edit_test

Apply changes to an existing test.

Apply changes to an existing test.

Destructive · Writes data · Idempotent · Calls external systems. This tool can modify or delete data. Review the arguments before letting an agent call it.

Parameters

FieldTypeRequiredDescription
acknowledgebooleanNoSet to true after user approves continuing past the edit limit. NEVER set without explicit user approval.
actionsstringNoJSON array of actions (replaces current actions)
archivedbooleanNoFlip the archived flag
auto_correct_kindsbooleanNoRewrite declared input_variable kinds to match the platform's variable types. Default false: kind mismatches are hard errors.
descriptionstringNoNew test description (only if explicitly changing it)
is_favoritebooleanNoFlip the is_favorite flag
namestringNoNew test name (only if explicitly changing it)
skip_validationbooleanNoBypass the client-side validator (the platform still validates). Use for action types the CLI doesn't recognise or top-level metadata edits.
statusstringNoChange test status (e.g. DRA, APP)
test_idnumberYesID of the test to edit
variablesstringNoJSON variables object (replaces current variables)
variantsarrayNoVariant pin specifiers for V3 bucket resolution. Each item is '5' (numeric ID) or 'API:Staging' (group:variant by name). Edit mode: groups not pinned retain their existing buckets.

Example arguments

Illustrative arguments an agent supplies when calling this tool:

{
  "test_id": 0
}

Description

DESTRUCTIVE — creates a new version automatically.

PREREQUISITE: call platform_guide topic=test_editing for the guardrailed workflow and protected fields.
ALWAYS preview first with preview_test_edit to check impact.
Protected fields: product and collection_uid are NEVER changed. Name and description only change when explicitly provided.
A per-test edit limit is enforced (5). Check edits_remaining in the response.

Top-level metadata that's now always preserved on PUT and editable via this tool: archived, is_favorite, status. Pass archived=true to archive the test without touching actions; use manage_test_lifecycle action=archive (or action=unarchive) for the common case.

skip_validation=true bypasses the client-side validator. Use only when the test has an action type this MCP server's registry doesn't know (future types), or when you're only flipping a top-level flag like archived. The platform still validates.

What's next?

All MCP tools

Browse the full tool reference by category.

Connect a client

Point Claude, Cursor, or Claude Code at the server.

Last updated on