Pllan — Autonomous Desktop & Operations Agent

The fourth pillar of the NxAI suite — a persistent, learning agent that transforms any computer into an intelligent workspace.

What is Pllan?

Pllan is an autonomous operations agent that:

• 🧠 Learns continuously across sessions via 5-layer persistent memory
• 🔐 Operates safely with L0-L4 graduated authority and sandbox isolation
• 🤝 Orchestrates teams with cross-agent delegation and shared workflows
• 🌐 Integrates deeply with the NxAI ecosystem (Kapllan, Llana, Ofelia)
• 🔌 Evolves skills through self-improvement and A/B testing
• 🖥️ Runs everywhere with cross-platform support (macOS, Linux, Windows)

Unlike single-session assistants, Pllan remembers everything, gets smarter over time, and compounds its effectiveness through continuous learning.

Quick Start

Prerequisites

• Rust 1.75+ — Install from rustup.rs
• Node.js 22 LTS — Install from nodejs.org
• pnpm 9+ — Install with npm install -g pnpm
• Docker Desktop — Install from docker.com

One-Command Setup

git clone https://github.com/nxai/pllan.git
cd pllan
make dev-setup

This will:

1. ✅ Check prerequisites
2. ✅ Start infrastructure (PostgreSQL, Qdrant, Redis, Kafka, MinIO)
3. ✅ Install dependencies
4. ✅ Build the workspace
5. ✅ Run database migrations
6. ✅ Start the daemon

Manual Setup

If you prefer step-by-step:

# 1. Start infrastructure
./scripts/pllan_infra_up.sh

# 2. Install Rust dependencies
cargo build --workspace

# 3. Install JavaScript dependencies (if packages exist)
cd packages && pnpm install && cd ..

# 4. Run database migrations
cargo sqlx migrate run

# 5. Start daemon
cargo run --bin plland

Architecture

Pllan consists of 8 core crates and a multi-agent system:

Core Crates

| Crate | Purpose | |-------|---------| | pllan-core | Main daemon orchestrator, task state machine, authority system | | pllan-memory | 5-layer persistent memory (episodic, semantic, procedural, working, org) | | pllan-sandbox | Cross-platform execution isolation (Firecracker, VZ, Hyper-V) | | pllan-gateway | Central LLM Gateway with PII scanning, routing, caching | | pllan-bridge | IPC (gRPC/WebSocket) + NxAI Bus (Kafka) integration | | pllan-sentinel | Background file watcher and event triggers | | pllan-cli | Command-line interface (pllan binary) | | pllan-ops | Self-improvement subsystem (A/B testing, skill evolution) |

Multi-Agent System

| Agent | Role | |-------|------| | Planner | DAG task decomposition with dependency resolution | | Executor | Direct execution (file ops, shell, documents) | | Researcher | Web search, RAG, citation tracking | | Communicator | Email, Slack, calendar management | | Data Analyst | SQL, pandas, visualization | | Ops | Cloud infrastructure, CI/CD, monitoring | | Meta (Pllan-Core) | Supervisory coordinator |

Development

Common Commands

# Infrastructure
make infra-up        # Start all services
make infra-down      # Stop all services
make infra-status    # Check service health

# Build & Test
make build           # Build Rust + TypeScript
make test            # Run all tests
make fmt             # Format code
make lint            # Lint code
make clean           # Clean build artifacts

# Database
make migrate         # Run migrations

# Run
make run             # Start plland daemon

Project Structure

pllan/
├── crates/               # Rust workspace
│   ├── pllan-core/       # Main daemon
│   ├── pllan-memory/     # Memory system
│   ├── pllan-sandbox/    # Execution isolation
│   ├── pllan-gateway/    # LLM Gateway
│   ├── pllan-bridge/     # IPC + NxAI Bus
│   ├── pllan-sentinel/   # File watcher
│   ├── pllan-cli/        # CLI
│   └── pllan-ops/        # Self-improvement
├── packages/             # TypeScript workspace (future: Tauri UI)
├── contracts/            # API specs (OpenAPI, Protobuf)
├── migrations/           # Database migrations
├── scripts/              # Infrastructure scripts
├── config/               # Configuration templates
└── tests/                # Integration tests

Documentation

• Implementation Roadmap — 36-week plan to Core GA
• Full Specification — Complete product vision & architecture
• v2.1.1 Errata — Recent corrections and clarifications

Current Status: Phase 0 — Bootstrap ✅

Progress:

• ✅ Infrastructure scripts (pllan_infra_up.sh, pllan_infra_down.sh, pllan_infra_status.sh)
• ✅ Monorepo structure (Rust workspace + packages)
• ✅ Docker Compose (PostgreSQL, Qdrant, Redis, Kafka, Schema Registry, MinIO)
• ✅ All crates compile with stubs
• ✅ Makefile with common commands
• 🚧 Database migrations (Week 1 Day 5)
• 🚧 Minimal working daemon (Week 2)

Next: Phase 1 — Core Runtime (Weeks 3-12)

See IMPLEMENTATION_ROADMAP.md for detailed timeline.

NxAI Product Suite

Pllan completes the NxAI suite alongside:

| Product | Domain | Status | |---------|--------|--------| | Ofelia | ML Training Lab | ✅ Production | | Kapllan | Training Supervisor | ✅ Production | | Llana | Coding Agent | ✅ Production | | Pllan | Desktop & Operations | 🔨 Building |

NxAI Bus connects all four products via Kafka event backbone.

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE)
• MIT license (LICENSE-MIT)

at your option.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md first.

Built with ❤️ by the NxAI team