Agentic Engineering
Agentic Engineering puts Rescale Edison and a full library of engineering agents to work on the tasks around your simulations.
Describe a task in plain language and Edison routes it to the right agent. Agents check input files before you submit, diagnose and fix failed runs, answer questions from your team's data, and report on finished work.
Agentic Projects keep your program's context across sessions and teammates, so you stop re-explaining your setup. Workflow Builder turns a process you worked out with Edison into a workflow template your team can run again. Agentic Engineering is built for simulation engineers, not AI specialists, and anything that would change your work asks you first.
What is an agent?
An agent handles one engineering task, such as validating an input file or diagnosing a failed job, and returns a result. Rescale Edison is the assistant you work with. Describe a task in plain language, or type / to pick an agent, and the Orchestrator routes the work.
Agents call each other when a task needs it: Job Troubleshooter uses Job Failure Summarizer, and several agents draw on Knowledge Navigator for cited answers from your data. Every agent works with your permissions, so it reads only the jobs, files, and budgets you can open. Anything that would change something, such as submitting a job or applying a fix, asks for your approval first.
Benefits
- Catch problems before you spend compute. Job Input Validator checks an input file before you submit and points to the line that would stop the job, with a suggested fix.
- Recover from failed runs faster. Job Troubleshooter reads the logs, proposes a fix, and resubmits once you approve it.
- Get answers you can check. Knowledge bases connect Edison to your own standards, specs, and reports so it can answer questions citing the document and where it is stored.
- Keep your team's context and process. Agentic Projects carry your objective, baseline jobs, and learnings into every chat, and Workflow Builder saves a process your teammates can run.
Key capabilities
Agentic Engineering includes everything in Advanced Modeling & Simulation and adds the capabilities below.
Rescale Edison and the full agent library
Ask Edison a question or describe a task, and it routes the work to the right agent. Agents validate inputs, troubleshoot failures, find answers in your data, manage budgets, and report on finished work.
Use it to: check input files before you submit a simulation run, fix a failed run, or turn a finished campaign into a PowerPoint report.
Agentic Projects
An Agentic Project holds one project's context: its objective, baseline jobs, workflow templates, standards and specs, report template, and learnings. Attach it to a chat, and Edison and every agent it calls start with that context. Share it with your workspace so teammates work from the same baseline.
Use it to: stop re-explaining your setup and keep what the team learns in one place.
Knowledge bases
A knowledge base indexes documents in your connected storage so Edison can answer from their contents, citing the document and the place in it. It works over AWS S3, Azure Blob Storage, and SharePoint. Agentic Engineering also adds SharePoint and Databricks to the data connectors you can browse.
Use it to: find what your standards, specs, and past reports say without searching folder by folder.
Agentic workflows
The Workflow Builder agent builds multi-step workflows with you and can turn a process you worked out with Edison into a template. With Agentic Engineering, workflows can include agent and Edison steps, so a step that needs judgment runs as an agent step.
Use it to: make one engineer's process something the whole team can run.
MCP Server
The Rescale MCP Server lets an AI client you already use, such as Claude Desktop, work with Rescale over the Model Context Protocol. It works on every tier, but using Edison or agents through it requires Agentic Engineering. Requests run under your account in one workspace, so the client gets no more access than you have.
Use it to: submit a local input file as a job, or ask why a run failed while your code is open.
AI Credits and usage dashboards
Edison and agent interactions draw from one AI Credits pool shared by the workspace, and larger interactions use more. Admins see the balance, daily burn, and usage by person, and get an email at 50%, 80%, and 100% consumed.
Use it to: track adoption and plan a top-up before anyone is blocked.
How Agentic Engineering compares
Each tier includes everything in the one before it. AI Physics is added on top of Advanced Modeling & Simulation or Agentic Engineering, with its own AI Credits pool.
