diff --git a/TOOLS.md b/TOOLS.md
index fb0861cf4..9084b7716 100644
--- a/TOOLS.md
+++ b/TOOLS.md
@@ -211,7 +211,9 @@ EXAMPLES:
 
 Creates a root component configuration using the specified name, component ID, configuration JSON, and description.
 Not for SQL transformations (`keboola.snowflake-transformation` / `keboola.google-bigquery-transformation`),
-data apps (`keboola.data-apps`) or flows — use the dedicated tools (see WHEN NOT TO USE).
+data apps (`keboola.data-apps`) or flows — use the dedicated tools (see WHEN NOT TO USE). This IS the tool for
+Python (`keboola.python-transformation-v2`), R (`keboola.r-transformation-v2`) and DuckDB
+(`keboola.duckdb-transformation`) transformations.
 
 BEFORE CALLING - REQUIRED STEPS:
 1. Call `get_components([component_id])` to retrieve the component's `configuration_schema`.
@@ -397,6 +399,8 @@ USAGE:
 - This is THE tool for creating `keboola.snowflake-transformation` and `keboola.google-bigquery-transformation`
   components (do NOT use `create_config` for these); the transformation ID is derived automatically from the
   workspace SQL dialect.
+- Snowflake/BigQuery only. For Python, R, or DuckDB transformations, use `create_config` with the appropriate
+  `component_id` instead — this tool cannot create them.
 
 EXAMPLES:
 - user_input: `Can you create a new transformation out of this sql query?`
@@ -1379,7 +1383,7 @@ the operations you want to perform. All other fields will remain unchanged.
 Use this for modifying SQL transformations created with create_sql_transformation.
 
