Available today

Getting in safely

• One login. You sign in with the NASA identity you already have. There is no extra platform password to remember.
• Strong login. Multi-factor and passwordless sign-in. (Multi-factor means a second proof, such as a code or a badge, on top of your password.)
• Partners can join. Collaborators from universities or other agencies can be given access. Their institution can also connect its own login system.
• Your data stays yours. The platform gives your notebook access to only the data you and your project may see. You do not handle any keys or secrets.
• Encryption. Your data is encrypted when stored and when moving.
• Web firewall. A managed firewall filters every web request to the platform.
• A real user account. You get your own username and your own file ownership, so shared files behave the way you expect.
• Reach other cloud accounts. A project admin can register a cloud role, and your notebook gets it at start-up. This lets you read data in your institution's or a mission's account.

Getting a machine that fits the work

• Start a workspace on demand. No ticket, no waiting.
• Pick the size. From a small 4 GB workspace up to a 64 GB / 8-CPU machine, or a GPU machine. (A GPU is a chip that is fast for AI and image work.)
• Software is ready. Earth-science, Python, and AI software stacks are pre-installed. Power users can bring their own container image. (A container image is a packaged set of software.)
• Many machines for one job. Spread a large analysis over many machines when you need to.
• Background jobs. Send a long notebook to run on its own. Your session stays free, and the job keeps running after you close the browser.
• Capacity follows demand. The platform adds and removes machines automatically.

Storing the science

• Your own space. Private storage that is still there when you come back.
• Team space. A shared filesystem for your project's data, models, and results.
• Project object storage. Private large-scale cloud storage for your project, reachable from the notebook.
• Data browser. Browse and download from your project storage and from public AWS science datasets, inside JupyterLab.

AI help

• AI in the notebook. Chat and agents inside JupyterLab, using several model providers.
• Help on one notebook or one cell. Ask for code, an explanation, or a fix in place.
• Agents and a secure gateway. Run agents on the platform, reached through one trusted entry point. Some workflows ask a human to approve before they act.
• The AI Registry, in JupyterLab. A hub of AI building blocks — MCP servers, skills, agents, models, prompts, and harnesses — shared by NASA colleagues and validated by NASA. Search it, deploy an item, and hand it to your AI assistant, from the AgentRegistry panel. Being rolled out to every project. See the tool reference.
• Model endpoints on demand. Import and call your own or a managed model from the notebook.

Working together

• Real-time co-editing. Several members can edit the same notebook, code file, or document at the same time, Google-Docs style. Markers show who else is online. Work lands in the shared project space.
• Your project turns it on. A project admin switches collaboration on for the project in the RPAdmin console. It is self-service. No ticket to the platform team.

See Working together on the same notebook.

Seeing what it costs

• Your own usage. You see your own compute and storage usage, and your remaining budget, inside JupyterLab.
• Your project's picture. Your project admin sees the whole project's spend against budget in the RPAdmin console, described next.

The RPAdmin console — self-service for project admins

RPAdmin is the web console for the person who runs a project. Everything in this list is self-service: the project admin does it, with no ticket to the platform team.

• Track cost. Compute hours, compute cost, and storage cost for the project.
• Set and share out budgets. Give each member an allocation. The platform blocks giving out more than the project holds.
• See platform usage. Who is using what, and how much.
• Manage members and roles. Add members, remove members, and promote a member to project admin.
• Turn on collaboration. The project-level switch described above.
• Cross-account AWS access. Register one of your own AWS roles, and the platform hands it to your notebooks at start-up. With it you can use any AWS service in your own account from the Research Platform — for example, start an HPC cluster there, or read a mission bucket. (HPC means high-performance computing: a large group of machines for heavy jobs.) No access keys are copied, and no secrets sit in a notebook.
• Audit trail. Every account, membership, and budget change, with who did it and what changed.

See RPAdmin — the admin console.

Coming

These are agreed and being worked on. No dates. They are listed in rough order of how far along they are.

In active development

• Faster start-up. Keep launch time short even when your home space grows large.
• One-click Hugging Face models. Deploy a public model with one click and call it from your notebook.
• Better "my usage" view. Usage and remaining budget shown on the launch screen as well.
• Roles in the AI Registry. Per-user and per-group permissions, instead of one level for everyone.

Planned next

• Scored AI tools. Every tool in the catalog scored for quality and safety, with history.
• Plain-language search. Find a model, agent, or tool by describing what you need.
• Budget stop at launch. Show a true remaining balance, and stop new workspaces when the project's money is used up.
• Storage cost per person. See what each member's object storage costs.

Being explored

• Project-defined machine sizes. A project admin sets the sizes their own team can pick.
• Session detail. Per-session CPU, memory, and storage detail, to understand a cost and size the next run better.
• Review before publish. A draft-and-approve step before an AI asset goes live in the catalog.
• More clouds in the data browser. Google Cloud and Microsoft Azure support.

Want the engineering detail?

The full catalog, with an identifier for every feature and its status, is the Product Roadmap. This page is the plain-language summary of Part I of that file.