Aller au contenu principal
This page is not translated yet. Français is switched on so the navigation and URLs can be tested; the text below is the English original. Use the language selector in the header to switch back, or see the documentation home.
concept · Reviewed 2026-11-02 GA

Advanced Modeling & Simulation agents

An agent handles one task and hands back a result. Advanced Modeling & Simulation includes agents for three of the jobs that take up time around a simulation: working out what to run on, finding out why a run failed, and answering questions about cost.

Each one reads your own account data, so a hardware recommendation reflects your organization's run history rather than a generic default, and a cost answer comes from your actual billing records.

What an agent is​

An agent is built for a single task rather than for conversation. Rescale Edison is the general interface you can ask anything; an agent is a specialist Edison can hand a task to, and you can also run one directly without going through chat.

Prerequisites​

  • Your workspace needs Advanced Modeling & Simulation. If it is not enabled, your Rescale account team can tell you what applies to your organization.
  • You need view permission on whatever the agent reads. An agent cannot see a job, workspace or budget that you cannot.

Where you invoke an agent​

Entry pointUse it when
Agent Library in the left navigationYou know which agent you want and want to run it directly
A prompt to Rescale EdisonYou would rather describe the task than pick a tool
A node in Workflow BuilderThe task should run as part of a pipeline
A control on the page you are already onFor example, troubleshooting a failure from a job's Results page

In Edison, type / to list the agents available to you, then pick one and give it a prompt or parameters. If you describe the task in plain language instead, the Orchestrator routes it to the right agent and returns the result.

The Agent Library shows the agents your workspace has, each tagged with the product it comes with, so what you see there is what you can run.

Anything that would change something asks you first. Applying a fix, resubmitting a job or editing a budget comes back for your approval rather than happening quietly.

Deciding what to run on​

Hardware Recommender​

Recommends hardware for a given solver and workload before you submit. The choice of coretype is what makes a run fast or slow and cheap or expensive, and it is often made from habit, so asking is usually worth the few seconds. See hardware for the same capability outside chat.

Finding out why a run failed​

Job Failure Summarizer​

Reads a failed job's logs and explains in plain language what went wrong and where. It reasons over the log text rather than matching known patterns, so it handles solvers it has not seen before, and it names the file and line the problem came from rather than returning a category.

Answering questions about cost​

Finance UI Agent​

Reports on billing, budgets, credits and deposits, broken down by workspace, project or member. It answers an end-of-month cost question without anybody exporting a CSV. See billing and reports.

Access more agents with Agentic Engineering​

Agentic Engineering includes these three agents and adds others, among them agents that validate an input file before submission, apply a fix to a failed job and resubmit it, and build workflows with you.