 WHEN TO USE:
-- SQL transformations only (Snowflake/BigQuery); use update_config for Python/R transformations
+- SQL transformations only (Snowflake/BigQuery); use update_config for Python/R/DuckDB transformations
 - Modifying SQL queries in transformation (add/edit/remove SQL statements)
 - Updating transformation block or code block names
 - Changing input/output table mappings for the transformation
diff --git a/feature_spec/duckdb_transformation_support/RFC.md b/feature_spec/duckdb_transformation_support/RFC.md
new file mode 100644
index 000000000..816ca7172
--- /dev/null
+++ b/feature_spec/duckdb_transformation_support/RFC.md
@@ -0,0 +1,99 @@
+# RFC: DuckDB transformation support
+
+Linear: [AI-3112](https://linear.app/keboola/issue/AI-3112/kai-supporting-creation-of-duckdb-transformations)
+
+## Problem
+
+Keboola added `keboola.duckdb-transformation` as a third transformation backend alongside
+Snowflake/BigQuery SQL transformations and Python/R transformations. The MCP server has no
+awareness of it:
+
+- `create_sql_transformation` / `update_sql_transformation` are hard-anchored to whichever
+  backend the project's **workspace** uses. The component ID is derived via
+  `WorkspaceManager.get_sql_dialect()` → `get_sql_transformation_id_from_sql_dialect()`
+  (`src/keboola_mcp_server/tools/components/utils.py:281-299`), which only maps `'snowflake'`
+  and `'bigquery'` and raises `ValueError` otherwise. Workspaces themselves
+  (`src/keboola_mcp_server/workspace.py:714-813`) only support those two backends. DuckDB has
+  no cloud workspace and was never going to plug into this path.
+- `create_config`'s docstring (`tools.py:1121-1138`) tells the caller to use `create_sql_transformation`
+  for Snowflake/BigQuery and says nothing about Python, R, or DuckDB — callers currently infer
+  by precedent that non-SQL transformation types go through `create_config`.
+- `update_sql_transformation_internal`'s 404 error message (`tools.py:996-1000`) already hints
+  "if this is a Python or R transformation, use `update_config`..." — DuckDB hits the same 404
+  and needs the same hint, or the guidance is incomplete/misleading.
+- `FOLDER_SUPPORTING_COMPONENT_IDS` (`utils.py:74`) only contains the Python/R component IDs, so
+  `update_config`'s folder metadata handling (`tools.py:1551-1555`) silently skips folder
+  management for DuckDB configs even though the UI supports organizing DuckDB transformations
+  into folders (per the DuckDB transformation docs).
+
+Visible symptom: asking Kai to create or update a DuckDB transformation either fails outright
+(if routed through the SQL-transformation tools, which will raise on non-snowflake/bigquery
+workspaces) or succeeds via `create_config` by accident, without folder support and without any
+tool guidance steering the model there in the first place.
+
+## Required Behavior
+
+| Scenario | Required behavior |
+| --- | --- |
+| User asks to create/update a `keboola.duckdb-transformation` config | Routed through `create_config` / `update_config`, same as Python/R today — **not** through `create_sql_transformation`/`update_sql_transformation`. |
+| User asks for a transformation but doesn't specify backend, and the workspace default isn't what's implied | Kai should recognize DuckDB and Python as the two non-workspace-backed alternatives and ask the user which one, rather than silently guessing. (Prompt-level guidance change, not new code — see Scope.) |
+| DuckDB config created/updated via `update_config` | Folder metadata (`folder` param) is applied, same as Python/R — i.e. `FOLDER_SUPPORTING_COMPONENT_IDS` includes the DuckDB component ID. |
+| Caller mistakenly calls `update_sql_transformation` on a DuckDB config | 404 error message mentions DuckDB as a valid alternative alongside Python/R, so the caller self-corrects to `update_config`. |
+| DuckDB-specific sync actions (`syntax_check`, `lineage_visualization`, `execution_plan_visualization`, `expected_input_tables`) | No new code — `run_sync_action` (`tools.py:1960-2030`) already dispatches generically to whatever sync action a component declares; this works automatically once the DuckDB component's API metadata lists them. |
+| DuckDB-specific config parameters (`backend_size`, `timeout`, `duckdb_version`, `use_parquet`, `infer_input_table_data_types`, etc.) | No new code — `create_config`/`update_config` already treat `parameters` as an opaque JSON blob validated against the component's `configuration_schema` fetched via `get_components`/`get_config_examples`. |
+
+## Resolution Strategy
+
+Minimal, additive change mirroring the existing Python/R pattern — no new tools, no workspace
+changes:
+
+1. **`utils.py:68`** — add `DUCKDB_TRANSFORMATION_ID = 'keboola.duckdb-transformation'` next to
+   `PYTHON_TRANSFORMATION_ID` / `R_TRANSFORMATION_ID`.
+2. **`utils.py:74`** — add `DUCKDB_TRANSFORMATION_ID` to `FOLDER_SUPPORTING_COMPONENT_IDS`.
+3. **`tools.py:1121-1138` (`create_config` docstring)** — extend the "Not for SQL transformations"
+   line and "WHEN NOT TO USE" bullet to name Python/R/DuckDB explicitly as transformation types
+   that *do* go through `create_config`, so the model doesn't have to infer this from precedent.
+4. **`tools.py:996-1000` (`update_sql_transformation_internal` `ToolError`)** — extend the message
+   to say "...Python, R, or DuckDB transformation, use `update_config` with component_id
+   `keboola.python-transformation-v2`, `keboola.r-transformation-v2`, or
+   `keboola.duckdb-transformation`...".
+5. **`create_sql_transformation`/`update_sql_transformation` docstrings (`tools.py:440-473`,
+   `563-...`)** — add one line clarifying these tools only ever produce Snowflake/BigQuery
+   transformations (component ID derived from the workspace backend) and that DuckDB is handled
+   via `create_config`/`update_config`.
+
+No changes to `workspace.py`, `sql_utils.py`, or the sync-action dispatch — those are already
+either backend-specific by design (workspace = Snowflake/BigQuery only, intentionally out of
+scope for DuckDB) or already generic enough to cover DuckDB for free.
+
+## Scope
+
+In scope:
+
+- The five code changes above (constants, docstrings, error message).
+- Unit tests: `DUCKDB_TRANSFORMATION_ID` present in `FOLDER_SUPPORTING_COMPONENT_IDS`; folder
+  metadata applied on `update_config` for a DuckDB component ID (extend the existing
+  parametrized Python/R folder test with a DuckDB case rather than adding a new test function);
+  updated error-message assertion in the `update_sql_transformation` 404 test.
+- Version bump (minor — new tool-facing guidance/behavior) + `uv lock` + `TOOLS.md` regen.
+
+Out of scope (tracked separately, each with its own RFC per CONTRIBUTING.md):
+
+- Consolidating `create_sql_transformation`/`create_config`/`update_config` into fewer, more
+  generic tools, and moving the guidance currently duplicated across tool docstrings into
+  `project_system_prompt.md` (or a future Agent Skill). Tracked in a follow-up Linear issue,
+  explicitly scoped to build on top of this change.
+- Any DuckDB-specific workspace/backend support for `query_data` — DuckDB is not a workspace
+  backend and this RFC does not add one.
+- New dedicated sync-action tools — `run_sync_action` already covers this generically.
+
+## Testing / Verification
+
+1. `tox` — pytest, black, isort, flake8, check-tools-docs all exit 0.
+2. Unit tests in `tests/tools/components/test_utils.py` / `test_tools.py`: extend existing
+   parametrized folder-metadata and error-message tests with a DuckDB axis, rather than adding
+   new test functions.
+3. Manual E2E via local `.mcp.json` against a project with a DuckDB transformation: confirm
+   `create_config`/`update_config` with `component_id='keboola.duckdb-transformation'` succeeds,
+   folder metadata is applied, and `run_sync_action` executes `syntax_check` successfully.
+4. `tox -e check-tools-docs` to regenerate `TOOLS.md` after docstring changes.
diff --git a/src/keboola_mcp_server/tools/components/tools.py b/src/keboola_mcp_server/tools/components/tools.py
index cc5d2a998..210c6f0c1 100644
--- a/src/keboola_mcp_server/tools/components/tools.py
+++ b/src/keboola_mcp_server/tools/components/tools.py
@@ -463,6 +463,8 @@ async def create_sql_transformation(
     - This is THE tool for creating `keboola.snowflake-transformation` and `keboola.google-bigquery-transformation`
       components (do NOT use `create_config` for these); the transformation ID is derived automatically from the
       workspace SQL dialect.
+    - Snowflake/BigQuery only. For Python, R, or DuckDB transformations, use `create_config` with the appropriate
+      `component_id` instead — this tool cannot create them.
 
