> **Building with AI coding agents?** Install the authstack plugin with one command. This equips your agent with accurate Scalekit implementation patterns.
>
> **Recommended** (the `-y` flags skip prompts, so the command runs without an interactive terminal):
> ```bash
> npx -y @scalekit-inc/cli setup -y
> ```
>
> Global:
> ```bash
> npm install -g @scalekit-inc/cli
> scalekit setup -y
> ```
>
> Supports Claude Code, Cursor, GitHub Copilot, Codex + skills for other Agent Skills-compatible agents in `~/.agents/skills`.
> Skills: integrate-agentkit, implement-saaskit, add-mcp-oauth, implement-sso, implement-scim.
> [Full setup guide](https://docs.scalekit.com/dev-kit/build-with-ai/)

---

# Connect AI agents to Qdrant

The Qdrant connector lets your AI agent act in each user's Qdrant account. Each user connects their own Qdrant API key once, and Scalekit sends it with every call, so your agent never handles credentials. It comes with 24 tools.

**Authentication:** API Key
**Categories:** Databases, AI
**Tools:** 24: 14 read, 4 write, 6 destructive
**Users sign in with:** API key
**Built by:** Scalekit connector
**Try it:** [Playground in the Scalekit dashboard](https://app.scalekit.com/ws/signup?sk_intent=playground&provider=QDRANT)

## Setup

1. ### Install the SDK

   **Node.js**

   ```bash
   npm install @scalekit-sdk/node dotenv
   ```

   **Python**

   ```bash
   pip install scalekit-sdk-python python-dotenv
   ```

2. ### Set your credentials

   Add your Scalekit credentials to your `.env` file. Find values in **[app.scalekit.com](https://app.scalekit.com)** > **Developers** > **API Credentials**.

   ```sh title=".env"
   SCALEKIT_ENVIRONMENT_URL=<your-environment-url>
   SCALEKIT_CLIENT_ID=<your-client-id>
   SCALEKIT_CLIENT_SECRET=<your-client-secret>
   ```

3. ### Create the Qdrant connection

   In **AgentKit > Connections**, create a Qdrant connection. The name you give it is the `connection_name` your code passes. See [Configure connections](/agentkit/connections/).

4. ### Authorize a user and make your first call

   **Node.js** (`quickstart.mts`)

   ```typescript
   import { ScalekitClient } from '@scalekit-sdk/node'
   import 'dotenv/config'
   import { createInterface } from 'node:readline/promises'

   const scalekit = new ScalekitClient(
     process.env.SCALEKIT_ENVIRONMENT_URL,
     process.env.SCALEKIT_CLIENT_ID,
     process.env.SCALEKIT_CLIENT_SECRET,
   )
   const actions = scalekit.actions

   const connector = 'qdrant'
   const identifier = 'user_123'

   // Generate an authorization link for the user
   const { link } = await actions.getAuthorizationLink({ connectionName: connector, identifier })
   console.log('Authorize Qdrant:', link)
   const rl = createInterface({ input: process.stdin, output: process.stdout })
   await rl.question('Press Enter after authorizing...')
   rl.close()

   // Make your first call
   const result = await actions.executeTool({
     connector,
     identifier,
     toolName: 'qdrant_aliases_list',
     toolInput: {},
   })
   console.log(result)
   ```

   ```bash
   npx tsx quickstart.mts
   ```

   **Python** (`quickstart.py`)

   ```python
   import os
   from scalekit import ScalekitClient
   from dotenv import load_dotenv
   load_dotenv()

   scalekit_client = ScalekitClient(
       env_url=os.getenv("SCALEKIT_ENVIRONMENT_URL"),
       client_id=os.getenv("SCALEKIT_CLIENT_ID"),
       client_secret=os.getenv("SCALEKIT_CLIENT_SECRET"),
   )
   actions = scalekit_client.actions

   connection_name = "qdrant"
   identifier = "user_123"

   # Generate an authorization link for the user
   link_response = actions.get_authorization_link(
       connection_name=connection_name,
       identifier=identifier,
   )
   print("Authorize Qdrant:", link_response.link)
   input("Press Enter after authorizing...")

   # Make your first call
   result = actions.execute_tool(
       tool_input={},
       tool_name="qdrant_aliases_list",
       connection_name=connection_name,
       identifier=identifier,
   )
   print(result)
   ```

   ```bash
   python quickstart.py
   ```

   Each user opens the link once and enters their Qdrant credentials there. See [Authorize a user](/agentkit/tools/authorize/) for the full flow and statuses.

