One username and password. Here is what it opens, and how to set up your own computer if you are going to run simulations.
Browse, read and download the whole collection.
Where simulation runs, the settings they used and the results they produced are all recorded.
A persistent notebook and kernel next to the data. Choose the smallest profile that fits; for large-array postprocessing, High memory is the usual starting point.
Temporary passwords expire after 7 days. If yours has lapsed, email support@ergodic.io and we will send a new one.
Keep https://continuum.ergodic.io/login bookmarked alongside the other two. It is where your password gets changed, and where Forgot your password? lives if you lose it. The library offers you that same sign-in page, so it works from there too. The experiment manager cannot change a password, only check one, so come back here when you want to.
To sign out of everything in one go, go to https://continuum.ergodic.io/logout. Closing the tab does not do it, which makes this the one worth using on a shared or lab computer.
About the library: it reads in a browser and downloads to your device. E-reader apps cannot connect to it directly.
Only needed if you are going to run them, rather than only look at other people’s. Two things to set up on your own computer, once each.
Your machine needs to sign in to the experiment manager without you there to do it. Add these
three lines to the end of your shell profile — the file is ~/.zshrc on a Mac,
~/.bashrc on Linux and on NERSC:
export MLFLOW_TRACKING_URI=https://continuum.ergodic.io/experiments/
export MLFLOW_TRACKING_USERNAME=your-username
export MLFLOW_TRACKING_PASSWORD='your-password'
Use your own username and password. Keep the single quotes around the password — it contains a symbol, and a few symbols mean something to the terminal if they are left bare. Then open a new terminal window, or the lines will not have taken effect yet.
If you ever change your password, change it here too. This copy does not follow along, and a run will stop being able to sign in until it matches.
Without these three, a run stops immediately and says it cannot reach the experiment manager.
Your invitation includes a private, per-person token for the Jupyter MCP endpoint. Store it in
your MCP client’s private configuration, never in a repository. Start your Jupyter server at
https://jupyter.ergodic.io before connecting the agent. The endpoint is
https://jupyter.ergodic.io/user/your-username/mcp.
The connector runs inside the notebook image, pinned to jupyter-mcp-server 1.3.8.
An idle server is culled after an hour and every server has a hard age ceiling.
The files a run produces — plots, data, checkpoints — are written from your own machine straight to our cloud storage, so your computer needs a key for it. Your key came with your invitation email; if you do not have one, email support@ergodic.io.
Save it in a file named exactly credentials, with no file extension, in a folder
named .aws in your home folder:
~/.aws/credentialsC:\Users\<your name>\.aws\credentialsCreate the .aws folder if it does not exist. On Windows, make sure Notepad has not
added .txt to the end of the file name. The file holds exactly this:
[default]
aws_access_key_id = your-access-key-id
aws_secret_access_key = your-secret-access-key
region = us-east-1
Nothing else needs setting up — the tools find that file by themselves. If it is missing, a run still records its settings and its numbers, and then fails when it tries to upload the files it produced.
The key reaches our results storage and nothing else, and it cannot delete anything.