     EXAMPLES:
     - user_input: `Can you create a new transformation out of this sql query?`
@@ -673,7 +675,7 @@ async def update_sql_transformation(
     Use this for modifying SQL transformations created with create_sql_transformation.
 
     WHEN TO USE:
-    - SQL transformations only (Snowflake/BigQuery); use update_config for Python/R transformations
+    - SQL transformations only (Snowflake/BigQuery); use update_config for Python/R/DuckDB transformations
     - Modifying SQL queries in transformation (add/edit/remove SQL statements)
     - Updating transformation block or code block names
     - Changing input/output table mappings for the transformation
@@ -995,9 +997,9 @@ async def update_sql_transformation_internal(
         if e.response.status_code == 404:
             raise ToolError(
                 f"Configuration '{configuration_id}' was not found under SQL transformation component "
-                f"'{sql_transformation_id}'. If this is a Python or R transformation, use 'update_config' "
-                f"with component_id 'keboola.python-transformation-v2' or 'keboola.r-transformation-v2' "
-                f"instead of 'update_sql_transformation'."
+                f"'{sql_transformation_id}'. If this is a Python, R, or DuckDB transformation, use 'update_config' "
+                f"with component_id 'keboola.python-transformation-v2', 'keboola.r-transformation-v2', or "
+                f"'keboola.duckdb-transformation' instead of 'update_sql_transformation'."
             ) from e
         raise
     api_component = await fetch_component(client=client, component_id=sql_transformation_id)
@@ -1121,7 +1123,9 @@ async def create_config(
     """
     Creates a root component configuration using the specified name, component ID, configuration JSON, and description.
     Not for SQL transformations (`keboola.snowflake-transformation` / `keboola.google-bigquery-transformation`),
-    data apps (`keboola.data-apps`) or flows — use the dedicated tools (see WHEN NOT TO USE).
+    data apps (`keboola.data-apps`) or flows — use the dedicated tools (see WHEN NOT TO USE). This IS the tool for
+    Python (`keboola.python-transformation-v2`), R (`keboola.r-transformation-v2`) and DuckDB
+    (`keboola.duckdb-transformation`) transformations.
 
