Chat completion | Scout | Documentation

curl -X POST https://core.usescout.sh/v1/chat/completions \
     -H "Authorization: Bearer <token>" \
     -H "Content-Type: application/json" \
     -d '{
  "messages": [
    {
      "role": "user",
      "content": "What did Anthropic announce this week?"
    }
  ],
  "model": "scout-web"
}'

{
  "id": "chatcmpl-2b8f1d6a4c9e",
  "object": "chat.completion",
  "created": 1737382991,
  "model": "scout-web",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "This week Anthropic announced a new Claude model with improved coding and reasoning performance."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 28,
    "total_tokens": 42
  }
}

OpenAI-compatible chat completions, grounded in live web search.

When web_search is false (or tools is false) we skip the SERP fetch entirely and route straight to Claude — sub-1s for simple chat. When stream is true we return Server-Sent Events with OpenAI-style chat.completion.chunk frames terminated by data: [DONE].

curl -X POST https://core.usescout.sh/v1/chat/completions \
     -H "Authorization: Bearer <token>" \
     -H "Content-Type: application/json" \
     -d '{
  "messages": [
    {
      "role": "user",
      "content": "What did Anthropic announce this week?"
    }
  ],
  "model": "scout-web"
}'

{
  "id": "chatcmpl-2b8f1d6a4c9e",
  "object": "chat.completion",
  "created": 1737382991,
  "model": "scout-web",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "This week Anthropic announced a new Claude model with improved coding and reasoning performance."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 28,
    "total_tokens": 42
  }
}

OpenAI-compatible chat completions, grounded in live web search.

Authentication

AuthorizationBearer

Your API key, sent as a Bearer token.

Request

This endpoint expects an object.

messageslist of objectsRequired

OpenAI-style messages (system / user / assistant).

modelstring or nullOptional

Echoed back in the response; does not affect routing.

streambooleanOptionalDefaults to false

When true, return Server-Sent Events with OpenAI-style chat.completion.chunk frames terminated by data: [DONE].

response_formatobject or nullOptional

Output format. text (default), json_object, or json_schema. For json_schema, pass {type:‘json_schema’, json_schema:{name, schema}}.

temperaturedouble or nullOptional0-2

OpenAI-range 0-2. Mapped to Anthropic’s 0-1 range (clamped).

top_pdouble or nullOptional0-1

Nucleus sampling cutoff. Forwarded to Anthropic.

nintegerOptional1-4Defaults to 1

Number of chat completions to generate. Capped at 4.

toolsboolean or list of maps from strings to any or nullOptional

false disables Scout’s built-in web search (pure chat). An array is accepted for OpenAI compatibility but client-defined tools are not yet wired up; Scout’s built-in search is always available.

web_searchbooleanOptionalDefaults to true

Explicit toggle for the built-in web grounding. Set to false for pure chat (sub-1s).

seedinteger or nullOptional

Deterministic sampling seed (forwarded when supported).

max_tokensintegerOptional1-8192Defaults to 2048

Max tokens in the response.

stopstring or list of strings or nullOptional

String or list of stop sequences. Forwarded as Anthropic stop_sequences.

Response

Successful Response

Errors

422

Unprocessable Entity Error

$	curl -X POST https://core.usescout.sh/v1/chat/completions \
>	-H "Authorization: Bearer <token>" \
>	-H "Content-Type: application/json" \
>	-d '{
>	"messages": [
>	{
>	"role": "user",
>	"content": "What did Anthropic announce this week?"
>	}
>	],
>	"model": "scout-web"
>	}'

1	{
2	"id": "chatcmpl-2b8f1d6a4c9e",
3	"object": "chat.completion",
4	"created": 1737382991,
5	"model": "scout-web",
6	"choices": [
7	{
8	"index": 0,
9	"message": {
10	"role": "assistant",
11	"content": "This week Anthropic announced a new Claude model with improved coding and reasoning performance."
12	},
13	"finish_reason": "stop"
14	}
15	],
16	"usage": {
17	"prompt_tokens": 14,
18	"completion_tokens": 28,
19	"total_tokens": 42
20	}
21	}