Common use cases
Find an existing run before you submit a new one
Before you set up a new job, ask Edison whether your workspace already has a run that meets your requirements. It searches the jobs you can access by software, hardware, tags, and custom fields, such as wing designs filtered by Mach number and lift-to-drag ratio. When a past run already answers the question, you can reuse its results instead of paying for and waiting on a repeat simulation.
Validate input files before a batch goes out
Before you submit a parameter sweep, ask Edison to check the input files. Job Input Validator flags anything that would stop a job, with the line number and a suggested fix. It can also run as a check at submission or as a step in a workflow.
Recover from an overnight failure
When a run fails, Job Troubleshooter reads the logs, explains the cause, and proposes a fix. Once you approve, it applies the fix and resubmits, and it tries again if the job fails a second time.
Answer a question from your team's documents
Ask what your standard requires for a load case or a mesh, and Edison answers from the standards and specs in your knowledge base. Each answer cites its source, and Knowledge Navigator tells you when the documents don't support an answer.
Report on a finished campaign
Jobs Report Generator turns a completed set of runs into a PowerPoint deck with findings, charts, and recommendations. Add your report template to an Agentic Project, and the deck follows your team's format.
Prerequisites
- Your workspace is on the Agentic Engineering tier. Your Rescale account team can confirm what your workspace includes.
- The workspace has credits in its Agentic Engineering AI Credits pool. If the pool reaches zero, Edison and agents stop for everyone in the workspace until credits are added.
- You have view permission on the jobs, files, and budgets you want agents to work with.
- To build a knowledge base or link files from external storage, a workspace admin has set up a data connector.
- You work on a platform where Edison is available. Edison and agents aren't available on the ITAR or FedRAMP platforms yet.
Get started
Each step links to its full guide.
- Ask Rescale Edison a question: open Edison from the left navigation and ask about one of your jobs.
- Run an agent: type
/in Edison to pick one, or describe the task and let Edison choose. - Create an Agentic Project: name it, state the objective, and link your golden jobs and standards.
- Set up a knowledge base: an admin connects your document storage and syncs it once.
FAQs
How is Agentic Engineering different from Advanced Modeling & Simulation? Agentic Engineering includes everything in Advanced Modeling & Simulation and adds the full agent library, Agentic Projects, knowledge bases, agent steps in workflows, and more data connectors. Edison draws from a shared AI Credits pool for the workspace instead of a daily allowance per user.
Do I need AI or machine learning experience? No. You work with Edison in plain language, and the agents handle the AI. You decide what to run and approve any change before it's made.
How is Agentic Engineering priced? The agents, Agentic Projects, and knowledge bases are included with the Agentic Engineering tier. Edison and agent interactions draw from the workspace's AI Credits pool, and compute bills at standard Rescale rates. Contact your Rescale account team for pricing.
How far do AI Credits go? It depends on the size of each interaction: a quick question uses less than a long analysis over a lot of context. Run for a month, then read your burn rate on the AI Credits Utilization dashboard to size future top-ups. Starting a new chat for unrelated work usually uses fewer credits.
What happens when the AI Credits pool runs out? Edison and agents stop for everyone in the workspace until credits are added. Standard jobs, storage, and software keep working, and so does AI Physics if its own pool has credits. Admins get emails at 50%, 80%, and 100%, so treat the 80% email as your cue to plan a top-up with your account team.
What if an agent doesn't recognize my solver? Job Input Validator and Job Troubleshooter rely on skills written for each solver. If yours isn't recognized, the agent prompts you to author a skill, which then applies across your organization.
Does Rescale use my data to train AI models? No. Rescale doesn't use your prompts, Edison's answers, or your engineering data to train or fine-tune AI models, and doesn't let third parties do so.
Support
Contact your Rescale representative or visit the Rescale Support Center.
Related
- Agentic Engineering agents — every agent, what it does, and how to run it
- Rescale Edison — the assistant, its chats, and how your data is handled
- AI credits — how Edison and agent usage is metered and monitored
- AI Physics — surrogate model training you can add to this tier