# Extella — Full Content Reference

> AI execution platform. Turns natural language into reusable automations that run locally on your device.

## What Extella Is

Extella is not a chatbot. It executes tasks: reads files, generates documents, calls APIs, sends messages, runs Python code, and saves results. Every task creates a reusable Expert saved to your personal library. The system grows more capable with every use.

Category: AI Execution Platform
Tagline: Intelligence that compounds
Founded: 2024
Company: Chariot Technologies Lab

## How It Works

Extella runs on a client-server architecture:

- **Server**: AI brain (language model, knowledge base, Expert and agent management)
- **Listener**: background process on your local device that receives tasks and runs them locally

When you describe a task, Extella creates an Expert (executable Python module), saves it to your library, runs it on your device via Listener, and returns a real result: a file, a message, processed data. Not text in a chat window.

## Core Concepts

### Experts
Atomic executable Python modules. Created on demand from natural language. Saved permanently to your personal library. Run locally via Listener with full filesystem access. Every Expert has: name (snake_case), description (used for semantic search), code (Python), parameters (with defaults).

### Concepts
Long-term semantic memory. Each Concept is stored with a vector embedding (OpenAI text-embedding-3-small, 1536 dimensions, cosine similarity). The agent finds relevant knowledge by meaning, not just keywords. Automatically accumulates technical patterns, API learnings, error resolutions.

### Rules
Behavioral instructions loaded at every session automatically. Persist between conversations. Examples: "always ask before deleting files", "save output to ~/Documents/Extella/", "describe the plan before executing multi-step tasks".

### KV Store
Encrypted key-value storage for credentials and configuration. All values encrypted with user PIN (AES-256-GCM). The agent never sees decrypted values outside the execution context. Semantic search across keys and descriptions.

### Listener
Background process running on your device (macOS, Windows, Linux). Receives task execution requests from the AI agent. Runs Experts locally with full OS access. No files leave your device unless you explicitly send them.

### Target / Device ID
UUID identifying a specific device. Allows running Experts on remote devices (e.g., run a task on your home Mac from a laptop at a cafe).

## What Extella Can Do

- Read and write local files (PDF, Excel, Word, CSV, images, code, any format)
- Call external APIs (Telegram, OpenAI, Google, Slack, GitHub, custom endpoints)
- Run CLI tools (ffmpeg, git, curl, any shell command via shell_runner CSPL mode)
- Process data with any Python library
- Schedule and automate multi-step workflows
- Orchestrate multiple AI agents in parallel
- Generate documents, reports, images, presentations
- Monitor services and trigger actions based on conditions
- Work with databases and data pipelines
- Access and control local applications

## Technical Architecture

### CSPL (Container Specific Programming Language)
Defines execution mode for each Expert:
- **fython** (default): standard Python function, synchronous, returns dict
- **nohup**: detached subprocess, runs as script, returns PID immediately. For long-running orchestrations.
- **parallel_task**: auto-registered in Task Registry, returns UUID. For parallel workers.
- **shell_runner**: executes shell commands, body is shell script.
- **wait_tasks**: synchronous barrier, waits for parallel workers to complete.

### REST API
Base URL: https://prod.extella.ai
Authentication: X-Auth-Token header
Key endpoints: /api/agent/run, /api/expert/run, /api/expert/save, /api/concept/add, /api/kv/set, /api/token/generate

### Agents and Profiles (Pro plan)
Each agent has its own Rules, Concepts, system prompt, model selection, and tool set. Profiles group agents. One user can have specialized agents: coding, research, customer support, each with different underlying models.

### BYOK (Bring Your Own Keys)
Connect your own LLM providers: OpenAI, Anthropic, Gemini, Azure OpenAI, OpenRouter, Ollama (local), any OpenAI-compatible endpoint.

## Platform Details

- **OS support**: macOS, Windows, Linux
- **Default model**: Claude (Anthropic)
- **Execution**: local via Listener, files never leave device unless explicitly sent
- **Pricing**: free to install
- **Security**: AES-256-GCM encryption for credentials, server-side user isolation by user_id
- **API**: full programmatic access via REST
- **Local models**: Ollama integration out of the box

## Pages

- Homepage: https://extella.ai
- Getting Started Guide: https://extella.ai/getting-started
- FAQ: https://extella.ai/faq
- Terms of Use: https://extella.ai/terms
- Privacy Policy: https://extella.ai/privacy
- API Documentation: https://docs.extella.ai

## Social / External

- Twitter/X: https://x.com/ExtellaAI
- GitHub: https://github.com/extella
- LinkedIn: https://www.linkedin.com/company/extella
- ProductHunt: https://www.producthunt.com/products/extella
- Crunchbase: https://www.crunchbase.com/organization/extella

## Getting Started (Summary)

1. Download Extella Desktop from extella.ai
2. Install and sign in — Listener starts automatically
3. Generate an API token (type "generate an API token" in chat)
4. Type your first task in plain English
5. Extella creates an Expert, runs it, returns a real result
6. The Expert is saved to your library permanently
7. Add your first Rule: behavioral instruction for every session
8. Concepts accumulate automatically with every task

## Compounding Effect

Day 1: one Expert. Day 30: 15 Experts, a personal toolbox. Day 90: 50+ Experts, institutional memory, a system that knows your context. Unlike chatbots that reset every session, Extella gets more capable with every task you run.

## Landing Pages (updated July 2026)

- Main page: https://extella.ai/ — positioning, six agent roles, architecture, downloads
- Product tour: https://extella.ai/product — how Extella works, catalog of capabilities, downloads
- API reference: https://extella.ai/api — 48 REST endpoints, 106 schemas
- User guide: https://extella.ai/guide.html — install, memory (KV/Concepts/Rules), Experts, agents & teams, local models, parallel execution, CSPL, REST API
- FAQ: https://extella.ai/faq
- Blog — AI 101 course: https://extella.ai/blog — 7 chapters + final assessment (what is an LLM; the chatbot you already know; from talking to doing; why one agent isn't enough; what an agentic platform is; the four side by side; what is the right AI for you)
- Russian mirror: https://extella.ai/ru/ — full localization including blog and guide

Downloads (v1.2.5, free install, 5,000 credits included):
- macOS (Apple Silicon): https://extella-install-files.fra1.digitaloceanspaces.com/extella-desktop/macos/Extella-1.2.5-arm64.dmg
- Windows: https://extella-install-files.fra1.digitaloceanspaces.com/extella-desktop/windows/Extella%20Setup%201.2.5.exe
- Linux Desktop (x86_64): https://extella-install-files.fra1.digitaloceanspaces.com/extella-desktop/linux/Extella-1.2.5-x86_64.AppImage
- Linux Desktop (aarch64): https://extella-install-files.fra1.digitaloceanspaces.com/extella-desktop/linux/Extella-1.2.5-arm64.AppImage
- Linux CLI Worker (x86_64): https://extella-install-files.fra1.digitaloceanspaces.com/extella-worker/linux/Extella-Worker-1.1.1-x86_64.run
- Linux CLI Worker (aarch64): https://extella-install-files.fra1.digitaloceanspaces.com/extella-worker/linux/Extella-Worker-1.1.1-aarch64.run

---
Generated: 2026-05-28, extended: 2026-07-10
Source: extella.ai