## Tools

Pass the exact name to `execute_tool`, with an input like each tool's example. To ask for a tool that's missing, use the [request form](https://scalekitsupport.portal.usepylon.com/forms/request-a-connector-tool).

**Node.js**

```typescript
const result = await actions.executeTool({
  toolName: 'qdrant_aliases_list',
  toolInput: {},
  connector: 'qdrant',
  identifier: 'user_123',
})
```

**Python**

```python
result = actions.execute_tool(
    tool_name="qdrant_aliases_list",
    tool_input={},
    connection_name="qdrant",
    identifier="user_123",
)
```

`result.data` is the app's response as JSON, and `result.execution_id` (`executionId` in Node.js) is the ID of the call.

### `qdrant_aliases_list`

List Aliases · Read-only

List every collection alias in the Qdrant cluster.
Returns each alias name with the collection it points to.
Use this to resolve an alias to its collection. Use qdrant_collection_aliases_list for the aliases of one collection.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs: none.

Example input: `{}`

### `qdrant_collection_aliases_list`

List Collection Aliases · Read-only

List the aliases that point to one Qdrant collection.
Returns each alias name with the collection name.
Use this for one collection's aliases. Use qdrant_aliases_list for every alias in the cluster.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_collection_exists`

Check Collection Exists · Read-only

Check whether a collection with a given name exists in the Qdrant cluster.
Returns exists: true or false.
Use this before creating a collection or to confirm a name. Use qdrant_collection_get for the collection's full configuration.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_collection_get`

Get Collection · Read-only

Get the details of one Qdrant collection.
Returns its status, point and indexed-vector counts, vector configuration (sizes, distance, named and sparse vectors), payload schema with indexed fields, and optimizer, HNSW and quantization settings.
Use this to learn a collection's vector size, vector names and indexed payload fields before writing or searching. Use qdrant_collection_exists for a yes/no check.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_collections_list`

List Collections · Read-only

List every collection in the Qdrant cluster.
Returns the name of each collection, without configuration or point counts.
Use this to discover collection names. Use qdrant_collection_get for one collection's configuration, status and point count.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs: none.

Example input: `{}`

### `qdrant_point_get`

Get Point · Read-only

Get one point from a Qdrant collection by its ID.
Returns the point's id, payload and vectors.
Use this for a single known ID. Use qdrant_points_retrieve for several IDs at once, and qdrant_points_scroll to browse points by filter.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `point_id` (`string`, required): ID of the point to fetch: an unsigned integer or a UUID, as stored. Example: 550e8400-e29b-41d4-a716-446655440000

Example input: `{"collection_name":"<collection_name>","point_id":"<point_id>"}`

### `qdrant_points_count`

Count Points · Read-only

Count the points in a Qdrant collection, optionally only those matching a filter.
Returns count.
Use this for totals. Use qdrant_points_facet for counts per payload value, and qdrant_points_scroll to list the points.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `exact` (`boolean`, optional): true counts exactly; false returns a faster estimate that can be off while the collection is indexing. Qdrant counts exactly by default. Default: `true`.
- `filter` (`object`, optional): Qdrant filter that limits which points are counted. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_points_facet`

Facet Points · Read-only

Count how many points have each value of a payload field in a Qdrant collection, optionally within a filter.
Returns hits, each with a value and its count, most frequent first.
Use this to see the distinct values of a field and how common they are. Use qdrant_points_count for a single total.
The field needs a keyword, integer, bool or uuid payload index (qdrant_payload_index_create).

