New from CompuTherm
PandatAi
AI-powered materials design platform
Ask in plain English. PandatAi sets up the calculation, runs it on the same Pandat engine and the same databases you already license, then returns the analysed result — tables, labelled diagrams and interactive plots.
An addition to Pandat — not a replacement. Same engine. Same databases.
The questions engineers actually ask
These are single sentences. Answering any one of them in a conventional workflow means choosing a database, picking the phase set, writing a batch file, running it, then post-processing the output — four separate steps, each with its own opportunity for error.
PandatAi prepares all four steps end-to-end, on the same engine.
How it works
PandatAi sits between your question and the Pandat calculation engine. It is connected through the Model Context Protocol (MCP), so the same tool set is available whether you work inside an AI assistant or in the standalone PandatAi app.
Your question
Plain English, in your own words. Attach a file or an image if the alloy list is already on disk.
PandatAi
Selects the database and the module, builds a standard Pandat batch file, and runs it on the Pandat engine.
Analysed results
Tables, labelled phase diagrams, interactive plots and reports — already interpreted, not just dumped.
Nothing is hidden
Every calculation PandatAi runs is written as a standard Pandat batch file in your run folder. Open it, check it, run it yourself in Pandat. Every recommended number cites the run it came from, and every assumption is labelled as an assumption.
What it can do
Calculations
Equilibria and phase diagrams, Scheil and kinetic solidification, precipitation and grain growth, 1-D diffusion, and high-throughput screening across a composition grid.
Knowledge Q&A
Ask how a calculation works or what a setting means. Answers come from a built-in knowledge base and the Pandat Example Book — including running an example for you.
Alloy design & workflows
From a property goal to a composition and a heat-treatment schedule, with feasibility gates applied along the way. Define a workflow once and reuse it as a named skill.
Analysis & rendering
Results come back analysed and plotted: interactive figures with their own font, axis-range and K/°C controls, labelled phase diagrams, ranked tables and written reports.
Pandat modules PandatAi drives
PandatAi currently drives four Pandat calculation modules, plus high-throughput calculation through the PanPython SDK.
PanPhaseDiagram
- point, line and section calculations
- stable and metastable phase diagrams
- liquidus projection, contour, property line
- Scheil solidification and CSI
PanEvolution
- precipitation kinetics (KWN)
- ageing curves, hardness and yield strength
- TTT diagrams and particle size distribution
- grain growth and recrystallisation
PanSolidification
- kinetic solidification with back-diffusion
- finite cooling rate
- dendrite arm spacing (SDAS)
- micro-segregation and solute trapping
PanDiffusion
- diffusion couples
- homogenisation
- carburisation and decarburisation
- particle dissolution
High-throughput (PanPython)
- composition grids and alloy lists
- process-variable matrices
- automated screening and ranking
- heat maps and x–y sweeps
More to come
- further Pandat modules as batch templates, tools and skills are added
- extended knowledge base
- shared user skills
Where you can run it
As a plugin in your AI assistant
Install PandatAi from the plugin marketplace of the assistant you already use — Claude Desktop and Claude Code, or Codex Desktop and the Codex CLI. Your existing Pandat installation and licensed databases are used as they are.
As the standalone PandatAi app
A dedicated desktop application with its own workspace browser, skills and tools panel, and conversation area. Runs on Windows and Linux. Works with the major LLM families; requires an API key.
You stay in control
PandatAi never runs a calculation you have not seen. Choose the level of confirmation once, under Configuration → Running Mode.
Guided recommended
PandatAi fills in every input it can reasonably suggest, then shows one summary before running — composition, database, temperature range, phase set. You confirm it, or change it.
Expert
PandatAi stops at every decision gate. You pick the database, the phases, the conditions. For when the defaults are the thing you are studying.
What a session looks like
AskAttached are some candidates we’re considering for this casting. Screen all of them for hot-cracking susceptibility.
PandatAiReads the attached file, runs a Scheil calculation for every alloy in it, ranks them by crack susceptibility index, and returns the ranked table with the plot.
AskI need a high-strength aluminium alloy for a structural application. Choose the Al–Zn–Mg–Cu composition space. First, give me a plan for the calculations needed.
PandatAiProposes a staged plan — screen the composition grid, apply a feasibility gate on the solution-treatment window, optimise the ageing schedule, verify — and runs it once you approve. The highest-strength screening candidate can and does get rejected at the feasibility gate when its solvus sits above the solidus.
AskWhich composition has the highest fcc spinodal in Co–Cr–Cu–Fe–Ni? Build it as a skill I can reuse.
PandatAiBuilds a 126-alloy grid, scans the eigenvalue criterion over temperature, ranks the highest-temperature zero crossings, refines the grid, verifies that the alloy is fully solid at the spinodal temperature — and saves the whole procedure as a named skill you can run again on a different system.
Built for real work
Shareable, resumable sessions
Session folders carry their own metadata. Share one and the recipient gets the conversations and figures, fully resumable.
Session summaries
One click writes a running summary.md beside the chat; later clicks append what is new.
Several instances at once
Every launch opens another window with its own workspace, so long jobs do not block the next question.
Remembered databases
A database used once by full path becomes a named key, reusable by name in any later run.
Interactive figures
Every plot carries its own controls: fonts, axis ranges, and a K/°C toggle.
Progress reporting
High-throughput sweeps and diffusion simulations report progress while they run.
Editable history
Edit a message you already sent and the session re-answers from that point, without starting over.
Python in your own skills
Custom skills can wrap heavy lifting in Python scripts the app runs, so they stay reliable even on smaller models.
One sentence in. A defensible calculation out.
PandatAi is an addition to Pandat — same calculation engine, same thermodynamic and kinetic databases. The free trial is limited to 3 elements and the example databases; full functionality requires a licensed Pandat installation and the corresponding databases.