From the Research Universities edition of October 29, 2026
Berkeley Lab Molecular Foundry connects a commercial LLM to an advanced electron microscope through MCP, letting users run experiments in plain language
Molecular Foundry, Lawrence Berkeley National Laboratory (DOE Office of Science user facility) · National laboratory user facility and research instrumentation · United States (California); applicable to university core facilities
- Publisher
- Molecular Foundry, Berkeley Lab
- Original publication
- 29 October 2026 (Molecular Foundry news, adapted from a Berkeley Lab press release)
- Source retrieved
- 2026-10-30
What happened
A Molecular Foundry team rewrote automation routines it had already built for a transmission electron microscope in Model Context Protocol (MCP) format, which gives a commercial language model a fixed set of pre-written actions rather than open-ended control. The resulting "TEM Agent" lets researchers describe an experiment in plain language and strings together steps such as changing magnification, tilting the sample and refocusing the beam. In its first tomography attempt it confused sample tilt with scan rotation, which the team fixed with a short written explanation. The team also tested ptychography. TEM Agent is linked to NERSC and to Crucible, the Foundry metadata platform, which the article says lets raw data be processed in seconds instead of hours and lets users ask what settings they used weeks earlier. A single detector can generate about 700 GB in 15 seconds, and visiting users can need weeks of training. The work is described in an npj Computational Materials study.
Why it matters
Shared instrument facilities at research universities face the same training and data bottleneck. This shows a lower-risk pattern: expose existing, tested automation to an LLM through a bounded tool interface and a metadata platform, rather than giving the model direct control.
Evidence and measured results
Read the full article. No accuracy, error-rate, safety-incident or user-outcome measures are given; the seconds-versus-hours figure is the lab's description. The cited journal paper was not inspected.
Limitations and uncertainty
Institutional news adapted from a press release; the agent's reliability is described through anecdotes, including one initial failure. Results come from one facility with mature existing automation, which the team says made integration easier than expected.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-29; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Advanced shared instruments need weeks of training and produce data faster than staff can process, limiting access for visiting researchers.
- Stakeholders
- Core facility director
- Vice president for research
- Research computing lead
- Instrument staff scientists
- Discovery questions
- Which instruments already have scripted automation?
- How long does user training take today?
- Where is experiment metadata stored, and can users search it?
- Value hypothesis
- A plain-language front end over existing automation may let more users run standard techniques with less staff time, if validated.
- Potential engagement
- A pilot exposing one instrument's existing scripts through MCP with logging and staff review.
- Unsupported claims
- Measured training-time reductions
- Error rates or safety of agent control
- Results on instruments without existing automation
Pre-sales engineering
Role takeaway
- Fit
- Fits facilities with scriptable instruments and a metadata platform.
- Architecture and integration
- MCP server wrapping pre-written instrument actions, a metadata catalog for recall, and HPC links for processing.
- Prerequisites
- Scriptable instrument control interface
- Experiment metadata store
- Approved commercial or local model service
- Constraints
- Vendor instrument software may lack open interfaces
- No published reliability data in the article
- Commercial model data terms
- Security
- Authenticate the MCP server, restrict actions to a reviewed list, and keep unpublished data on approved services.
- Proof of value
- Run established techniques and count operator corrections
- Log and review every tool call
- Compare processing time with and without HPC linkage
Delivery
Role takeaway
- Work
- Convert scripts to MCP tools, write model guidance, connect metadata and compute, train staff and users.
- Dependencies
- Instrument vendor interfaces
- Metadata platform
- Model service approval
- Ownership
- Core facility manager with research computing support.
- Skills and adoption
- Staff learn to maintain tools and correct agent behavior; users learn to state experiments clearly and check results.
- Governance checkpoints
- List of actions requiring human confirmation
- Data classification before model access
- Review after each new technique
- Acceptance criteria
- Agreed share of test tasks completed without correction (proposal)
- All actions logged and attributable
- No unapproved data leaves the facility
- Risks
- Agent confusing instrument actions
- Over-reliance by inexperienced users
- Model or vendor changes breaking tools
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Bounded MCP tools wrap existing instrument scripts; links to an HPC center and a metadata catalog provide processing and recall of past settings.
Governance
Who approves, reviews and stays accountable for outcomes?
Restricting the model to pre-written actions is a control choice; facilities should still define which actions need human confirmation.
Security and privacy
What data, permissions and controls need testing?
A commercial model is connected to instrument control and experiment metadata; the article does not describe data handling, authentication or model hosting.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Aims to let less-experienced visiting users run advanced techniques, potentially reducing weeks of hands-on training, though this is not measured.
Procurement
What should contracts, pricing and exit terms secure?
Relies on a commercial model and in-house tooling; Crucible is being extended to other nanoscale research centers.
Operating model
Which teams own the service once it runs?
Staff scientists maintain the tool layer and teach the agent experiment rules; the agent becomes a front end to several systems.
Publication history
- 2026-10-29Research Universities · Issue 514 resources
Stable resource ID: lbnl-molecular-foundry-tem-agent-mcp-microscope-20261029