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `key` (`string`, required): Payload key to count values for; it must have a keyword, integer, bool or uuid payload index. Nested keys use dot notation. Example: category
- `exact` (`boolean`, optional): true counts each value exactly (slower); false (Qdrant's default) returns faster approximate counts.
- `filter` (`object`, optional): Qdrant filter that limits which points are counted. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `limit` (`integer`, optional): Maximum number of distinct values to return. Qdrant returns 10 when omitted. Example: 10

Example input: `{"collection_name":"<collection_name>","key":"<key>"}`

### `qdrant_points_query`

Query Points · Read-only

Run any query through Qdrant's universal Query API: nearest-neighbour, recommendation, hybrid dense and sparse search with fusion, multi-stage prefetch and re-ranking, ordering by a payload field or random sampling.
Returns points ranked by score, each with id, score, version and, if requested, payload and vectors.
Use this for hybrid, fusion or multi-stage queries and other query types. Use qdrant_points_search for a plain vector search, qdrant_points_recommend for examples, and qdrant_points_query_groups to group results by a payload field.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `filter` (`object`, optional): Qdrant filter that limits which points can be returned. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `limit` (`integer`, optional): Maximum number of results to return. Qdrant returns 10 when omitted. Example: 10
- `offset` (`integer`, optional): Number of top results to skip, for paging through results. Qdrant uses 0 when omitted. Example: 10
- `params` (`object`, optional): Advanced search settings: hnsw_ef (search breadth; higher is more accurate and slower), exact (true for exact brute-force search) and quantization options. Example: {"hnsw_ef": 128, "exact": false}
- `payload_fields` (`array`, optional): Return only these payload keys instead of the whole payload. Overrides Include Payload when set. Nested keys use dot notation. Example: ["title", "metadata.author"]
- `prefetch` (`array`, optional): Sub-queries that run first; the main query then re-ranks or fuses their results. Each sub-query is an object with its own query, using, filter and limit. Typical hybrid search: a dense and a sparse prefetch fused with query {"fusion": "rrf"}. Example: [{"query": [0.2, 0.1, 0.9, 0.7], "using": "dense", "limit": 20}, {"query": {"indices": [1, 42], "values": [0.22, 0.8]}, "using": "sparse", "limit": 20}]
- `query` (`object`, optional): The query to run, as a Qdrant Query API object. Common forms: nearest search with a vector or an existing point ID ({"nearest": [0.2, 0.1, 0.9]} or {"nearest": 42}); recommendations ({"recommend": {"positive": [100], "negative": [718]}}); fusion of prefetch results ({"fusion": "rrf"}); ordering by a payload field ({"order_by": "created_at"}); and random sampling ({"sample": "random"}). Leave empty with no prefetch to list points by ID.
- `score_threshold` (`number`, optional): Return only points whose score is better than this value. For Cosine and Dot distance a higher score is better; for Euclid and Manhattan a lower score is better. Example: 0.75
- `using` (`string`, optional): Name of the named vector to search with, for collections that store several vectors per point (see the vectors section of qdrant_collection_get). Leave empty for a collection with a single unnamed vector. Example: image
- `with_payload` (`boolean`, optional): Whether to include each point's payload (its JSON metadata) in the response. Qdrant leaves the payload out when omitted, so set true to see point metadata.
- `with_vector` (`boolean`, optional): Whether to include the stored vectors of each point in the response. Qdrant leaves vectors out unless this is true. Vectors can be large, so enable it only when you need them.

