Your instruments.
Part of the workflow.

Not listed? The route is the same: if the instrument’s software saves a file, Seal IoT collects it, and each connection is scoped and tested with a representative export from your installed version. Send us the model and an example output.

Install Seal IoT on a Windows or macOS machine with access to the files: the instrument workstation, a local staging machine or a host that can read the network share. One host can run several folder triggers if its account can reach each source.

For each source, choose the folder, whether to include subfolders and which Seal template receives the files. A file-watch trigger detects a new or changed file, waits until its size is stable and runs the upload script. A Seal Script then checks the export’s layout, resolves the run and sample it belongs to and writes the selected values into typed fields.

Acquisition and vendor analysis stay in the instrument software: Seal collects completed results and sends no instructions to the instrument. The released Seal IoT workflow collects files; it does not read controllers directly. Preserving a proprietary file and interpreting its contents are separate capabilities, so the formats, extraction and mapping are agreed and verified for each source. Scientific data management explains the capture route in full.

A traditional SDMS archives the file beside the work, and a point-to-point interface copies a value from one system into another. Seal keeps the original and its values together, related to the sample, method and run, so a reviewer, an investigator and Neil reach the same preserved evidence from the work they are doing.

Matching uses the run or sequence, injection and sample rather than a filename, which can collide across experiments, and ambiguous matches go to review instead of attaching a result to the wrong sample. A file that arrives does not by itself show that the acquisition is complete: define the files each run must produce, because a finished PDF does not prove that the raw data, method and audit trail arrived.

Reprocessing creates a derived record with its reason, and the earlier result stays available for comparison. Capture is not qualification: keeping an original file does not establish that the instrument was suitable for the work, which equipment management records.

Test connectivity from the actual host before relying on it. Windows network shares use the running account’s permissions, so an unattended installation needs its own account and the host’s startup arrangement. Authentication and agent updates use service domains beyond seal.run; review the full allowlist and update behaviour before deployment.

If the agent stops, files that arrived meanwhile can be picked up when it starts again, provided they are still at the source. It shows connection diagnostics, recent runs and errors, and upload scripts can retry transient failures. Agree a source-retention window and how repeats are reconciled: a changed file can trigger another upload, and the basic folder-to-record script does not deduplicate by filename.

The scientific data blueprint sets out installation in full, and connected systems in the regulatory guide covers how a connection is qualified.

Neil can compare methods, lots and preparation steps across runs, then propose a workflow improvement for your team to test and approve. Unsupported formats, missing files and unresolved matches stay visible rather than becoming confident guesses, and receiving a file never approves a result.

In line with the EU’s draft GMP Annex 22 on AI, Neil is not used in GMP execution. It helps set up configuration, which your team verifies and releases under change control. See how Neil fits Annex 22 and more about Neil.