SKIP TO CONTENT
(CASE STUDY — 03)← ALL PROJECTS

IGFarm Studio

A Windows control panel that turns a rack of real Android phones into one point-and-click control room — driving Instagram growth across dozens of devices over ADB, with a vision AI that only steps in on screens it has never seen.

  • PHONE FARM
  • REAL-DEVICE ADB
  • VISION AI AGENT
  • LICENSED PRODUCT
  • 2026
IGFarm Studio brand composite showing the live device wall
ROLE
Design, engineering, delivery
PLATFORM
Windows desktop app + phone farm
SCALE
Dozens of physical phones in parallel
DELIVERY
Machine-bound license, auto-update
(ABOUT)

A phone farm,
run from
one screen

IGFarm Studio controls a physical farm of Android phones over ADB — real devices tapping real screens, not API calls. Human-like pacing, per-phone stagger and randomized delays keep every account behaving like a person.

The engine is deterministic first: every step replays a known or previously learned flow for free. A vision AI agent engages only when it hits a screen it has never seen, then permanently learns that branch — so the system self-heals when Instagram's UI changes and gets cheaper the longer it runs.

The operator sees every phone streamed live in the app, builds multi-step tasks from eleven chainable actions, schedules posts, and manages accounts and proxies from one dark UI.

(FEATURES)

What's inside

  • (01)

    Real-device automation

    Drives dozens of physical Android phones in parallel over ADB with an authorized key — real taps on real screens instead of risky API calls. Human-like pacing, per-phone stagger and randomized delays throughout.

    • ADB CONTROL
    • PARALLEL DEVICES
    • HUMAN PACING
  • (02)

    Deterministic engine with an AI safety net

    Every step replays a known or previously learned flow for free. A vision AI agent engages only on a screen it has never seen, then permanently learns that branch — cheap, resilient automation that self-heals when the app UI changes.

    • LEARNED FLOWS
    • VISION AGENT
    • SELF-HEALING
    • NEAR-ZERO COST
  • (03)

    Live device wall

    Every phone streamed live inside the app with click-to-focus spotlight, bundled so nothing extra needs installing. Clear gallery, keyboard reset, paste and rescan controls built in.

    • LIVE STREAMS
    • CLICK-TO-FOCUS
    • BUNDLED
  • (04)

    Automation toolset

    Eleven configurable, chainable actions: Warmup, Reel Watch, Like, Follow with rich profile filters, Unfollow with ratio and whitelist rules, Comment from a no-repeat spintax pool with optional AI, Save, Share to Story, Scrape, Story Viewer and Edit Profile.

    • 11 ACTIONS
    • CHAINABLE
    • SPINTAX + AI
    • RICH FILTERS
  • (05)

    Post scheduling

    Queue photos, reels and stories at a set time with spintax or AI captions, and per-image uniquification that produces a byte-distinct copy per account to avoid cross-account duplicate-content detection.

    • SCHEDULED POSTS
    • AI CAPTIONS
    • IMAGE UNIQUIFY
  • (06)

    Safety & scale

    Per-account daily limits, auto-incrementing warm-up ramps and weekday scheduling across every action. Unique-target distribution with atomic reservation means no two accounts ever touch the same target, and every action is verified from the UI before it counts.

    • DAILY LIMITS
    • WARM-UP RAMPS
    • ATOMIC TARGETS
    • UI VERIFICATION
  • (07)

    Operations & delivery

    Accounts spreadsheet with live health colors, task builder with run-order priority and pause/resume/stop, proxy manager, run and AI-activity logs, in-app docs, and a demo mode. Ships as a per-user installer with machine-bound licensing and silent auto-updates.

    • HEALTH COLORS
    • PROXY MANAGER
    • IN-APP DOCS
    • AUTO-UPDATE
(SCREENS)

Inside the product

IGFarm Studio device wall streaming every connected Android phone live
LIVE DEVICE WALL — every phone streamed live in-app with click-to-focus spotlight — bundled so nothing extra needs installing
IGFarm Studio task builder listing eleven chainable actions with targets and filters
TASK BUILDER — eleven chainable actions — Warmup, Reel Watch, Like, Follow, Unfollow, Comment, Save, Share to Story, Scrape, Story Viewer, Edit Profile — each with its own targets, filters and delays
IGFarm Studio task runner with live per-account progress and pause and stop controls
TASK RUNNER — run tasks in parallel with pause, resume, restart and stop — live per-account progress without leaving the screen
IGFarm Studio accounts spreadsheet with colour-coded account health states
ACCOUNTS SPREADSHEET — live health colors for logged-in, banned, checkpoint, locked and needs-phone states, with bulk login, sync, logout and grouping
IGFarm Studio post scheduler queueing photos, reels and stories with AI captions
POST SCHEDULING — queue photos, reels and stories with spintax or AI captions and per-image uniquification to avoid cross-account duplicate detection
IGFarm Studio proxy manager assigning a proxy to each connected device
PROXY MANAGER — per-device proxy assignment pushed over the current Wi-Fi, with round-robin auto-assign and a local auth-proxy so phones are never prompted
IGFarm Studio configuration screen with ADB settings, AI model selection and integrations
CONFIGURATION — ADB path and device count, AI model selection with learned-flow replay, deterministic captcha and phone-verification integrations, and per-account image pools
(UNDER THE HOOD)

Technical highlights

  1. (01)

    Python 3.12 core with a modern dark desktop UI (CustomTkinter), packaged with PyInstaller and an Inno Setup installer.

  2. (02)

    Device layer over ADB and scrcpy, bundled with the app so operators install nothing extra.

  3. (03)

    Learning engine: screen-fingerprint flow memory that records during a run, commits only on success, and accumulates every branch over time.

  4. (04)

    Vision AI agent (Claude) invoked only on unrecognized screens, keeping running cost a fraction of an AI-every-step approach.

  5. (05)

    Deterministic captcha and phone verification via 2Captcha and 5sim, plus IMAP/POP3 email-code retrieval for email logins.

  6. (06)

    Per-user install with no admin rights, machine-bound licensing, and silent self-updates from GitHub Releases.

LEARNING-ENGINE INTERNALS ARE PROPRIETARY.

(WANT SOMETHING LIKE THIS?)Hardware, software,
one system — let's talk ↗

Device automation, AI agents with learned flows, licensing and delivery — built as a product you own and can run unattended.