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_points_query_groups`

Query Points Grouped · Read-only

Run a Query API query in a Qdrant collection and group the results by a payload field, returning the best points of each group.
Returns groups, each with the group id (the field value) and its top hits with id, score and, if requested, payload.
Use this to avoid many results from the same document or entity, for example one best chunk per document. Use qdrant_points_query for ungrouped results.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `group_by` (`string`, required): Payload field to group results by; it must hold a string or number (a point with several values joins several groups). Example: document_id
- `filter` (`object`, optional): Qdrant filter that limits which points can be returned. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `group_size` (`integer`, optional): Maximum number of points returned in each group. Qdrant uses 3 when omitted. Example: 1
- `limit` (`integer`, optional): Maximum number of groups to return. Qdrant returns 10 when omitted. Example: 10
- `params` (`object`, optional): Advanced search settings: hnsw_ef (search breadth; higher is more accurate and slower), exact (true for exact brute-force search) and quantization options. Example: {"hnsw_ef": 128, "exact": false}
- `payload_fields` (`array`, optional): Return only these payload keys instead of the whole payload. Overrides Include Payload when set. Nested keys use dot notation. Example: ["title", "metadata.author"]
- `prefetch` (`array`, optional): Sub-queries that run first; the main query then re-ranks or fuses their results. Each sub-query is an object with its own query, using, filter and limit. Typical hybrid search: a dense and a sparse prefetch fused with query {"fusion": "rrf"}. Example: [{"query": [0.2, 0.1, 0.9, 0.7], "using": "dense", "limit": 20}, {"query": {"indices": [1, 42], "values": [0.22, 0.8]}, "using": "sparse", "limit": 20}]
- `query` (`object`, optional): The query to run, as a Qdrant Query API object. Common forms: nearest search with a vector or an existing point ID ({"nearest": [0.2, 0.1, 0.9]} or {"nearest": 42}); recommendations ({"recommend": {"positive": [100], "negative": [718]}}); fusion of prefetch results ({"fusion": "rrf"}); ordering by a payload field ({"order_by": "created_at"}); and random sampling ({"sample": "random"}). Leave empty with no prefetch to list points by ID.
- `score_threshold` (`number`, optional): Return only points whose score is better than this value. For Cosine and Dot distance a higher score is better; for Euclid and Manhattan a lower score is better. Example: 0.75
- `using` (`string`, optional): Name of the named vector to search with, for collections that store several vectors per point (see the vectors section of qdrant_collection_get). Leave empty for a collection with a single unnamed vector. Example: image
- `with_payload` (`boolean`, optional): Whether to include each point's payload (its JSON metadata) in the response. Qdrant leaves the payload out when omitted, so set true to see point metadata.
- `with_vector` (`boolean`, optional): Whether to include the stored vectors of each point in the response. Qdrant leaves vectors out unless this is true. Vectors can be large, so enable it only when you need them.

Example input: `{"collection_name":"<collection_name>","group_by":"<group_by>"}`

### `qdrant_points_recommend`

Recommend Points · Read-only

Find points in a Qdrant collection similar to positive examples and unlike negative examples, given as point IDs or vectors.
Returns points ranked by score, each with id, score, version and, if requested, payload and vectors; the example points themselves are excluded.
Use this for more-like-this recommendations from existing points. Use qdrant_points_search when you have a single query vector.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `positive` (`array`, required): Examples the results should resemble: point IDs from the collection (integers or UUID strings) or raw vectors (arrays of numbers). Example: [100, 231]
- `filter` (`object`, optional): Qdrant filter that limits which points can be returned. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `limit` (`integer`, optional): Maximum number of results to return. Qdrant returns 10 when omitted. Example: 10
- `negative` (`array`, optional): Examples the results should be unlike: point IDs or raw vectors. Example: [718]
- `payload_fields` (`array`, optional): Return only these payload keys instead of the whole payload. Overrides Include Payload when set. Nested keys use dot notation. Example: ["title", "metadata.author"]
- `score_threshold` (`number`, optional): Return only points whose score is better than this value. For Cosine and Dot distance a higher score is better; for Euclid and Manhattan a lower score is better. Example: 0.75
- `strategy` (`string`, optional): How examples are combined: average_vector (default) averages them into one query vector; best_score scores each candidate against every example and works better with many or negative examples; sum_scores adds positive and subtracts negative similarities. One of: `average_vector`, `best_score`, `sum_scores`.
- `using` (`string`, optional): Name of the named vector to search with, for collections that store several vectors per point (see the vectors section of qdrant_collection_get). Leave empty for a collection with a single unnamed vector. Example: image
- `with_payload` (`boolean`, optional): Whether to include each point's payload (its JSON metadata) in the response. Qdrant leaves the payload out when omitted, so set true to see point metadata.
- `with_vector` (`boolean`, optional): Whether to include the stored vectors of each point in the response. Qdrant leaves vectors out unless this is true. Vectors can be large, so enable it only when you need them.

