Pilot — Turn Instagram DMs Into Qualified Leads

---

AI-powered Instagram sales system for lead management and revenue operations - built for creators, entrepreneurs, and social teams who want conversions, not brittle flow-builder bots.

Supported by teams building alongside us.

Free Greptile license through their open source program, with setup support.

Startup Program support: $1,000 in Pro credits for 6 months, product team access, early feature access, and collaboration opportunities.

Table of Contents

• About
• Pilot vs ManyChat
• Architecture
• Monorepo Setup
• Turborepo Commands
• Roadmap
• Documentation
• Contributing
• License
• Authors
• Acknowledgements

About

Pilot is an Instagram automation and deal management platform designed to help creators, entrepreneurs, small businesses, and social media managers turn Instagram conversations into qualified pipeline.

Pilot is positioned as a sales system, not a bot:

• Intent-aware conversation handling over rigid decision trees
• CRM depth (lead score, stage, sentiment, tags, notes, follow-up context)
• Human Response Needed (HRN) guardrails for risky threads
• Open source and self-hostable for data and platform control

The AI-powered Sidekick supports lead qualification, response generation, and next-step recommendations while preserving human oversight where needed.

Pilot vs ManyChat

Common ManyChat pain points

• Unreliable follow-to-DM behavior and delayed sends
• Shadow-ban/account-risk anxiety from aggressive automation patterns
• Per-contact pricing shock as contact count scales
• Flow-builder complexity for simple use cases

Head-to-head comparison

Dimension	ManyChat	Pilot	Advantage
AI Intelligence	Basic rule-based flows + AI add-on ($)	Native AI-first with intent detection, sentiment analysis, lead scoring	Pilot
Conversation Handling	One reply per trigger pattern	Sidekick tracks conversation context and tone	Pilot
Lead Management	Basic contact list	Full contact CRM with scoring, tags, stages, sentiment	Pilot
Human Handoff	Mostly manual	HRN system with classifier + risk heuristics	Pilot
Pricing Model	Per-contact scaling	Open-source, self-hostable, predictable model direction	Pilot
Shadow Ban Protection	Manual mitigation	Guardrails-first automation design	Pilot
Setup Complexity	Visual flow builder	Trigger + AI handling	Pilot
Open Source	Closed, vendor lock-in	Fully open-source and forkable	Pilot
Multi-Step Sequences	Supported	Not yet built	ManyChat
Visual Flow Builder	Supported	Not yet built	ManyChat
Multi-Channel	Instagram + FB + WhatsApp + SMS	Instagram-first today	ManyChat
Integrations	50+ integrations	Early-stage integration surface	ManyChat

Full competitive breakdown: competitive-analysis.md

Architecture

flowchart LR
U["User: Web App"] --> A["Auth & Access"]
A --> O["Onboarding"]
O --> W["Workspace"]
W --> S1["Sidekick AI Chat"] & S2["Contacts CRM"] & S3["Automations"] & S4["Settings & Instagram Connection"] & S5["Billing & Plans"]
IG["Instagram Platform"] --> WH["Webhook Ingestion"]
WH --> E["Decision Engine"]
E -- Trigger matched --> S3
E -- No trigger / AI reply --> S1
E -- Risky thread --> HRN["Human Response Needed Queue"]
E --> OUT["Instagram Reply APIs"]
S1 --> AI["Gemini AI Layer"]
S2 --> DB[("Postgres + Drizzle + RLS")]
S3 --> DB
S4 --> DB
S5 --> POLAR["Polar Billing"]
AI --> DB
JOBS["Background Jobs: Inngest"] --> S2 & IG & DB
OBS["Monitoring & Analytics"] --- W & WH & JOBS

Loading

Monorepo Setup

This repository uses pnpm workspaces + Turborepo.

Structure

• apps/app -> Main product app
• apps/web -> Marketing website and public waitlist
• packages/ui -> Shared shadcn UI + Tailwind layer
• packages/db -> Shared Drizzle schema/client/migrations
• packages/instagram -> Shared Instagram Graph API client, token refresh, retries, webhook helpers
• packages/core -> Shared product brain (automation logic, sidekick logic, contact workflows)
• packages/config -> Shared eslint/postcss/tsconfig
• packages/types -> Shared domain types

Install

pnpm install

App-specific docs

• Main product app: apps/app/README.md
• Marketing app: apps/web/README.md

Turborepo Commands

Run from repo root:

# run both apps/dev tasks via turbo graph
pnpm dev
# targeted dev
pnpm dev:app
pnpm dev:web
# run builds with turbo graph + cache
pnpm build
# targeted builds
pnpm build:app
pnpm build:web
# changed-only graph build
pnpm build:affected
# type checks
pnpm check-types

Turbo local cache is stored in .turbo/.

For remote cache in CI/team machines, set:

TURBO_TEAM=...
TURBO_TOKEN=...

Roadmap

Roadmap is tracked in ROADMAP.md.

Documentation

• Product strategy and comparison: competitive-analysis.md
• Engineering and contribution workflows: CONTRIBUTING.md

Contributing

All code setup, development workflow, testing, and deployment notes live in CONTRIBUTING.md.

License

This project is licensed under the GNU Affero General Public License v3.0.

See LICENSE for the full text.

Authors

• ArjunCodess - Project development and maintenance

Acknowledgements

• Instagram for providing the API that powers social integrations
• Vercel for the deployment platform and AI SDK
• Neon for serverless PostgreSQL hosting
• shadcn/ui and Radix UI for accessible UI components
• Drizzle Team for the ORM toolkit
• Open Source Community for the ecosystem of libraries and tools

---

Pilot - Transforming Instagram interactions into business opportunities