Understanding Credits
Credits are the currency you use to run analyses on Pipette.bio. This guide explains how they work and how to get the most value from them.
What Are Credits?
Think of credits as a combined measure of the computational resources your analysis consumes. When you run an analysis, we track two things:
- AI tokens — the language model reasoning that powers our intelligent agent
- Compute time — the actual processing time for running bioinformatics tools
Credits capture both of these costs in a single, easy-to-understand unit.
What You're Actually Paying For
AI Reasoning (Tokens)
Our agent uses Claude, a state-of-the-art language model, together with Skills—specialized bioinformatics modules designed to follow the most advanced tools and best practices in the literature. The agent orchestrates these Skills to:
- Understand your research question
- Plan the analysis pipeline
- Execute and adapt to results in real-time
- Generate comprehensive reports with citations
Tokens measure the "thinking" the AI does—both reading your data context and generating responses.
Compute Processing
This covers the actual bioinformatics work:
- Running alignment tools (Salmon, STAR, BWA)
- Quality control analysis (FastQC, MultiQC)
- Statistical analysis (DESeq2, edgeR)
- Single-cell analysis (Scanpy, Seurat)
- Any custom scripts the agent writes and executes
Reporting
Pipette always generates visualizations and plots for each analysis by default—these are included with every run. The reporting agent reads these outputs and compiles them into a detailed, publication-quality report with QC summaries, methodology descriptions, and result interpretation.
Important: Report generation can use 20-30% of the total credit cost per query. If you're doing exploratory work or just need the raw outputs (count matrices, DEG tables, plots), turn off reporting before running your job to save credits. Save detailed reports for your final analysis.
Why Costs Don't Scale Linearly with Samples
Here's something that surprises many users: AI token costs are largely independent of your sample count.
The AI Overhead Is Fixed
Whether you're analyzing 3 samples or 30, the agent performs similar reasoning:
- Parse the experimental design
- Select appropriate tools and parameters
- Monitor quality metrics
- Interpret results and write the report
A 10-sample RNA-seq experiment doesn't require 10x the AI reasoning of a 1-sample analysis. The agent thinks through the pipeline once and applies it across all your samples.
What Actually Scales
- Compute time scales with data volume (more samples = longer runtime)
- Token usage stays relatively flat regardless of sample count
This means batching samples together is significantly more economical than running them individually.
The Power of Batching
When you submit multiple samples that need the same pipeline:
| Approach | AI Overhead | Compute | Total Cost |
|---|---|---|---|
| 10 separate runs | 10x reasoning cycles | 10 separate jobs | High |
| 1 batched run | 1x reasoning cycle | 1 parallelized job | Lower |
Example
Running 8 RNA-seq samples as a single batch might cost 25-30 credits total. Running them as 8 individual samples could cost 20+ credits each—paying the AI reasoning overhead eight times.
Batching Best Practices
- Group samples by experiment — samples with the same reference genome, annotation, and analysis goals should run together
- Include your full sample sheet — let the agent process all conditions and replicates in one go
When to Run Separately
- Different species or reference genomes
- Fundamentally different analysis types (RNA-seq vs. variant calling)
- Exploratory runs where you're testing parameters on a subset first
Credit Packs and Pricing
Every month to explore the platform. Resets at the start of each calendar month.
Need more? Purchase credit packs that never expire:
| Pack | Credits | Price | Per Credit |
|---|---|---|---|
| Starter | 20 | $10 | $0.50 |
| Standard | 55 | $25 | $0.45 |
| Pro | 120 | $50 | $0.42 |
| Power | 250 | $100 | $0.40 |
Larger packs offer better value per credit. The Pro pack hits the sweet spot for most active researchers.
How Credits Are Used
- Free credits first — your monthly allowance is always used before purchased credits
- Purchased credits never expire — buy with confidence; they stay in your account until used
- Refunds on failures — if an analysis fails due to a system error, credits are automatically refunded
Estimating Credit Costs
Typical credit usage by analysis type:
| Analysis Type | Typical Credits | Notes (incl. reports) |
|---|---|---|
| RNA-seq (bulk, 8 PE samples / 16 fastq files) | 30-35 | ~55 mins |
| Single-cell (68K cells PBMC) | 25-30 | h5 to marker genes |
| Variant calling | — | Depends on VCF file and type of analysis |
| Clinical variant analysis | 20 | ClinVar, GenoMad |
These estimates assume standard input formats (FASTQ, H5, VCF, etc.) and include report generation. Actual costs depend on data size, analysis complexity, and how much back-and-forth reasoning is needed. Turn off reports to reduce costs by 20-30%.
Real-World Examples
These examples include full report generation. Turning off reports before running can reduce costs by 20-30%.
Bulk RNA-seq: FASTQ to DEGs (Salmon pipeline)
8 samples (16 paired-end FASTQ files), ~20 GB input data
Trim Galore → Salmon → DESeq2
~27 credits, ~55 minutes
Bulk RNA-seq: FASTQ to DEGs (STAR pipeline)
8 samples (16 paired-end FASTQ files), ~20 GB input data
Trim Galore → STAR → featureCounts → DESeq2
~37 credits, ~90 minutes
Single-Cell RNA-seq: H5 to Marker Genes
68,000 cells from H5 input
QC → Normalization → Clustering → Marker Gene Detection
~25 credits, ~15 mins
AI costs do not scale with cell count. Compute time increases with more cells, but the orchestration cost stays fixed.
Tips for Optimizing Credit Usage
Do
- Batch related samples together — process entire experiments in one submission
- Turn off reports for exploration — reports use 20-30% of credits; save them for final runs
- Provide clear instructions — ambiguous requests require more AI reasoning
- Include sample metadata upfront — condition labels, batch info, comparisons
- Start with a pilot — test your pipeline on 2-3 samples first
Avoid
- Running samples one at a time — this multiplies the AI overhead
- Leaving reports on for exploratory runs — turn them off before rerunning jobs
- Vague requests — "Analyze this data" costs more than specific instructions
- Unnecessary re-runs — download and save your results
Checking Your Credit Balance
View your current credit status anytime from your account dashboard. You'll see:
- Free credits remaining this month
- Purchased credit balance
- Usage history
Key Takeaways
| Concept | What It Means |
|---|---|
| Credits combine token + compute costs | You pay for AI reasoning and processing time |
| AI costs don't scale with samples | The agent thinks once, executes many |
| Batching is cheaper | Group samples to share AI overhead |
| Reports use 20-30% of credits | Turn off reports for exploratory runs |
| Free credits reset monthly | 20 credits/month to explore |
| Purchased credits never expire | Buy packs as needed |
Questions?
If your analysis used more credits than expected, or you're unsure how to structure your submission for best value, reach out to our support team. We're happy to help you get the most from your credits.
Ready to start?
Your dashboard shows the estimated credit cost before each run.
Start Free View Pricing