Example input: `{"collection_name":"<collection_name>","positive":[100,231]}`

### `qdrant_points_retrieve`

Retrieve Points · Read-only

Get several points from a Qdrant collection by their IDs in one call.
Returns each found point's id with its payload (included by default) and, if requested, its vectors; unknown IDs are skipped.
Use this when you already have point IDs. Use qdrant_point_get for one ID, qdrant_points_scroll to browse by filter, and qdrant_points_search to find similar points.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `ids` (`array`, required): IDs of the points to fetch. Each ID is an unsigned integer or a UUID string. Example: [1, 42, "550e8400-e29b-41d4-a716-446655440000"]
- `payload_fields` (`array`, optional): Return only these payload keys instead of the whole payload. Overrides Include Payload when set. Nested keys use dot notation. Example: ["title", "metadata.author"]
- `with_payload` (`boolean`, optional): Whether to include each point's payload (its JSON metadata) in the response. Qdrant includes the payload when omitted.
- `with_vector` (`boolean`, optional): Whether to include the stored vectors of each point in the response. Qdrant leaves vectors out unless this is true. Vectors can be large, so enable it only when you need them.

Example input: `{"collection_name":"<collection_name>","ids":[1,42,"550e8400-e29b-41d4-a716-446655440000"]}`

### `qdrant_points_scroll`

Scroll Points · Read-only

Page through the points of a Qdrant collection, optionally filtered by payload and ordered by a payload field.
Returns a page of points (id, payload and optionally vectors) and next_page_offset, the ID to pass as offset for the next page (null on the last page).
Use this to browse or export points by metadata without a query vector. Use qdrant_points_search for similarity search and qdrant_points_count for a total.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `filter` (`object`, optional): Qdrant filter that limits which points are returned. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `limit` (`integer`, optional): Maximum number of points to return in this page. Qdrant returns 10 when omitted. Example: 10
- `offset` (`integer`, optional): Point ID to start the page from, taken from next_page_offset of the previous response. Leave empty for the first page. Not used together with order_by. Example: 1043
- `order_by` (`string`, optional): Payload key to sort the points by, instead of by ID. The key needs a payload index of type integer, float or datetime (see qdrant_payload_index_create). Page with a range filter on that key rather than offset. Example: created_at
- `order_direction` (`string`, optional): Sort direction when order_by is set: asc (smallest first, the default) or desc. One of: `asc`, `desc`.
- `payload_fields` (`array`, optional): Return only these payload keys instead of the whole payload. Overrides Include Payload when set. Nested keys use dot notation. Example: ["title", "metadata.author"]
- `with_payload` (`boolean`, optional): Whether to include each point's payload (its JSON metadata) in the response. Qdrant includes the payload when omitted.
- `with_vector` (`boolean`, optional): Whether to include the stored vectors of each point in the response. Qdrant leaves vectors out unless this is true. Vectors can be large, so enable it only when you need them.

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_points_search`

Search Similar Points · Read-only

Find the points in a Qdrant collection whose vectors are most similar to a query vector, optionally filtered by payload.
Returns points ranked by score, each with id, score, version and, if requested, payload and vectors.
Use this for semantic search with an embedding you already computed. Use qdrant_points_recommend to search from example point IDs, and qdrant_points_query for hybrid, fusion or multi-stage queries.
The vector length must match the collection's vector size from qdrant_collection_get.

