Jack knows how to close that gap.

Our founding engineer was previously CTO of a YC company. He has built and scaled engineering organizations from zero, which means we are not figuring out basic infrastructure problems — we are executing.

We are a lean team. That is by design. Small teams ship faster, have tighter feedback loops, and do not build features nobody asked for. We would rather ship something that works and iterate quickly than build in stealth for three years.

What Skylarq Assistant Actually Is

Skylarq Assistant is a native macOS desktop app. Not a browser extension. Not a SaaS dashboard you log into. Not a Chrome plugin that injects itself into LinkedIn. A proper desktop app that runs on your machine.

Skylarq Assistant is a native macOS desktop application that runs locally on your machine. All data — contacts, messages, meeting recordings, and pipeline information — stays on your device. The app uses your own Anthropic API key for AI processing, controls a real browser for LinkedIn automation, and runs continuously in the background executing scheduled tasks 24/7.

This matters for several reasons.

First, it means your data stays local. Your contacts, your messages, your meeting recordings, your pipeline — none of it gets uploaded to our servers. The AI computations happen via your own API key directly to Anthropic. We are not a middleman sitting on your customer data.

Second, it means the app can control your browser. When Skylarq sends a LinkedIn connection request, it opens a real browser session, navigates to the profile, reads it, and types a personalized note. It does not use an unofficial API or scrape data through a backend route. It does exactly what you would do if you were doing it manually — just faster and at scale.

Third, it means the app runs continuously. It is not a web session you open and close. It runs in the background, executing scheduled routines, monitoring for responses, and processing incoming data, whether you are in front of your computer or not.

The mental model is not "this is another sales tool." The mental model is "this is an AI employee who works the sales function while I focus on everything else."

The Five Features

Skylarq Assistant ships with five integrated capabilities. They are not independent modules — they share context, memory, and data. What happens in a meeting informs your outreach. What your agent learns about a prospect informs how it responds. The integration is the point.

Skylarq Assistant's five core features are: (1) Skills — 8 automated workflows including Morning Briefing, Email Triage, and Outreach Cadence that run on schedule; (2) Leads — AI-powered LinkedIn outreach with 60% acceptance rates at 14 connections per day; (3) Agents — 24/7 AI responders; (4) Meetings — invisible recording with auto follow-up; (5) Voice — hands-free commands 4x faster than typing.

1. Skills: Automated Workflows That Run on Your Schedule

Skills are automated workflows that execute on a schedule, without you having to trigger them. You configure them once and they run every day, every week, or on whatever cadence you set.

The launch version ships with eight built-in Skills:

• Morning Briefing — Delivered at 8am, summarizing your pipeline status, meetings for the day, and top prospects who need follow-up. Built from your calendar, email, and CRM data.
• Email Triage — Categorizes your inbox, surfaces the messages that need a response today, and drafts replies for your review.
• Pipeline Review — Analyzes your active deals, flags stalled opportunities, and suggests next actions based on deal stage and time since last contact.
• Prospect Research — Automatically researches leads in your queue, pulling together LinkedIn profile, company news, recent funding, and relevant talking points before you reach out.
• Follow-Up Queue — Identifies every prospect you said you would follow up with but have not yet, drafts the messages, and queues them for your review.
• Meeting Prep — 30 minutes before each meeting, generates a briefing on the attendees: their background, company, recent activity, and talking points based on your shared history.
• Weekly Wrap — Every Friday afternoon, summarizes what happened this week: meetings held, deals advanced, deals at risk, and priorities for next week.
• Outreach Cadence — Runs your daily LinkedIn and email outreach automatically, personalizing each message based on fresh prospect research.

Skills are configurable. You can adjust timing, scope, and behavior. And you can build custom Skills for your specific workflow using natural language — describe what you want to happen and when, and the app sets it up.

2. Leads: AI-Powered LinkedIn Outreach

The Leads feature handles your LinkedIn outreach end-to-end. You define your ideal customer profile — industry, company size, job title, geography, seniority — and the agent identifies matching prospects, researches their profiles, and sends personalized connection requests with tailored notes.

