POST
/
api
/
v1
/
agent
/
execute
Execute Agent (Beta)
curl --request POST \
  --url https://agents.bigdata.com/api/v1/agent/execute \
  --header 'Content-Type: application/json' \
  --header 'X-API-Key: <api-key>' \
  --data '{
  "chat_id": "EPH-12345-683279",
  "message": "What'\''s the latest news about Nvidia'\''s AI chip developments?",
  "mode": "research",
  "persistence_mode": "disabled",
  "tools_configs": {
    "search": {
      "query_filters": {
        "entities": {
          "any_of": [
            "ID1",
            "ID2"
          ]
        }
      }
    }
  }
}'
{
  "content": "I need to search for the latest news about Nvidia...",
  "message_id": "thinking_123",
  "role": "assistant",
  "type": "THINKING"
}

Authorizations

X-API-Key
string
header
required

Enter your API key

Body

application/json
mode
enum<string>
required

Agent execution mode: 'research' for comprehensive analysis, 'chat' for conversational interaction

Available options:
research,
chat
message
string
required

The user's message/question to send to the agent

toolsConfigs
object | null

Optional runtime configurations for tools. Available tools: 'search' - configure search parameters and filters.

chatId
string | null

Optional chat ID to continue an existing conversation thread. If not provided, a new conversation is started.

persistenceMode
enum<string>

Whether to persist the agent state in memory or in the database

Available options:
enabled,
disabled

Response

Successful Response

type
enum<string>
required
Available options:
THINKING,
ACTION,
ANSWER,
COMPLETE,
ERROR,
AUDIT,
GROUNDING,
TOOL_ERROR,
NOT_ENTITLED
chat_id
string
required

The ID of the chat

role
string
default:assistant
Allowed value: "assistant"
content
string | null

Text content of the message

message_id
string | null

Unique identifier for grouping related chunks

tool_name
string | null

Name of the tool being called (ACTION type only)

tool_arguments
object | null

Arguments for tool call (ACTION type only)

error
string | null

Error message (ERROR type only)

audit_traces
Audit Traces · array

Audit trace data (AUDIT type only)

references
ResolvedReference · object[]

Grounding references (GROUNDING type only)