Read only: yes. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `vector` (`array`, required): The query embedding: an array of numbers with the same length as the collection's (or named vector's) size, produced by the same model as the stored vectors. Example: [0.2, 0.1, 0.9, 0.7]
- `filter` (`object`, optional): Qdrant filter that limits which points can be returned. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `limit` (`integer`, optional): Maximum number of results to return. Qdrant returns 10 when omitted. Example: 10
- `offset` (`integer`, optional): Number of top results to skip, for paging through results. Qdrant uses 0 when omitted. Example: 10
- `payload_fields` (`array`, optional): Return only these payload keys instead of the whole payload. Overrides Include Payload when set. Nested keys use dot notation. Example: ["title", "metadata.author"]
- `score_threshold` (`number`, optional): Return only points whose score is better than this value. For Cosine and Dot distance a higher score is better; for Euclid and Manhattan a lower score is better. Example: 0.75
- `using` (`string`, optional): Name of the named vector to search with, for collections that store several vectors per point (see the vectors section of qdrant_collection_get). Leave empty for a collection with a single unnamed vector. Example: image
- `with_payload` (`boolean`, optional): Whether to include each point's payload (its JSON metadata) in the response. Qdrant leaves the payload out when omitted, so set true to see point metadata.
- `with_vector` (`boolean`, optional): Whether to include the stored vectors of each point in the response. Qdrant leaves vectors out unless this is true. Vectors can be large, so enable it only when you need them.

Example input: `{"collection_name":"<collection_name>","vector":[0.2,0.1,0.9,0.7]}`

### `qdrant_collection_create`

Create Collection · Write

Create a new Qdrant collection with its vector configuration.
Returns result: true when the collection is created.
Use this once per collection before upserting points; check the name with qdrant_collection_exists first. Use qdrant_payload_index_create afterwards to index payload fields you filter on.

Read only: no. Destructive: no. Idempotent (safe to retry): no.

Inputs:

- `collection_name` (`string`, required): Name for the new collection. Letters, digits, dashes and underscores work everywhere. It must not already exist. Example: products
- `distance` (`string`, optional): Distance function used to compare vectors of the single unnamed vector: Cosine, Euclid, Dot or Manhattan. Used with vector_size and ignored when vectors is set. Cosine is used when omitted. Example: Cosine. One of: `Cosine`, `Euclid`, `Dot`, `Manhattan`.
- `on_disk_payload` (`boolean`, optional): Whether to keep point payloads on disk instead of in RAM, saving memory at a small latency cost. Indexed payload fields stay in RAM. Qdrant's server default applies when omitted.
- `replication_factor` (`integer`, optional): Number of copies of each shard, for multi-node clusters. Qdrant uses 1 when omitted. Example: 2
- `shard_number` (`integer`, optional): Number of shards for the collection. Qdrant picks a value from the cluster size when omitted. Example: 2
- `sparse_vectors` (`object`, optional): Configuration for sparse vectors (for example BM25 or SPLADE keyword vectors). Each key is a sparse vector name; its value can be an empty object or set a modifier such as idf. Example: {"text-sparse": {"modifier": "idf"}}
- `vector_size` (`integer`, optional): Number of dimensions of the single unnamed dense vector each point will store. It must match the output size of the embedding model you use (for example 384, 768 or 1536). Required unless you pass vectors for named vectors or create a sparse-only collection. Example: 1536
- `vectors` (`object`, optional): Configuration for named vectors, for a collection that stores several vectors per point. Each key is a vector name and its value gives that vector's size and distance (and optionally on_disk or datatype). When set, vector_size and distance are ignored. Example: {"image": {"size": 512, "distance": "Cosine"}, "text": {"size": 768, "distance": "Dot"}}

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_payload_index_create`

Create Payload Index · Write

Create an index on a payload field of a Qdrant collection so filters, ordering and facets on that field are fast.
Returns the operation id and status.
Use this for each payload field you filter, order or facet on. Use qdrant_collection_get to see the fields already indexed, and qdrant_payload_index_delete to remove an index.