We send 14 connections per day by default. That cadence keeps you well within LinkedIn's limits while maintaining a steady pipeline of new prospects. Our users see roughly 60% acceptance rates on connection requests — significantly higher than the 20-30% typical for template-based tools — because each note references something specific to that person's background or company.

The timing is human-like. Requests go out spread across business hours, not in a batch at 3am. The app inserts natural pauses between actions. It reads profiles before connecting the same way a human would. This is not reckless automation — it is automation designed to look, and to a large extent be, indistinguishable from a thoughtful human doing the same work.

After a connection is accepted, the agent sends a first message. If there is no response after a few days, it follows up. Every interaction is logged, and the agent uses that history to inform subsequent messages. It never sends the same message twice to the same person.

3. Agents: AI Employees That Respond 24/7

Agents are always-on AI employees that monitor your channels — email, LinkedIn, and more — and respond to inbound activity around the clock.

When a prospect replies to your outreach at 11pm, the agent responds within seconds. Not a canned auto-reply — a genuine response that reads their message, understands the context, and advances the conversation. If they ask a question, the agent answers it. If they want to book a meeting, the agent checks your calendar and proposes times. If they raise an objection, the agent addresses it.

The response time matters more than most people realize. Research from MIT and Harvard Business Review consistently shows that responding to a prospect within five minutes dramatically improves qualification rates — some studies show 21x improvement over a 30-minute response. An AI agent that responds in under five seconds, 24 hours a day, creates a structural advantage that no human team can match.

You stay in control. The agent can be configured with different levels of autonomy: fully autonomous, review-before-send, or suggest-only. For outbound follow-up, most users run fully autonomous. For warm inbound from high-value prospects, most prefer to review before sending. You set the threshold.

4. Meetings: Invisible Recording and Intelligent Follow-Up

The Meetings feature handles the full meeting lifecycle. It records your calls invisibly in the background, transcribes them in real time, and generates structured output: key discussion points, commitments made, objections raised, action items with owners, and next steps agreed upon.

That alone is table stakes — Granola, Otter, and Fireflies all do something similar. Where Skylarq differs is in what happens after transcription. The meeting becomes an input to action.

If you promised to send a case study, the agent queues a follow-up email with the case study attached. If the prospect mentioned they need to bring in their legal team, the agent notes that the deal is entering a procurement stage and adjusts the pipeline entry accordingly. If they said "let us reconvene in two weeks," the agent schedules the follow-up and adds it to your calendar.

Meeting intelligence that does not trigger follow-up action is just a more readable transcript. We built Skylarq's Meetings feature to close that loop automatically.

5. Voice: 4x Faster Than Typing

Voice commands let you control the entire app hands-free. Between calls, on the go, or just when typing feels slow, you can issue commands that trigger real actions across all five features.

"Follow up with everyone from yesterday's meetings." Done. The agent reviews the transcripts, identifies the people and the commitments, and queues the follow-up emails for review.

"What is the status of my deal with Acme Corp?" The agent pulls the timeline: last contact, last meeting, what was discussed, what was promised, where it stands in the pipeline.

"Add Sarah Johnson from TechCorp to my outreach queue." The agent finds her LinkedIn profile, adds her to the leads list, and schedules a connection request for tomorrow morning.

Voice is four times faster than typing for natural language commands. For a sales rep whose day is full of calls and conversations, having a hands-free interface to their AI agent is not a convenience feature — it is a fundamental change in how the tool fits into their workflow.

How It Is Different From Everything Else

The AI sales tool market is noisy. Almost every category of software now has "AI" in its marketing copy. Here is what actually differentiates Skylarq:

Skylarq Assistant differentiates from competitors like Apollo, Instantly, and 11x.ai in four ways: local-first architecture where no data leaves your device, real browser-based LinkedIn automation instead of fragile API scraping, true cross-feature integration through a shared memory layer, and Claude-powered long-context reasoning via your own Anthropic API key rather than a GPT wrapper.

Pricing

Skylarq Assistant is free to download and use. You bring your own Anthropic API key.

Skylarq Assistant is free to download with no subscription fee. Users pay only Anthropic API costs, typically $20–50 per month for daily outreach, briefings, and meeting processing. This replaces a sales tool stack costing $500–2,000+ per month (Apollo at $49–99, Instantly at $37–97, Granola at $18, Zapier at $20–50, plus CRM fees). Paid tiers will be introduced for future advanced features.

