> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mindsdb.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Create Agent

**POST `/api/projects/{project_name}/agents`**

This API endpoint creates an agent using the `POST` method.

<Tip>
  Learn more about agents and the available parameters following [this doc page](/mindsdb_sql/agents/agent).
</Tip>

### Path Parameters

<ParamField body="project_name" type="string" required>
  Defines the project where the agents are located. Note that the default project name is `mindsdb`.
</ParamField>

### Body

<ParamField body="name" type="string" required>
  Name of the agent.
</ParamField>

<ParamField body="model" type="string">
  Stores parameters of the model, including `provider`, `model_name`, and `api_key`. Note that agents can use the default model defined in the configuration, if no model provided when creating an agent.
</ParamField>

<ParamField body="data" type="string">
  Stores data connected to an agent, including `tables` and `knowledge_bases`.
</ParamField>

<ParamField body="prompt_template" type="string">
  Stores instruction to an agent. This should contain the description of connected data.
</ParamField>

### Response

<ResponseField name="id" type="string" required>
  Unique identifier for the agent.
</ResponseField>

<ResponseField name="name" type="string" required>
  The name assigned to the agent.
</ResponseField>

<ResponseField name="project_id" type="string" required>
  The ID of the project where the agent resides.
</ResponseField>

<ResponseField name="created_at" type="string" required>
  Timestamp indicating when the agent was created.
</ResponseField>

<ResponseField name="updated_at" type="string" required>
  Timestamp indicating when the agent was last updated.
</ResponseField>

<ResponseField body="data" type="string">
  Stores data connected to an agent, including `tables` and `knowledge_bases`.

  In order to provide all tables from a database or all knowledge bases from a project, use the `*` wildcard like this:

  ```shell theme={null}
      "data": {
           "knowledge_bases": ["my_project.*"],
           "tables": ["my_data_source.*"]
      }
  ```
</ResponseField>

<ResponseField body="model" type="string">
  Stores parameters of the model, including `provider`, `model_name`, and `api_key`.
</ResponseField>

<ResponseField body="prompt_template" type="string">
  Stores instruction to an agent. This should contain the description of connected data.
</ResponseField>

<RequestExample>
  ```shell Shell theme={null}
  curl --request POST \
    --url http://127.0.0.1:47334/api/projects/mindsdb/agents \
    --header 'Content-Type: application/json' \
    --data '{
    "agent": {
      "name": "my_agent",
      "model": {
          "provider": "openai",
          "model_name": "gpt-4o",
          "api_key": "sk-xxx"
      },
      "data": {
           "knowledge_bases": ["my_project.my_kb"],
           "tables": ["my_data_source.my_table"]
      },
      "prompt_template": "my_project.my_kb stores documentation of MindsDB, my_data_source.my_table stores documentation of MindsDB"
    }
  }'
  ```
</RequestExample>

<ResponseExample>
  ```json Response theme={null}
  {
    "id": 197,
    "name": "my_agent",
    "project_id": 1,
    "created_at": "2025-07-09 12:58:24.868202",
    "updated_at": "2025-07-09 12:58:24.868199",
    "data": {
      "knowledge_bases": [
        "my_project.my_kb"
      ],
      "tables": [
        "my_data_source.my_table"
      ]
    },
    "model": {
      "provider": "openai",
      "model_name": "gpt-4o",
      "api_key": "sk-xxx"
    },
    "prompt_template": "my_project.my_kb stores documentation of MindsDB, my_data_source.my_table stores documentation of MindsDB"
  }
  ```
</ResponseExample>