generated: '2026-08-23'
method: searched
source: https://docs.kore.ai/agent-platform/testing
note: >-
  Kore.ai publishes NO public sandbox in the payments sense — there are no test API keys,
  no test-vs-live key prefixes, no magic identifiers, no hosted test tokens and no time
  simulation. What it does publish is a substantial IN-PRODUCT rehearsal layer behind the
  login wall: a Studio test console, a scripted evaluation harness, a debug/trace surface,
  and package validation with an import preview. All of it requires an authenticated
  workspace, so an unauthenticated consumer cannot exercise the API at all before signing
  up. No test values are recorded here because the provider publishes none.
public_sandbox: false
test_credentials_published: false
key_prefixes:
  note: >-
    Key prefixes exist but denote SCOPE, not test-vs-live — abl_ (long-lived API key),
    ak_ (service-to-service), pk_ (origin-restricted public widget key), sdk_ (short-lived
    SDK session token). There is no test-mode prefix.
environments:
  model: named-environment-segment
  note: >-
    Runtime routes carry an {env} path segment
    (/api/v1/project/{projectSlug}/{env}/agent/{agentSlug}/chat), so a project can be
    exercised in a non-production environment, and deployments are typed and promotable
    between them. The environment names are customer-defined, not a fixed sandbox/live pair.
facilities:
- name: Studio test console
  description: Interactive testing of agents from within the platform UI.
  docs: https://docs.kore.ai/agent-platform/testing
  access: authenticated workspace
- name: Evaluations
  description: >-
    Generate eval personas and scenarios, define evaluators, assemble eval sets and run them
    with preflight, per-case results and heatmaps. Drivable from the UI or over MCP.
  docs: https://docs.kore.ai/agent-platform/evals
  mcp_tools: [platform_eval_personas, platform_eval_scenarios, platform_eval_evaluators, platform_eval_sets, platform_eval_runs]
- name: Debug sessions
  description: >-
    Load an agent into a debug session, send messages, inspect gather progress and flow
    state, search traces, view the span tree and ask the platform to explain a decision.
  docs: https://docs.kore.ai/agent-platform/runtime
  mcp_tools: [debug_load_agent, debug_send_message, debug_get_current_state, debug_traces, debug_get_span_tree, debug_explain_decision]
- name: Package validation with import preview
  description: Validate a local package before importing it and preview what the import would change.
  mcp_tools: [platform_validate_package, platform_package_model]
- name: Batch and conversation test suites (AI for Service)
  description: >-
    The older estate ships importable/exportable batch test suites and conversation test
    suites with asynchronous execution-status APIs.
  docs: https://docs.kore.ai/ai-for-service/apis/automation/api-list
signup:
  free_signup_claimed: true
  claim_source: https://docs.kore.ai/agent-platform/getting-started
  claim_text: 'Agent Platform Account — Sign up for free using your region-specific URL.'
  probed_url: https://agents.kore.ai/signup
  probed_status: 200
  probed_note: Redirects to https://agents.kore.ai/auth/login; the signup form itself is behind that wall.
ip_allowlisting:
  note: >-
    The platform sends requests from a fixed set of region-specific egress IPs, published so
    customers can allowlist them. That list is at
    https://docs.kore.ai/agent-platform/getting-started#platform-access-and-allowlist-of-ips