Query all AI interactions

POST

/ai-governance/interactions

Query and filter AI interactions: per-action telemetry from coding agents
and AI apps (file reads, shell commands, MCP tool calls, plugin and skill
invocations, config tampering, prompt activity).

Filtering by time range: Use the timestamp field with comparison
operators. For example: timestamp >= '2026-01-01'

Filtering by kind: interaction_kind selects the category of event and
determines which evidence fields are populated. For example:
interaction_kind = 'mcp_tool_call' or
interaction_kind IN ('sensitive_file_read', 'sensitive_command').
Only exact enum string values are supported.

MCP tool calls: tool carries the MCP tool name on mcp_tool_call rows
and the agent's own tool name otherwise, so it is the single field to
filter on either way.

Relations:

employee (Person): The employee whose agent produced the interaction

Request

Authorization

Body Params application/jsonRequired

Examples

Responses

🟢200OK

application/json

Successful operation

Bodyapplication/json

🟠400Bad Request

🟠401Unauthorized

🟠403Forbidden

Request Example

Shell

JavaScript

Java

Swift

curl --location '/ai-governance/interactions' \
--header 'Content-Type: application/json' \
--data '{
    "distinct_count": "string",
    "distinct_on": "host_name",
    "include_total_count": true,
    "limit": 10,
    "offset": 0,
    "query": "name='\''WINPC-4291'\'' AND status IN ('\''active'\'', '\''pending'\'')",
    "sort": "name desc, id"
}'

Response Example

{
    "distinct_values": [
        {
            "count": 0,
            "value": null
        }
    ],
    "pagination": {
        "count": 0,
        "total_count": 1000
    },
    "items": [
        {
            "agent_version": "1.2.3",
            "app_kind": "unspecified",
            "bytes_returned": 1024,
            "bytes_sent": 4096,
            "chained_subcommand_count": 73,
            "command": "security find-generic-password",
            "corporate_determination_method": "authenticated_user",
            "denial_reason": "Editing the agent's own settings.json is self-modification of agent configuration.",
            "description": "List files in the backup directory",
            "device_name": "janes-macbook-pro",
            "device_serial_number": "C02XK1ZGJGH5",
            "duration_ms": 350,
            "employee_email": "jane.doe@example.com",
            "employee_name": "Jane Doe",
            "error_message": "permission denied",
            "file_path": "/Users/jdoe/.ssh/id_ed25519",
            "from_corporate_domain": true,
            "hook_version": "v1.4.0",
            "host_user_account": "jane@example.com",
            "id": "8CA67511-744C-4D74-B26E-7281CF88712F",
            "interaction_kind": "sensitive_file_read",
            "matched_rule": "~/.ssh",
            "mcp_destination": "unspecified",
            "mcp_destination_path": "/backups/id_ed25519.txt",
            "mcp_server": "dropbox",
            "mcp_server_name": "Dropbox",
            "mcp_server_transport": "stdio",
            "mcp_server_url": "https://mcp.dropbox.com/sse",
            "mdm_device_id": "412",
            "mdm_type": "unspecified",
            "outcome": "success",
            "permission_mode": "default",
            "plugin_name": "anzenna-toolkit",
            "session_id": "0c75df52-2c41-4a8f-bf16-5c1f4e2c9a31",
            "skill_name": "system-health",
            "timestamp": "2026-05-19T15:32:00Z",
            "tool": "create_file",
            "tool_use_id": "toolu_01abc123xyz",
            "username": "jdoe",
            "web_host_name": "claude.ai"
        }
    ]
}

Modified at 2026-09-21 04:45:45

Get a Shadow AI instance by id

Get an AI interaction by id