Samples & groups
Replicates, conditions, and study structure.
Give Claude or ChatGPT the context behind your results. Ask about quality, follow an analysis, find the right files, and move your research forward without retelling the whole experiment.
Preview rollout · Check availability in lab Settings · OAuth sign-in
I checked the recorded trimming summary and comparison for Analysis 07.
The comparison is Treated / Control. Control is the baseline; positive log₂ fold change points toward Treated.
I found the outputs attached to Analysis 07, with the comparison preserved.
Save complete files using the links in your AI app. Inline previews depend on the app.
Here is a starting draft based on the saved analysis, with a few details for you to verify.
Read quality was assessed with fastp. The recorded preprocessing parameters and Salmon reference were used to quantify expression. Differential expression was evaluated with DESeq2 for Treated versus Control.
Before submission: verify software versions, reference annotation, model design, and thresholds against the original records.
Your context, in the AI tools you use.
Connect through custom MCP settings or developer mode where required. Client permissions, OAuth support, and preview access apply.
Less context switching. More scientific context.
Model Context Protocol connects your AI app to NotchBio tools. Instead of pasting disconnected screenshots and remembering every setting, let the assistant retrieve the relevant records from the project you choose.
Replicates, conditions, and study structure.
Recorded read counts and preprocessing settings.
The run and upstream stages behind a result.
Original outputs and supported figure previews.
Numerator and baseline made explicit.
The agent handbook and current app defaults.
From a question to the underlying evidence
“Why did my read count change?” deserves an answer from the trimming record. “Which genes went up?” needs a specific numerator and baseline. “Can I use this figure?” should lead back to the data that created it.
NotchBio MCP helps your assistant carry those details into the conversation, so you can spend less time finding files and more time checking the biology.
Explore the research workspaceTreatment study / Analysis 07
Treated versus ControlPositive log₂FC → Treated · Baseline → Control
Result tables, saved plot data, and run settings
A shorter path to your next question
Review recorded QC and trimming, check run state, and ask the assistant to explain settings in the context of your experiment.
“Summarize quality and trimming for this run. Flag anything I should inspect.”
Locate accessible outputs, inspect supported tables, and bring supported scientific figures into the chat alongside the exact comparison.
“Find this comparison’s results CSV and show a volcano preview.”
Assemble methods notes from saved settings. Ask for missing details, provenance, and a review checklist before you write the paper.
“Draft methods from this analysis. List every detail that still needs verification.”
What’s in this preview: project and run inspection, public archive metadata assessment, files, supported figures, and guidance. Starting or retrying runs, applying parameter changes, and archive imports through MCP are planned. Use the lab workspace to run analyses today.
From results to a research discussion
Retrieve the saved tables and supported figures, then discuss your comparison with the context intact. If your AI app provides a data-analysis environment, use the complete files to create a different plot or investigate a follow-up question.
Running new tests from chat is planned. The current MCP does not start RNA-seq or statistical analysis jobs. Its built-in renderer supports volcano, MA, PCA, and heatmap figures; custom plotting and additional tests depend on your AI app’s own tools.
Your AI app. Your NotchBio account.
When the preview is enabled, setup starts in your lab account. There is no separate data upload for the conversation.
Open MCP settingsOpen Settings → MCP & AI apps. Your server URL appears once setup is enabled.
Follow the instructions for your AI app, paste the server URL, and sign in through NotchBio.
Start with the agent guide and available tools. Name the project, run, and comparison you want to review.
The initial integration is in preview. Open lab Settings → MCP & AI apps to check whether connection setup is enabled. Your verified server URL appears there when available.
The current MCP preview reads and explains existing work. Starting, changing, or retrying analyses and importing archive files through MCP are planned. Run those actions in the NotchBio lab workspace today.
It works with the NotchBio projects your account can access, subject to the permissions granted to the connection. It does not browse your computer or upload local files.
Supported images can appear inline, and complete files use expiring download links. Rendering, attachments, and saving to your device depend on the AI app. A preview is not a substitute for the original file.
Use it to assemble evidence and draft methods from saved records. Check the actual settings, reference, sample design, thresholds, and original results before submission. AI interpretation is not independent biological validation.
Connect the context. Ask better questions. Keep the evidence close.