How to vibe codeSalesTouch
LinkedIn prospecting infrastructure for AI agents
salestouch.io ↗B2B SaaS / Developer Tools
The verdict: can you vibe code SalesTouch?
Build a personal subset for local script execution, but keep paying if you want production reliability; the single biggest catch is LinkedIn's aggressive anti-abuse systems and proxy management.
SalesTouch is essentially a specialized scraping and automation wrapper disguised as an MCP server. While a solo developer can spin up a TypeScript Node.js server that speaks the Model Context Protocol in a weekend, replicating the underlying plumbing—residential proxy rotation, automated session recovery, cookie persistence, and rate-limiting jitter that survives LinkedIn's aggressive bot detection—is a Sisyphean engineering nightmare. One silent DOM update or bad IP block results in account termination. If you value your LinkedIn profile, paying $49/mo is far cheaper than the opportunity cost of debugging ban notices.
Estimated effort: 3-4 weeks of debugging and proxy configuration
What you can't replicate
- Production-grade residential proxy pool matched to user geographies
- Battle-tested anti-detection heuristics and automated cooldown algorithms that prevent account bans
- Up-to-the-minute resilient scrapers for LinkedIn's frequently shifting DOM
Founded
—
Raised
—
Team
—
Cheapest paid tier
$49/mo
What SalesTouch does
SalesTouch exposes a remote Model Context Protocol (MCP) server with 26 typed tools that connect AI clients like Claude, Codex, and Cursor directly to LinkedIn for prospecting, research, outreach, and engagement.
Core features
- Remote MCP server with HTTP transport implementing OAuth 2.1 authentication
- 26 typed tool definitions for audience building, profile research, messaging, and public engagement
- Residential IP proxy rotation mapped to user country zones to evade detection
- Asynchronous background worker queue with randomized jitter and human-paced timing windows
- Encrypted credential and 2FA session management store
- Safety throttle with automated circuit breakers and daily action caps
- Multi-account workspace orchestration and audit log storage
The business
Pricing
- Monthly Plan$49/mo
- Annual Plan$389/yr
Funding
Unknown / bootstrapped
Pay vs build, cumulative
Break-even at month 1 — after that, every month is money kept.
The hard parts of vibe coding SalesTouch
- Bypassing LinkedIn's sophisticated bot-detection and anti-scraping firewalls without getting accounts permanently banned
- Maintained DOM/API structure wrappers against silent breaking changes from LinkedIn front-end updates
- Stateful background queuing that respects complex calendar windows, working timezones, and automatic cooldown pauses
- OAuth 2.1 token management and secure credential re-authentication loops for multi-factor challenges
How to vibecode SalesTouch
Prerequisites
Node.jsfree
Required runtime for executing the TypeScript MCP server backend.
GitHubfree
Repository hosting and deployment pipeline integration.
Residential Proxy Provider Account~$15/mo
Required to route scraping requests through IP addresses matching your country to avoid immediate flagging.
AI coding tools
Recommended stack
| Frontend | Next.js |
|---|---|
| Backend | TypeScript / Express MCP Server |
| Database | Turso |
| Auth | better-auth |
| Payments | Stripe |
| Other | Playwright for browser automation fallback, Zod for type-safe MCP tool schemas |
Hosting & infrastructure
| Fly.io | Hosting the Node.js remote MCP server with persistent IP routing and background queues | $5/mo |
Build guide
01Scaffold the Remote MCP Server
Initialize a TypeScript Node.js project using Express and the Model Context Protocol SDK to expose HTTP-transport endpoints.
Create a new TypeScript project configured as an HTTP-transport MCP server using the official Model Context Protocol SDK and Express. Set up structured logging with Winston, Zod for tool schema validation, and health check endpoints. Ensure the server accepts incoming POST requests at /api/mcp with proper CORS and authentication headers. Add a modular structure where tools are registered dynamically from a tools directory.02Implement Session & Proxy Management
Build a secure session manager that handles encrypted LinkedIn cookies, authentication tokens, and residential proxy routing.
Implement a session manager module in TypeScript that securely stores encrypted LinkedIn session cookies and proxy configurations using Node's crypto module and Turso for persistence. Configure a custom Axios instance or Playwright browser context injector that dynamically routes all outgoing requests through a configured residential proxy with geo-matching headers to prevent immediate detection by LinkedIn anti-bot systems.03Build Core Extraction & Research Tools
Develop the first batch of MCP tools for scraping people searches, post reactions, profile views, and company feeds.
Implement the core extraction and research MCP tools: linkedin.lookup_profile, linkedin.search_people, linkedin.post_reactions, and linkedin.read_feed. Write robust parsing logic using cheerio or structured DOM selectors over fetched LinkedIn markup, incorporating randomized user-agent strings and human-like browsing headers. Handle pagination gracefully and return clean, normalized JSON structures suitable for AI agent consumption.04Implement Outreach & Engagement Actions
Add tools for sending connection requests, direct messages, comments, and likes with built-in daily safety caps.
Add outreach and engagement tools to the MCP server: linkedin.send_invitation, linkedin.send_message, linkedin.like_post, and linkedin.comment_post. Implement strict validation checks against daily limits stored in Turso, and integrate an asynchronous background queue (using BullMQ or native timers with randomized jitter) to enforce human-paced sending windows across configured working hours.05Deploy and Integrate with AI Clients
Deploy the server to Fly.io and configure local AI clients like Claude Code or Cursor to connect via HTTP transport.
Create a Dockerfile optimized for a Node.js TypeScript application running on Fly.io, configuring persistent volumes for encrypted credential storage and environment variables for Turso and proxy providers. Document the exact CLI commands (e.g., claude mcp add) required to connect local AI clients securely to the deployed remote MCP endpoint.
Cost vs paying for SalesTouch
What will you build it with?
Starting total with Claude Code~$15 one-time
Starting costs (one-time)
- Residential Proxy Starter Package$15
- Domain registration$12
Total~$27 one-time
Ongoing costs (monthly)
- Fly.io Server Hosting$5/mo
- Residential Proxy Subscription$15/mo
Total~$20/mo
Paying for SalesTouch
$49/mo
Your time to build
40 hours
AI tool credits
$20
Break-even
Never (risk of account ban makes paid SaaS safer)
Vibe code SalesTouch: FAQ
- Can you vibe code SalesTouch yourself?
- Don't bother — 25/100 vibecodeable. Build a personal subset for local script execution, but keep paying if you want production reliability; the single biggest catch is LinkedIn's aggressive anti-abuse systems and proxy management.
- How long does it take to vibe code SalesTouch?
- 3-4 weeks of debugging and proxy configuration — roughly 40 hours of hands-on time with an AI coding agent.
- How do you build your own SalesTouch?
- Scoped to personal use: Next.js on the front, TypeScript / Express MCP Server behind it, Turso for data. Follow the 5-step build guide on this page — each step has a paste-ready prompt for an AI coding agent.
- How do you code your own SalesTouch without being an expert?
- Use an AI coding tool (Claude Code or Cursor) and work in small steps: scaffold, data model, core screens, then deploy. Realistic effort: 3-4 weeks of debugging and proxy configuration. The prompts on this page are written so the AI does the heavy lifting.
- How much does it cost to vibe code SalesTouch instead of paying?
- About ~$27 one-time to start and ~$20/mo to run, versus $49/mo for SalesTouch. Break-even: Never (risk of account ban makes paid SaaS safer).
- What stack should you use to vibe code SalesTouch?
- Next.js; TypeScript / Express MCP Server; Turso; plus Playwright for browser automation fallback, Zod for type-safe MCP tool schemas.