Use smart parameters

Turn any question you have about your mentions into an AI-filled column that runs your prompt against every post and stores the answer.

A Smart Parameter turns any question you have about your mentions into a column that AI fills in for you. Your prompt runs against every mention in scope and its answer becomes the column's value — so you can tag, classify, or extract one thing across the whole collection without reading each post by hand. If you run an e-bike DTC brand, you might add a column for buyer intent, the pain point each post raises, or whether it names a competitor, and have it answered for thousands of mentions at once.

Prerequisites

A dataset with collected mentions to run against — collect one first (see Create a mention collection).
Available credits. Running costs 0.5 credits per mention; the test-run preview is free.

In the app this feature is called Smart Parameter. In the API and MCP the same entity is named custom_parameter (with instructions and value_type fields). The walkthrough below uses the in-app names.

Walkthrough

Add column

Open the Library at /library, or open a dataset table at /datasets/[id]. In the column header, click the + Add column button to open the column-settings popover. Under Add a property, click Smart Parameter (labelled AI auto-fill column) to open the New Smart Parameter modal.

Configure

This is phase one of three. Optionally pick a starting point under Start from preset — choose Custom to Start blank, or use one of the named presets (Topic, Buyer intent, Pain points, Competitor mentions, Content type, Language, Action summary).

Then fill in the fields. Enter a Name (the placeholder shows e.g. Buyer intent). Write your Prompt template — the prompt is sent to the AI for each mention. Optionally add an Expected output, and choose an Output type: Text, List, or Object.

For the e-bike brand you might name the column Buyer intent, set the output type to Text, and use a prompt template like: classify whether this mention is from someone researching a purchase, comparing models, or just discussing e-bikes generally.

Test run

In the Test run section, click Run test to preview results on the first 3 visible mentions. This is free and uses no credits, so iterate on your prompt here before committing to a full run. Each row shows Thinking… while it works, then the AI's Output.

Save

Click Save to create the parameter. The modal stays open and switches to edit mode for the parameter you just created — saving never closes the modal by design. Close it later with Cancel, the X, or Escape.

Scope

This is phase two. Under Scope, pick which mentions to analyse — your current saved View or the dataset. The helper text reads: "Choose which mentions to analyse. The run starts after you click Run." A saved View is required to run; if there is no scope context, the picker and Run buttons are hidden and a note tells you to save a View first.

Confirm & run

This is phase three. Review the mention count and the credit estimate, then click Run. If you have not picked a scope yet you will see "Pick a scope above to see the credit estimate."

Watch the column fill in

After the run starts, values populate per mention over time. Cells show — for mentions not yet run, Running… while in progress, the AI value when done, and Error on failure. There is no inline retry. Mentions already analysed are skipped automatically if you re-run.

Re-edit or run again later

Use the column kebab menu to manage the parameter. Edit prompt & test run… reopens the modal, Hide column / Show column toggle visibility, and Delete column removes it — note that deleting the column stops AI computations. Run now computes the parameter against the selected scope, skipping mentions already analysed.

Now use the column

Once a column is filled in, it behaves like any other column in the mentions table — and it travels with your data:

Filter and sort by it in a saved View. Combine a Smart Parameter with sentiment, platform, or posting date to slice mentions exactly the way you need.
Export it. When you select rows and Export to CSV, the file (library-<date>.csv) includes every Smart Parameter — regardless of which columns are currently toggled on — alongside the standard mention fields. CSV export is a Pro+ feature.
Feed it into analysis. Smart Parameters create the enriched columns that many skills and insights read from, so a buyer-intent or pain-point column makes the next analysis sharper.

Gotchas

Three stages, and only one costs credits. Create, then Test run (a free, stateless preview on the first 3 mentions), then run on your collection — only the collection run charges credits.
Cost is 0.5 credits per mention. The estimate is count × 0.5 rounded up. The numbers shown in the app are estimates only; the API charges the precise amount when it runs.
Per-run cap of 5,000 mentions. Larger scopes are sliced: "Up to 5,000 mentions will be processed in this run. Re-run after completion to continue with the next batch."
A saved View or dataset is required to run. With no scope context the picker and Run buttons are hidden, and a note tells you to save a View first.
Save never closes the modal. Dismiss it with Cancel, the X, or Escape. Only a run-only flow closes on a successful run.
Run is disabled while the form is unsaved or dirty. You will see "Save first to run" before the parameter exists, and "Save changes first" while you have unsaved edits.
Credits gate the run. With a zero balance you see "Top up to run a Smart Parameter — you have 0 credits." A nonzero but insufficient balance still runs the first affordable batch.
Values fill in asynchronously. Cell statuses move from — to Running… to a value, or Error on failure (including rows stuck over 60 minutes). There is no inline retry.
Re-runs skip analysed mentions. A re-run may report nothing to do with "All mentions already analyzed", or block if a run is already in progress.
Duplicate names are rejected by the API, and an empty name surfaces inline under the Name field.
Chat can suggest one too. The assistant may emit a Suggested Smart Parameter card with Edit / Confirm and a Run on dataset picker (or Skip — just save); confirming creates the parameter and can queue a run.