Batch

Classify a batch of messages

Import a UTF-8 CSV, TXT or JSONL file up to 2 MB, or paste one message per line: 1,000 rows per batch and 3–6,000 characters per message. Each completed row uses one credit.

  1. Import messages
  2. Choose rules
  3. Run and export

Import messages

Preview

Rows are parsed in your browser: only the row you run, with its labels and threshold, is sent to the API.

Labels and review rules

How each row is judged

Send one label decision per row, or up to four typed questions answered together in one request. Only what these fields show is sent.

Start from a rules preset

A preset fills the instructions and labels below. Edit either field before you run: only what those two fields show is sent.

Route each message to the team that owns it.

Decision instructions Choose the team responsible for resolving the main customer issue.

Allowed labels, separated by commas billing, technical support, account access, feedback, other

Review threshold 0.85

Sort each message by how the writer feels.

Decision instructions Judge the overall sentiment of the writer towards the product or the service, not the topic they mention.

Allowed labels, separated by commas positive, neutral, negative, mixed, unclear

Review threshold 0.8

Flag comments a human moderator should review.

Decision instructions Choose the label that best describes the comment. Use "unclear" when the text cannot be judged, and never infer intent that is not written.

Allowed labels, separated by commas acceptable, insult, harassment, spam, unclear

Review threshold 0.9

Multi-question preset

A preset fills the questions below with typed questions a visitor can inspect and edit before running.

Route each message, flag defects and set urgency in one request.

  1. team choice

    Which team should own this ticket?

    • billing Invoice, refund, plan or payment issue.
    • technical Broken or unexpected product behaviour.
    • account Sign-in, permissions or profile issue.
    • other Anything that does not fit the three above.
  2. is_bug noul

    Is the writer reporting a defect in the product?

    • true They describe broken or unexpected behaviour.
    • false They ask a question or request a feature; nothing is broken.
  3. urgency score

    How urgently does this need an answer?

    • 0 Can wait for the next release.
    • 1 Should be answered this week.
    • 2 The customer cannot work right now.

Review threshold 0.85

Generate the rules with AI

Describe the decision you need. The model proposes the typed questions, and you review and edit them before anything is sent.

Generation costs 1 credit, and a failed generation is refunded. It is not a decision: each row still costs 1 credit, and nothing runs until you press Run.

Keep an "other" or "unclear" label, and review a low-confidence row before acting on it.

One completed row uses one credit. Running out of credits stops the batch, and nothing is purchased automatically. Keep this tab open and export before you leave.

Guests can run 1 row per source address per UTC day. New accounts get 20 credits; one completed row uses one credit, and nothing is purchased automatically. Sign in and keep this batch · Pricing

Run and export

Your decision

DAILY BONUS

2 free credits, every day

Claim 2 free credits each UTC day — 14 across seven claims. No purchase needed. Credits do not expire.

    Resets at 00:00 UTC. No streak required.

    Try a decision →

    Take the next decision into your workspace.

    1 anonymous attempt per day · 20 signup credits · No card for the trial

    Sign in ↗

    Tell us what you need

    Describe the workflow you want to classify, the volume you expect, and the outcome you need. We reply by email.

    10–2,000 characters. Never include passwords, API keys, or payment details.