diff --git a/docs/source/augmentation_tutorials/intent_description.rst b/docs/source/augmentation_tutorials/intent_description.rst
index 3227f97f4..a56194c2d 100644
--- a/docs/source/augmentation_tutorials/intent_description.rst
+++ b/docs/source/augmentation_tutorials/intent_description.rst
@@ -33,7 +33,7 @@ Ensure you have the necessary dependencies installed:
 
 .. code-block:: bash
 
-    pip install autointent openai
+    pip install "autointent[openai]"
 
 Usage
 -----
diff --git a/docs/source/index.rst b/docs/source/index.rst
index a7e921322..9072da36e 100644
--- a/docs/source/index.rst
+++ b/docs/source/index.rst
@@ -47,6 +47,9 @@ Getting Started
 :doc:`🚀 Quickstart <quickstart>`
    Jump right in! Install AutoIntent and build your first text classifier in minutes. Perfect for users who want to get up and running quickly with practical examples.
 
+:doc:`📦 Installation & optional extras <installation>`
+   Choose a minimal core install or add optional dependency groups (transformers, OpenAI, CatBoost, servers, tracking, and more). Includes recommended bundles and resolver notes.
+
 :doc:`📚 Key Concepts <concepts>`
    Essential terminology and concepts used throughout AutoIntent. Understanding these will help you navigate the documentation and make the most of the library's features.
 
