Practical guide
How to Apply AI to Every Row in a CSV
A practical, test-first workflow for enriching, classifying, summarizing, or transforming every row in a CSV with an AI model.
Read guideNow reviewing design-partner applications
Apply AI to Every Row — Without Building Another Batch Script
BatchForge helps operations, data, and content teams turn CSV or JSON files into repeatable AI workflows. Test representative rows, inspect the output and cost, then approve the full run.
Test before spendingRow-level recoveryCSV & JSON exports
See how it works →Bring one repetitive spreadsheet workflow. Approved partners start with a real 10-row test.
Built to reduce batch-processing risk
Representative sample first
Selective row retries
Cost review before full run
7.3M+ Rows processed
3.6B+ Tokens processed
The Test-First BatchForge Process
Upload, define, test, then run only when the sample is good enough.
1
Upload Data
Upload your CSV or JSON file. We automatically parse columns and make them available as template variables.
2
Define Prompts
Create prompt templates using {{variable}} placeholders. Preview rendered prompts before running the batch.
3
Test a Sample
Run representative rows first so you can review quality and estimated cost before committing the full file.
4
Run & Export
Confirm the full run, follow progress, then export completed results as CSV or JSON.
See Your Jobs in Action
Track every job with real-time progress updates, token counts, and cost tracking.
batchforge.io/dashboard/jobs
Real-time progress tracking for every job
Choose the Right Run Mode
Use immediate mode for fast feedback or eligible batch execution when cost matters more than turnaround.
Immediate Mode
Start processing promptly and follow row-level progress while the run is active.
Best for tests and time-sensitive work
Batch Mode
Queue eligible provider work for lower-cost asynchronous processing when turnaround is flexible.
Up to 50% lower platform-funded model cost when eligible
Life Before and After BatchForge
See what changes when you stop doing data processing the hard way.
Without BatchForge
- Write and debug Python scripts for every new LLM task
- Manually manage API keys, rate limits, and retries
- Wait hours with no visibility into progress or costs
- Export results by hand, hoping nothing broke mid-run
With BatchForge
- Upload a CSV and define a prompt — no code needed
- Choose among supported LLM providers per job, with API keys encrypted
- Watch every row process live with real-time cost tracking
- Export structured results as CSV or JSON in one click
Everything You Need
A complete toolkit for batch LLM processing, from data upload to result export.
Multi-Provider LLMs
Choose supported models from OpenAI, Anthropic, Google, Mistral, DeepSeek, or xAI per job.
CSV & JSON Upload
Upload structured data in CSV or JSON format. Columns are parsed automatically as template variables.
Template Variables
Use {{variable}} placeholders in prompts. Preview rendered output before running your batch.
Real-Time Progress
Watch your batch process row by row with live progress updates, token counts, and cost tracking.
BYOK Support
Bring your own API keys for direct billing, or use BatchForge-managed processing.
Export Results
Export completed batch results as CSV or JSON. Download with one click when processing finishes.
What Changes for Your Team
Practical transformations supported by the current product, without requiring web research or unreleased integrations.
Data Enrichment
Classify or enrich lead records using the context already present in your file.
Add normalized industry and company-size fields
Content Generation
Generate consistent draft copy from the product attributes already present in each row.
Draft product descriptions from structured specs
Classification & Tagging
Apply one approved taxonomy across support tickets, products, or survey responses.
Apply a consistent taxonomy to support tickets
Data Extraction
Turn text already present in spreadsheet rows into structured fields your systems can use.
Extract dates, names, and categories from row text
Sentiment Analysis
Apply a consistent sentiment rubric to reviews, surveys, or NPS responses already in your file.
Score and summarize customer-review themes
Data Normalization
Standardize inconsistent names, labels, formats, and free-text values across a file.
Normalize messy categories into an approved schema
Bring us a repetitive workflow
We prioritize design partners with a specific file, prompt, and repeatable business outcome.
A focused design-partner program
Is BatchForge the Right Fit?
We are proving one workflow exceptionally well before opening broader automation and collaboration features.
A strong fit today
- You repeat the same AI prompt across rows in a CSV or JSON file
- A human can review sample outputs before the full run
- You need structured results, row status, retries, and exports
- The workflow repeats often enough to save and version
Not available yet
- The job depends on live web research or enrichment
- Multiple teammates need shared roles and approvals
- You need a public API, Zapier, or Google Sheets automation
- You want an autonomous agent with no review checkpoint
Secure Transport
HTTPS protects data in transit
Encrypted API Keys
BYOK credentials use AES-256-GCM
Deletion Controls
Delete source files and job data from the app
Service Readiness
App, worker, queue, storage, and database checks
Batch AI guides
Plan the workflow before you process the file.
Practical answers for preparing data, testing prompts, reviewing cost, and recovering failed rows.
Workflow design
Batch LLM Processing: A Test-First Workflow
Learn how batch LLM processing works and why sampling, structured outputs, cost review, and selective retries should happen before a full run.
Read guideData preparation
CSV vs. JSON for Batch AI Processing
Choose between CSV and JSON for batch AI jobs by comparing structure, nested data, validation, editing, and export requirements.
Read guideSimple, Credit-Based Pricing
Start small, then scale as you grow. All plans include access to every LLM provider.
Free
Try it out with no commitment
$0/mo
- 50 credits/mo
- Usage varies by model and token volume
- Free-tier models
- CSV/JSON export
Starter
For individuals getting started
$9.99/mo
- 1,000 credits/mo
- Usage varies by model and token volume
- All LLM providers
- CSV/JSON export
Recommended
Pro
For professionals and power users
$49.99/mo
- 5,000 credits/mo
- Usage varies by model and token volume
- All LLM providers
- BYOK support
- Priority processing
Business
For higher-volume individual workspaces
$299.99/mo
- 30,000 credits/mo
- Usage varies by model and token volume
- All LLM providers
- BYOK support
- Priority processing
- Dedicated support
Frequently Asked Questions
Everything you need to know about BatchForge.
Ready to Get Hours Back Every Week?
Request design-partner access to test a representative sample, review the result and estimated cost, then decide whether to run the full file.