Skip to main content
Prompt Studio proposes prompts to monitor. These tools start a generation and turn selected suggestions into prompts. Read MCP write access first for plans, scopes, roles and confirmation tokens. Both tools require the write:prompts scope and the Manager role. Examples use synthetic data. Read the results with get_prompt_suggestions and get_prompt_suggestion_run.

generate_prompt_suggestions

Start a Prompt Studio generation. A generation uses no customer credits. It is limited per workspace per day.

Parameters

Within a rolling 24 hours, a workspace can start 3 workspace generations and 10 topic generations. At most one workspace generation and one generation per topic can run at a time. A call for the same running scope is refused. Use a new requestId for each intentional generation, including a retry after a completed generation returned FAILED. Keep the same value when retrying a lost response. Without requestId, identical calls replay the previous result for 24 hours.

Example


accept_prompt_suggestions

Create prompts from up to 20 suggestions of the latest generation. Monitored prompts add the recurring credit use shown in the preview, so every execution needs a confirmation token.

Parameters

Inspect the outcome of every suggestion. A suggestion can be created, already tracked, not found or invalid. Already tracked means that an equivalent prompt is active and monitored. IDs are opaque and bound to the suggestion content and taxonomy. Read them again after regeneration; an old ID cannot accept a different suggestion. Acceptance uses the same import as Prompt Studio. A suggestion equivalent to an archived or paused prompt reactivates that prompt and reports created. Its text, funnel stage and topic are rewritten, its countries are cleared, and its personas and models return to the workspace defaults; its frequency, boost and last collection are kept. Accepting starts no collection itself, but on a LIVE workspace it can make prompts due at the next scheduler pass: a new prompt when automatic answers are on, or a reactivated prompt whose last collection is older than its frequency. That first collection uses credits once, on top of the recurring use. The preview shows it in credits.oneOff and the recurring change in credits.dailyChange. Execution refuses the confirmation when either amount changed since the preview.

Example