Read only: no. Destructive: no. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `field_name` (`string`, required): Payload field to index; nested fields use dot notation. Example: category
- `field_schema` (`string`, optional): Index type for the field: keyword (exact string match), integer, float, bool, geo (coordinates), datetime, uuid or text (full-text search). Required unless field_schema_params is set. Example: keyword. One of: `keyword`, `integer`, `float`, `geo`, `text`, `bool`, `datetime`, `uuid`.
- `field_schema_params` (`object`, optional): Advanced index settings as an object with a type plus type-specific options, used instead of field_schema. For example a tenant keyword index or a text index with a tokenizer. Example: {"type": "keyword", "is_tenant": true} or {"type": "text", "tokenizer": "word", "lowercase": true}
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>","field_name":"<field_name>"}`

### `qdrant_payload_set`

Set Payload · Write

Add or update payload keys on existing points in a Qdrant collection, chosen by ID or by filter; keys not named are kept.
Returns the operation id and status.
Use this to change metadata without re-sending vectors. Use qdrant_payload_overwrite to replace the whole payload, and qdrant_payload_delete_keys to remove keys.
Pass point_ids, or a filter when point_ids is empty.

Read only: no. Destructive: no. Idempotent (safe to retry): no.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `payload` (`object`, required): JSON object of payload keys and values to write. Example: {"category": "shoes", "price": 79.9, "tags": ["sale"]}
- `filter` (`object`, optional): Qdrant filter that limits which points are updated when no point IDs are given. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `key` (`string`, optional): Optional nested payload path to write under instead of the top level, in dot notation; the given payload is merged into the object at that path. Example: metadata.source
- `point_ids` (`array`, optional): Point IDs of the points to update; takes priority over filter. Each ID is an unsigned integer or a UUID string, exactly as stored. Example: [1, 42, "550e8400-e29b-41d4-a716-446655440000"]
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>","payload":{"category":"shoes","price":79.9,"tags":["sale"]}}`

### `qdrant_points_upsert`

Upsert Points · Write

Insert points into a Qdrant collection, or fully replace points whose IDs already exist.
Returns the operation id and status (completed when wait is true, acknowledged otherwise).
Use this to add or replace vectors with their payloads. Use qdrant_payload_set to change only the payload of existing points.
The collection must exist (qdrant_collection_create), and vectors must match its size from qdrant_collection_get.

Read only: no. Destructive: no. Idempotent (safe to retry): no.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `points` (`array`, required): Points to insert or replace. Each point needs an id (an unsigned integer or a UUID string) and a vector, and may carry a payload object of JSON metadata. The vector is an array of numbers whose length matches the collection's vector size, or, for named vectors, an object keyed by vector name. A point whose id already exists is fully replaced. Example: [{"id": 1, "vector": [0.05, 0.61, 0.76, 0.74], "payload": {"city": "Berlin"}}]
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>","points":[]}`

### `qdrant_collection_delete`

Delete Collection · Destructive

Permanently delete a Qdrant collection together with all of its points, payloads and indexes.
Returns result: true when the collection was deleted, false when it did not exist.
Use this only to remove a whole collection. Use qdrant_points_delete to remove some points and keep the collection.

Read only: no. Destructive: yes. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_payload_clear`

Clear Payload · Destructive

Remove the entire payload from points in a Qdrant collection, chosen by ID or by filter; the points and their vectors stay.
Returns the operation id and status.
Use this to strip all metadata from points. Use qdrant_payload_delete_keys to remove only some keys, and qdrant_points_delete to remove the points.
Pass point_ids, or a filter when point_ids is empty.

Read only: no. Destructive: yes. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `filter` (`object`, optional): Qdrant filter that limits which points are cleared when no point IDs are given. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `point_ids` (`array`, optional): Point IDs whose payload is cleared; takes priority over filter. Each ID is an unsigned integer or a UUID string, exactly as stored. Example: [1, 42, "550e8400-e29b-41d4-a716-446655440000"]
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>"}`

### `qdrant_payload_delete_keys`

Delete Payload Keys · Destructive

Remove the named payload keys from points in a Qdrant collection, chosen by ID or by filter; other keys and the vectors stay.
Returns the operation id and status.
Use this to drop specific metadata fields. Use qdrant_payload_clear to remove the whole payload.
Pass point_ids, or a filter when point_ids is empty.