@@ -73,6 +76,7 @@ Reference
    :maxdepth: 1
 
    quickstart
+   installation
    concepts
    user_guides
    learn/index
diff --git a/docs/source/installation.rst b/docs/source/installation.rst
new file mode 100644
index 000000000..b81d39d35
--- /dev/null
+++ b/docs/source/installation.rst
@@ -0,0 +1,128 @@
+.. _installation:
+
+Installation and optional extras
+================================
+
+AutoIntent supports **Python** ``>=3.10,<3.15`` (see ``requires-python`` in the package metadata).
+
+Since **v0.3.0**, ``pip install autointent`` installs only the **core** runtime: PyTorch, scikit-learn, Optuna, vector search, datasets, and other always-on dependencies. Heavier or integration-specific libraries are shipped as **optional extras** (see ``[project.optional-dependencies]`` in ``pyproject.toml``).
+
+Minimal install
+---------------
+
+For the smallest footprint (core AutoML and modules that do not require optional stacks):
+
+.. code-block:: bash
+
+   pip install autointent
+
+If a preset or module needs an extra, install it with ``pip install "autointent[<extra>]"`` or combine several names in the brackets (comma-separated).
+
+Optional extras matrix
+----------------------
+
+The table below lists each published extra, what it is typically used for in AutoIntent, and the matching ``pip`` command.
+
+.. list-table::
+   :header-rows: 1
+   :widths: 18 52 30
+
+   * - Extra
+     - What it enables
+     - Install command
+   * - ``sentence-transformers``
+     - SentenceTransformer embedders and pipelines that rely on them (for example many ``classic-*`` and embedding-based scorers).
+     - ``pip install "autointent[sentence-transformers]"``
+   * - ``catboost``
+     - ``CatBoostScorer`` and CatBoost-based tuning paths (see the modules API).
+     - ``pip install "autointent[catboost]"``
+   * - ``transformers``
+     - Hugging Face ``transformers`` models used by transformer presets and modules.
+     - ``pip install "autointent[transformers]"``
+   * - ``peft``
+     - Parameter-efficient fine-tuning (LoRA and similar) used together with transformer presets.
+     - ``pip install "autointent[peft]"``
+   * - ``openai``
+     - OpenAI API clients for OpenAI-backed embedders and the ``zero-shot-llm`` style workflows.
+     - ``pip install "autointent[openai]"``
+   * - ``dspy``
+     - DSPy-based augmentation and related generation utilities.
+     - ``pip install "autointent[dspy]"``
+   * - ``wandb``
+     - Weights & Biases experiment logging integration.
+     - ``pip install "autointent[wandb]"``
+   * - ``codecarbon``
+     - CodeCarbon energy / emissions tracking during runs.
+     - ``pip install "autointent[codecarbon]"``
+   * - ``fastapi``
+     - HTTP serving stack (FastAPI, Uvicorn, settings) for the AutoIntent server mode.
+     - ``pip install "autointent[fastapi]"``
+   * - ``fastmcp``
+     - FastMCP-based MCP server integration.
+     - ``pip install "autointent[fastmcp]"``
+   * - ``opensearch``
+     - OpenSearch client for OpenSearch-backed vector / retrieval integrations.
+     - ``pip install "autointent[opensearch]"``
+   * - ``vllm``
+     - vLLM as an optional high-throughput inference backend where supported.
+     - ``pip install "autointent[vllm]"``
+
+Recommended bundles
+-------------------
+
+Combine extras in one install by listing them inside the brackets (no spaces inside the list).
+
+**Classic embedding + gradient boosting pipeline** (typical ``classic-light`` / ``classic-heavy`` style workflows with SentenceTransformers and CatBoost):
+
+.. code-block:: bash
+
+   pip install "autointent[sentence-transformers,catboost]"
+
+**Transformer presets** (``transformers-*`` presets and fine-tuning): install both Hugging Face transformers and PEFT:
+
+.. code-block:: bash
+
+   pip install "autointent[transformers,peft]"
+
+**Zero-shot LLM preset** (OpenAI API):
+
+.. code-block:: bash
+
+   pip install "autointent[openai]"
+
+**HTTP API server**:
+
+.. code-block:: bash
+
+   pip install "autointent[fastapi]"
+
+**MCP server** (FastMCP):
+
+.. code-block:: bash
+
+   pip install "autointent[fastmcp]"
+
+**Experiment tracking** (pick what you use):
+
+.. code-block:: bash
+
+   pip install "autointent[wandb]"
+   pip install "autointent[codecarbon]"
+
+``codecarbon`` vs ``fastmcp`` (uv)
+----------------------------------
+
+When you use **uv** to resolve the project, ``codecarbon`` and ``fastmcp`` are declared as **conflicting extras** in ``pyproject.toml`` (``[tool.uv]`` ``conflicts``). A single environment should not select both at once; pick the extra that matches your goal (emissions tracking **or** FastMCP), or use separate environments if you genuinely need both stacks isolated.
+
+Development install from Git
+----------------------------
+
+Clone the upstream repository and use **uv** with the project ``Makefile`` (see ``CONTRIBUTING.md`` in the repository root for details):
+
+.. code-block:: bash
+
+   git clone https://github.com/deeppavlov/AutoIntent.git
+   cd AutoIntent
+   make install
+
+This installs the full contributor dependency set (tests, typing, docs, and so on), not only the minimal PyPI ``autointent`` wheel dependencies.
diff --git a/docs/source/quickstart.rst b/docs/source/quickstart.rst
index 79cfc0227..be0f632b6 100644
--- a/docs/source/quickstart.rst
+++ b/docs/source/quickstart.rst
@@ -19,34 +19,45 @@ Key Features
 Installation
 ------------
 
+AutoIntent supports **Python** ``>=3.10,<3.15``. The full optional-extra matrix, recommended bundles, and notes on dependency conflicts are on :doc:`installation`.
+
 Basic Installation
 ..................
 
-AutoIntent is compatible with Python 3.10+. For core functionality:
+The PyPI package installs **core** dependencies only. For a minimal install:
 
 .. code-block:: bash
 
     pip install autointent
 
+Add stacks you need with extras, for example:
+
+.. code-block:: bash
+
+    pip install "autointent[sentence-transformers,catboost]"
+    pip install "autointent[openai]"
+    pip install "autointent[transformers,peft]"
+
 With Experiment Tracking
 ........................
 
-To include experiment tracking capabilities:
+Optional integrations are installed by name, for example:
 
 .. code-block:: bash
 
-    pip install autointent[wandb,codecarbon]
+    pip install "autointent[wandb]"
+    pip install "autointent[codecarbon]"
 
 Development Installation
 ........................
 
-To install the latest development version:
+To work from a Git checkout, use **uv** and the project ``Makefile`` (see ``CONTRIBUTING.md``):
 
 .. code-block:: bash
 
-    git clone https://github.com/voorhs/AutoIntent.git
+    git clone https://github.com/deeppavlov/AutoIntent.git
     cd AutoIntent
-    pip install .
+    make install
 
 Quick Example
 -------------