| name | chemtools-inspect-run |
| description | Inspect completed, failed, or partial output from NWChem, OpenMolcas, DIRAC, GRASP, Quantum ESPRESSO, QMCPACK, or ORCA. Use when the user asks whether a run is usable, why it failed, what the output proves, or what evidence supports a next step. |
Inspect a chemistry run
Use the Chemtools inspect_run MCP tool to establish the scientific evidence
before recommending another calculation.
- Require one primary output path. Do not scan its directory.
- Pass only related artifacts the user identified, such as the matching input,
stderr, checkpoint, or orbital file. Preserve their order.
- Pass
program only when supplied by the user or needed to resolve an
ambiguity. Do not force a program that conflicts with a positive detector.
- Call
inspect_run and separate the response into execution or convergence
verdict, scientific evidence, warnings, uncertainty, and next actions.
- Distinguish facts parsed from the output from conclusions that remain
conditional on missing input, geometry, method, basis, or state evidence.
Use compare_runs when the question depends on two outputs. Use
plan_recovery only for NWChem after the inspection establishes a failure or
instability and the matching input is available. Recovery planning returns
candidate text for review; it does not write or launch anything.
If the user supplies only a PID, scheduler ID, or launch ID, do not treat that
identifier as an output. Use chemtools-monitor-run for a Chemtools-owned
launch ID, or ask for the primary output path.
