The AI project partner inside your notebook
Runcell is the Jupyter-native AI agent for domain experts running real, multi-week ML and data projects.
It reads your notebook and its outputs, remembers where your project left off, and runs long tasks end to end — not just snippets of code.
Free to start · runs in your existing JupyterLab · download the extension
Where did we leave off?
Tuesday we shipped v3 of the default-risk model — AUC 0.81 on holdout. You flagged PSI drift in bureau_score and wanted to try monotonic constraints next. Want me to set that up?
Your project doesn't fit in one prompt.
A real model, paper, or pipeline takes weeks, dozens of notebooks, and hundreds of decisions. Generic AI tools treat every message like a fresh start — so you become the glue.
They forget everything
Every session you re-explain the dataset, the decisions, and where the project left off. The tool never remembers — you do.
They can't see your results
Most assistants read your code but not the loss curve, the confusion matrix, or the figure your cell just produced. They debug blind.
They give up halfway
A multi-hour training run or a thousand-step analysis is exactly where chat tools time out and lose the thread.
Runcell works at the project level — it remembers your project, reads your outputs, and runs long tasks end to end, inside the JupyterLab you already use.
Code editors finish your lines. Runcell carries your project.
Cursor, Copilot, and Continue are code editors with AI — great at the next line and the next file. A multi-week ML project is a different job: a sequence of experiments, results, and decisions that has to hold together over time. That's what Runcell is built for.
Remembers your project
Cross-session memory. Pick up days later and ask "what did we do so far?" — Runcell knows the dataset, the decisions, and the state of the work.
Goes the distance
Long-running tasks that span hours or days. Runcell runs thousands of steps end to end without losing the thread.
Reads your results
It sees the charts, plots, and image outputs your cells produce — reasoning about real outcomes like a diverging loss curve, not just your code.
Lives in your notebook
Runs inside JupyterLab on your existing .ipynb files. No new editor, no copy-paste.
From a goal to a finished project
Hand Runcell an outcome and it plans the steps, writes the code, runs the cells, reads the results, and keeps going — remembering the project as it works, inside the JupyterLab you already use.
Give it a goal, it runs the notebook
Describe the outcome you want. Runcell plans the steps, writes the code, executes the cells, and fixes errors as it goes — an autonomous agent working end to end inside your notebook.
Multi-step workflows, executed for you
Ask about any cell, apply the fix in one click
Select a cell and ask what is wrong or what comes next. Runcell reads the surrounding code and outputs, generates the fix, and recommends your next step — all without leaving the notebook.
AI that understands the cells around it
Understand unfamiliar code with live examples
Hit an algorithm or library you do not know well? Runcell explains it with runnable examples right in your notebook — like comparing K-means vs DBSCAN on real data instead of digging through docs.
Concepts explained with runnable cells
You bring the domain, the agent writes the code
For domain experts who think in problems, not Python. Describe the analysis or figure you need — like a genomics circos plot — and Runcell handles the implementation.
Research-grade code, minus the boilerplate
Built for the work you actually ship.
The deepest Runcell users aren't hobbyists learning Python — they're domain experts shipping real projects.
field 01
Risk & fintech modeling
Iterate dozens of model versions, read PSI/IV and confusion matrices, and ship to production.
field 02
Academic research
Take a study from raw data to publication-ready figures across one long project.
field 03
Supply chain & forecasting
Keep multi-notebook demand and commodity models consistent over months.
field 04
Medical & life-science ML
Train and evaluate on small, sensitive datasets with the rigor your field demands.
What is Runcell?
Runcell is a Jupyter-native AI project partner for domain experts running real, multi-week ML and data projects. It lives inside JupyterLab as a lightweight extension — writing and debugging your Python, executing cells, and reading the charts and outputs your code produces. Unlike chat tools bolted onto a notebook, Runcell remembers your project across sessions and runs long tasks that span days, so it carries the whole project with you instead of answering one prompt at a time.
what runcell can do
It remembers your project
Cross-session memory keeps your dataset, decisions, and progress in context. Pick up a multi-week project days later and ask “what did we do so far?” — Runcell knows.
explore the features or the documentation
It sees your outputs
Runcell reads the charts, plots, and image outputs your cells produce — so it reasons about real results, like a diverging loss curve, instead of guessing from your code alone.
It runs the long jobs
Multi-hour training runs and thousand-step analyses. Runcell executes long tasks end to end without losing the thread or timing out halfway through.
Key capabilities
everything you need to run a project, not just write code
how it compares
Code editors vs. a project partner
vs. Cursor & AI IDEs
Cursor is an excellent code editor, but it is built around files and the next line — it pulls you out of Jupyter and does not carry a multi-week project. Runcell is notebook-native and built for the project, not the snippet.
vs. Copilot autocomplete
Autocomplete guesses the next line. Runcell plans a multi-step analysis, runs it, reads the results, remembers what happened, and iterates — a project partner, not code completion.
vs. notebook AI chat
Chat bolted onto a notebook answers questions about your code. Runcell acts on the notebook — running cells, reading outputs, remembering the project, and moving the work forward.
faq
Frequently asked questions
Start your next project with Runcell
Install in one line, sign in, and put a project partner inside the JupyterLab you already use — one that remembers your work and runs the long jobs with you.
free to start · runs in your JupyterLab · no API key required