Agent Computers
All Agents have multiple ways to spin up their own computers.
Running Python With an Agent
Last Updated: 2026.07.07
Overview
Some Agents have dedicated and powerful computers always attached to them (e.g. the Data Analyst Agent, the Sapience Coding Agent, the Claude Code Agent).
But, EVERY Agent in Sapience can spin up a cloud server on-demand to work in.
That means that when YOU use Sapience, you are getting multiple servers to power your work all built-in. Its crazy if you think about it, but that’s the power of Sapience.
Sometimes a question is really a small computing task — crunch these numbers, reshape this spreadsheet, generate a chart, parse this data. Sapience agents can write and run Python for you to do exactly that, on Sapience's own infrastructure, and bring the results back into your chat. You don't write any code — you just ask.
What It Does
When your request needs real computation, the agent can:
- Write the Python, Bash, C, C#, Javascript/Typescript needed to do the job.
- Run it in a secure, isolated sandbox — optionally handing it your Sapience files (by reference) or some data to work on.
- Bring back whatever it produced — any files the code generates (a chart, a cleaned spreadsheet, an output document) are harvested straight back into your conversation.
- Fun fact: Note that AI Agents gravitate to Python, because most LLMs are trained on far more python code than any other language.
This runs on Sapience's own infrastructure, and the capability is available to every agent — so you don't need a special "coding" agent to get it.
When It Helps
Reach for it (or just ask, and the agent will reach for it) when your task involves:
- Calculations — totals, growth rates, statistics, projections.
- Data wrangling — cleaning, filtering, reshaping, or combining tabular data.
- Generating output from data — a chart, an HTML table, a reformatted file.
- Working over a file you already have — point the agent at a Sapience file and ask it to process the contents.
You describe the outcome you want; the agent handles the code.
How You Use It
Just ask in plain language. For example:
"Take the sales spreadsheet I uploaded and give me total revenue by region as a chart.""Clean up this CSV — drop the blank rows and standardise the date column.""Work out the compound growth rate across these figures."
The agent writes the Python, runs it against your data, and returns the answer plus any files it created.
What Happens Behind the Scenes
- The agent writes a small Python program for your task.
- Sapience runs it in a fresh, isolated sandbox, giving it any files or data you referenced.
- Anything the program writes out — charts, spreadsheets, documents — is collected and returned into your conversation automatically.
The sandbox is isolated per run, so your code and data aren't mixed with anyone else's.
Good to Know
- No coding required from you. You describe the result; the agent writes and runs the code.
- It can work on your files. Hand it a Sapience file to process, not just inline data.
- You get the outputs back. Generated files come straight into the chat — no separate download step to fetch them from somewhere else.
- Available on every agent. This isn't limited to a special coding agent.
Summary
Sapience agents can write and run Python and other languages for you to handle real computation — calculations, data cleaning, chart generation, processing your files — all on Sapience's infrastructure, on any agent. Just describe what you want; the agent writes the code, runs it in a secure sandbox, and brings the results (including any generated files) back into your conversation.