About Extella
Overview
Extella is a desktop AI workspace that brings models, GitHub repositories, Hugging Face libraries, plugins and MCP connectors into one environment. You describe a task in plain language, and the system installs what it needs, works with real files, data and APIs, and returns a finished artifact such as a file, report or completed action instead of advice in a chat.
Successful runs become Experts — reusable tools stored in a personal library that run again with new inputs, get improved over time or pass to a team. Extella ships for macOS, Windows and Linux, and exposes a REST API for builders who wire it into their own pipelines and products.
Key Benefits
- Turns a plain-language task description into a working tool saved in your library.
- Installs GitHub repos, Hugging Face models, datasets and libraries from a catalog with no manual setup.
- Runs each Expert as an isolated module, so code, model and library dependencies never conflict.
- Keeps rules and concepts persistent, so agents hold your preferences and process knowledge across sessions.
- Encrypts API keys and tokens on your device and grants each task only the access it requires.
- Gives every agent its own provider, temperature and reasoning mode, including local models through Ollama.
How It Works
You state the result you need in plain words, with no prompt engineering, diagrams or manual setup. Extella executes the task against real files, data, APIs and tools, and a successful solution is saved to your library as a tool you can rerun, refine, hand to the team or build new processes on top of.
Use Cases
- Researchers gather sources, remove duplicates and prepare briefs and reports that follow a preferred structure.
- Ops and RevOps teams connect CRMs, spreadsheets, drives and APIs into one process and drop the manual hand-offs between tools.
- Marketing and brand teams move from plan to finished publication while Extella holds their rules, references and style.
- Developers spin up internal tools, scripts and integrations from a single instruction, then rerun them from the library.
- Analysts turn spreadsheets, exports and dashboards into finished reports, files and visualizations.
- Individuals keep notes, decisions and recurring routines in a personal system that retains context across sessions.
Why Choose This Product
Extella separates a task's intent from its mechanics through CSPL, its own container-specific runtime: the model writes a short declarative description and a deterministic handler turns it into code, which the page measures at up to 15 times fewer tokens on repeated tasks. Isolation runs through the whole design — every agent carries its own system prompt, model, tools and memory, split across Global, Team and Agent context levels, so parallel agents never collide.
Extella Pros & Cons
- Runs experts locally on your device with isolated execution modules
- REST API with 48 endpoints plus OpenAI-compatible mode
- Each agent can use a different model provider or a local model via Ollama
- API keys encrypted on the device and scoped per task
- Start free with 5,000 credits and no card required
Key Features
Catalog Installs
Add apps, models, skills, tools, knowledge bases, agents and GitHub repos from the catalog, and Extella installs, connects and saves the setup.
Reusable Experts
Each successful solution becomes an Expert in your personal library that runs again later with different inputs.
Concepts Memory
Concepts store processes, documents, clients, decisions and preferences, and pull the right context by meaning rather than chat history.
Behavioral Rules
Rules define an agent's working style, constraints, roles and safety preferences, and hold in future sessions.
Multi-Agent Teams
A master agent splits a request across specialist agents for analyst, developer, editor, researcher and data roles, then assembles the result.
Per-Agent Models
Every agent sets its own provider, temperature and reasoning mode across OpenAI, Anthropic, Google, xAI, Qwen, MiniMax, custom endpoints or a local model via Ollama.
Encrypted Key Store
API keys, tokens and configuration sit in a kv store encrypted on your device and stay out of code, chat and model memory.
CSPL Runtime
CSPL separates a task's intent from execution, so the model declares the what and a deterministic handler generates the exact code.
REST API
A 48-endpoint REST API with token auth runs experts, manages memory and agents, handles tasks on remote devices and includes OpenAI-compatible mode.
Isolated Execution
Experts run as isolated modules for code, models, libraries and external tools, so dependency conflicts stay out of the way.
Frequently asked questions about Extella
What is Extella?
Extella is a customizable AI superagent that performs tasks using a system of Experts, Concepts and Rules. It automates repetitive work such as sending reports, processing files and fetching data, builds reusable tools you can run anytime, remembers your preferences and learns from your workflows, and connects your services like Telegram, email, APIs and files.
Do I need to know how to code to use Extella?
No. You describe what you want in plain language, Extella creates an Expert as a coded script and saves it, and you run that Expert anytime with different inputs. Experts chain together for complex workflows, and you never write the code yourself.
What exactly are Experts?
An Expert is a reusable capability Extella creates to perform a specific task. They range from simple single-purpose tools like convert_pdf_to_text, to complex multi-step pipelines such as decompile_binary_to_pseudocode, to nested orchestrators that coordinate other Experts, to integration modules built around Telegram, Notion, PyTorch, ffmpeg or PostgreSQL.
What are Concepts in Extella?
Concepts are technical knowledge and solutions Extella learns and retains over time, so it gets smarter the more you use it. Use them for API patterns and endpoints, error-handling solutions, library recommendations and technical workarounds, such as retrying 429 errors with exponential backoff.
What are Rules used for?
Rules define how Extella should behave, covering communication style, safety behaviors and workflow habits. Typical uses include tone, format and length preferences, confirmations and warnings, and how tasks get handled — for example, always confirming before deleting anything or keeping responses short.
How does Extella execute a task?
You describe the result you need in plain words. Extella works with real files, data, APIs and tools, and the output is the finished task — a file, a report, an action or a working process — rather than advice in a chat. A successful solution is then saved to your library as a tool.