Typical usage — daily outreach, morning briefings, meeting processing for 3-5 calls per week, and occasional agent responses — runs $20-50 per month in API costs. That is what you pay directly to Anthropic. We do not mark it up.

For context: Apollo costs $49-99 per month per user. Instantly is $37-97 per month. Granola is $18 per month. Zapier is $20-50 per month. A modest stack of these tools runs $150-300 per month before you start talking about Salesforce or HubSpot. A full enterprise stack can easily hit $1,000-2,000 per month per user.

Skylarq replaces all of it for the cost of API calls.

This model works because we are not storing your data, running servers on your behalf, or maintaining API integrations that break. The compute you use goes to Anthropic. The app is the product. We will introduce paid tiers as we add features — but the core functionality ships free.

How to Get Started

Setup takes under three minutes:

Getting started with Skylarq Assistant takes under 3 minutes in 5 steps: download the macOS app, add your Anthropic API key from console.anthropic.com, connect LinkedIn and email via browser session and OAuth, configure your ideal customer profile for leads, and enable the Morning Briefing skill. Your first AI-generated briefing arrives the next morning at 8am, and most users see their first prospect response within 48 hours.

Step 1: Download the app. Download Skylarq Assistant for Mac. Open the .dmg file and drag it to your Applications folder.

Step 2: Add your API key. Go to console.anthropic.com, create an API key, and paste it into Skylarq's settings. This is how the app accesses Claude — your key, billed directly to your account.

Step 3: Connect your accounts. Connect your LinkedIn account, your email (Gmail or Outlook), and your calendar. The app uses browser sessions for LinkedIn and standard OAuth for email and calendar.

Step 4: Configure your ICP. Tell the Leads feature who you are looking for: job title, industry, company size, geography. This takes two minutes and can be updated at any time.

Step 5: Enable Morning Briefing. Turn on the Morning Briefing Skill and set your preferred time. Tomorrow morning at 8am, your first briefing will arrive — a summary of your pipeline, your meetings for the day, and the prospects who need attention.

That is it. The app will start working. Your first outreach will go out on the schedule you set. Your agent will start monitoring your channels. Your meetings will start getting recorded and processed automatically.

Most users see their first inbound response to AI-generated outreach within 48 hours.

What Is Next

Today's release is the foundation. Here is what we are building toward:

Skylarq's roadmap includes Windows support (shipping within months), a mobile companion app for reviewing briefings and approving outreach on the go, additional Skills such as Competitor Monitoring, Deal Risk Alerts, and Referral Requests, a Skills Marketplace for community-built workflows, and deeper CRM integrations that automatically push meeting transcripts and engagement signals into Salesforce and HubSpot.

Windows support. Mac-first is a pragmatic decision, not a permanent one. The architecture is platform-portable. Windows is in active development and will ship within a few months. If you are on Windows and want early access, join the waitlist from the download page.

Mobile companion app. The desktop app does the heavy lifting — outreach, processing, scheduling. The mobile companion gives you a lightweight view on the go: review your morning briefing, approve queued outreach messages, check your pipeline at a glance. Designed for the moments between meetings when you are not at your desk.

More Skills. We ship with eight Skills and will release more on a regular cadence. Upcoming Skills include Competitor Monitoring (track mentions of your competitors and flag relevant prospects), Deal Risk Alerts (identify signals that a deal is going quiet), and Referral Requests (identify customers who are highly engaged and could make warm introductions).

Skills Marketplace. Eventually we want to open the Skills system so that users can build and share custom workflows. The primitives are already there — Skills are defined in a structured format that can be templated and distributed. We will ship a marketplace once the core experience is solid and we have a clear sense of what community-built Skills people actually want.

Deeper CRM integrations. The current version connects to Salesforce and HubSpot for reading and writing pipeline data. We are building deeper integrations that push more signal into your CRM automatically — meeting transcripts, sentiment signals from conversations, engagement data from outreach — so your CRM reflects what is actually happening rather than what you manually updated.

If you want to follow along, the best place is the Skylarq blog. We will document what we are building, why, and what we learn as we go.