     BEFORE CALLING - REQUIRED STEPS:
     1. Call `get_components([component_id])` to retrieve the component's `configuration_schema`.
diff --git a/src/keboola_mcp_server/tools/components/utils.py b/src/keboola_mcp_server/tools/components/utils.py
index 7bcad1e6c..4ad2beb51 100644
--- a/src/keboola_mcp_server/tools/components/utils.py
+++ b/src/keboola_mcp_server/tools/components/utils.py
@@ -67,12 +67,15 @@
 BIGQUERY_TRANSFORMATION_ID = 'keboola.google-bigquery-transformation'
 PYTHON_TRANSFORMATION_ID = 'keboola.python-transformation-v2'
 R_TRANSFORMATION_ID = 'keboola.r-transformation-v2'
+DUCKDB_TRANSFORMATION_ID = 'keboola.duckdb-transformation'
 VARIABLES_COMPONENT_ID = 'keboola.variables'
 
 # Component IDs for which update_config actively manages folder metadata (set/clear/hint).
 # For all other components the folder parameter is accepted but silently skipped to avoid
 # unnecessary API calls on components where folder organisation is not expected.
-FOLDER_SUPPORTING_COMPONENT_IDS: frozenset[str] = frozenset({PYTHON_TRANSFORMATION_ID, R_TRANSFORMATION_ID})
+FOLDER_SUPPORTING_COMPONENT_IDS: frozenset[str] = frozenset(
+    {PYTHON_TRANSFORMATION_ID, R_TRANSFORMATION_ID, DUCKDB_TRANSFORMATION_ID}
+)
 
 
 # ============================================================================
diff --git a/tests/tools/components/test_tools.py b/tests/tools/components/test_tools.py
index c14f501c4..26022b94b 100644
--- a/tests/tools/components/test_tools.py
+++ b/tests/tools/components/test_tools.py
@@ -1314,7 +1314,7 @@ async def test_update_sql_transformation_wrong_component_type(
 ) -> None:
     """
     update_sql_transformation should raise ToolError with actionable guidance when the
-    configuration belongs to a Python/R transformation (Storage returns 404 for the SQL
+    configuration belongs to a Python/R/DuckDB transformation (Storage returns 404 for the SQL
     component + config-ID combination).
     """
     context = mcp_context_components_configs
@@ -1342,6 +1342,8 @@ async def test_update_sql_transformation_wrong_component_type(
     assert 'keboola.snowflake-transformation' in error_msg
     assert 'update_config' in error_msg
     assert 'keboola.python-transformation-v2' in error_msg
+    assert 'keboola.r-transformation-v2' in error_msg
+    assert 'keboola.duckdb-transformation' in error_msg
 
 
 @pytest.mark.asyncio
@@ -1655,14 +1657,23 @@ async def test_update_config(
 
 
 @pytest.mark.parametrize(
-    ('folder', 'cfg_count', 'cfg_folders', 'expect_folder_metadata', 'expect_folder_delete', 'expect_hint'),
+    (
+        'component_id',
+        'folder',
+        'cfg_count',
+        'cfg_folders',
+        'expect_folder_metadata',
+        'expect_folder_delete',
+        'expect_hint',
+    ),
     [
-        ('Analytics', 0, [], True, False, False),
-        ('  Analytics  ', 0, [], True, False, False),
-        (None, 5, [], False, False, False),
-        (None, 25, ['Analytics'], False, False, True),
-        (None, 25, [], False, False, True),
-        ('', 5, [], False, True, False),
+        ('keboola.python-transformation-v2', 'Analytics', 0, [], True, False, False),
+        ('keboola.python-transformation-v2', '  Analytics  ', 0, [], True, False, False),
+        ('keboola.python-transformation-v2', None, 5, [], False, False, False),
+        ('keboola.python-transformation-v2', None, 25, ['Analytics'], False, False, True),
+        ('keboola.python-transformation-v2', None, 25, [], False, False, True),
+        ('keboola.python-transformation-v2', '', 5, [], False, True, False),
+        ('keboola.duckdb-transformation', 'Analytics', 0, [], True, False, False),
     ],
     ids=[
         'folder_provided',
@@ -1671,6 +1682,7 @@ async def test_update_config(
         'no_folder_many_with_folders',
         'no_folder_many_no_folders',
         'folder_empty_deletes',
+        'duckdb_folder_provided',
     ],
 )
 @pytest.mark.asyncio
@@ -1678,6 +1690,7 @@ async def test_update_config_folder(
     mocker: MockerFixture,
     mcp_context_components_configs: Context,
     mock_component: dict[str, Any],
+    component_id: str,
     folder: Any,
     cfg_count: int,
     cfg_folders: list[str],
@@ -1688,7 +1701,6 @@ async def test_update_config_folder(
     """Test folder metadata is set/cleared and folder hint is returned by update_config."""
     context = mcp_context_components_configs
     keboola_client = KeboolaClient.from_state(context.session.state)
-    component_id = 'keboola.python-transformation-v2'
     mock_component['id'] = component_id
     configuration_id = 'cfg-folder-test'
     existing = {