Read only: no. Destructive: yes. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `keys` (`array`, required): Payload keys to remove; nested keys use dot notation. Example: ["discount", "metadata.draft"]
- `filter` (`object`, optional): Qdrant filter that limits which points lose the keys when no point IDs are given. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `point_ids` (`array`, optional): Point IDs whose keys are removed; takes priority over filter. Each ID is an unsigned integer or a UUID string, exactly as stored. Example: [1, 42, "550e8400-e29b-41d4-a716-446655440000"]
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>","keys":["discount","metadata.draft"]}`

### `qdrant_payload_index_delete`

Delete Payload Index · Destructive

Delete the index on a payload field of a Qdrant collection; the payload data itself is kept.
Returns the operation id and status.
Use this to drop an index you no longer filter on. Use qdrant_payload_index_create to add one.

Read only: no. Destructive: yes. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `field_name` (`string`, required): Indexed payload field whose index to delete, as shown in payload_schema of qdrant_collection_get. Example: category
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>","field_name":"<field_name>"}`

### `qdrant_payload_overwrite`

Overwrite Payload · Destructive

Replace the entire payload of existing points in a Qdrant collection, chosen by ID or by filter; keys not in the new payload are removed.
Returns the operation id and status.
Use this to reset metadata completely. Use qdrant_payload_set to add or change some keys and keep the rest.
Pass point_ids, or a filter when point_ids is empty.

Read only: no. Destructive: yes. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `payload` (`object`, required): The complete new payload for each selected point; it replaces everything stored before. Example: {"category": "shoes", "price": 79.9}
- `filter` (`object`, optional): Qdrant filter that limits which points are overwritten when no point IDs are given. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `point_ids` (`array`, optional): Point IDs of the points to overwrite; takes priority over filter. Each ID is an unsigned integer or a UUID string, exactly as stored. Example: [1, 42, "550e8400-e29b-41d4-a716-446655440000"]
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>","payload":{"category":"shoes","price":79.9}}`

### `qdrant_points_delete`

Delete Points · Destructive

Permanently delete points from a Qdrant collection, chosen by ID or by filter.
Returns the operation id and status.
Use this to remove specific points and keep the collection. Use qdrant_collection_delete to drop the whole collection.
Pass point_ids, or a filter when point_ids is empty.

Read only: no. Destructive: yes. Idempotent (safe to retry): yes.

Inputs:

- `collection_name` (`string`, required): Name of the Qdrant collection to act on. Get it from qdrant_collections_list. Example: products
- `filter` (`object`, optional): Qdrant filter that limits which points are deleted when no point IDs are given. It is an object with up to three condition lists: must (all must match), should (at least one must match) and must_not (none may match). Each condition usually names a payload key and a match, range or geo rule. Example: {"must": [{"key": "city", "match": {"value": "London"}}, {"key": "price", "range": {"lte": 100}}]}
- `point_ids` (`array`, optional): Point IDs of the points to delete; takes priority over filter. Each ID is an unsigned integer or a UUID string, exactly as stored. Example: [1, 42, "550e8400-e29b-41d4-a716-446655440000"]
- `wait` (`boolean`, optional): Whether to wait until the change is applied before responding. true (the default here) returns status completed once the data is searchable; false returns status acknowledged right away and applies the change in the background. Default: `true`.

Example input: `{"collection_name":"<collection_name>"}`

## Related

Other Databases connectors ([all 39](/agentkit/connectors/?category=databases)):

- [AWS Redshift](/agentkit/connectors/redshift/): Scalekit connector, Trusted identity provider, 10 tools
- [Google BigQuery](/agentkit/connectors/bigquery/): Scalekit connector, OAuth, 47 tools
- [Airtable](/agentkit/connectors/airtable/): Scalekit connector, OAuth, 26 tools


---

## More Scalekit documentation

| Resource | What it contains | When to use it |
|----------|-----------------|----------------|
| [/llms.txt](/llms.txt) | Structured index with routing hints per product area | Start here — find which documentation set covers your topic before loading full content |
| [/llms-full.txt](/llms-full.txt) | Complete documentation for all Scalekit products in one file | Use when you need exhaustive context across multiple products or when the topic spans several areas |
| [sitemap-0.xml](https://docs.scalekit.com/sitemap-0.xml) | Full URL list of every documentation page | Use to discover specific page URLs you can fetch for targeted, page-level answers |
