backed by Combinator

Strand AI Connector for Claude

Spatial proteomics from H&E through an authorized connector

What It Does

Lattice predicts spatially resolved protein-marker maps from routine H&E, delivered as virtual mIF channels. The Strand AI connector lets an AI assistant securely upload a whole-slide image, estimate and submit a Lattice prediction, monitor or cancel the job, and obtain OME-Zarr or OME-TIFF results. It calls Strand AI's first-party API and uses your existing Strand AI user and organization permissions.

Step-by-step setup, the full tool list, and data-handling details live in the Strand AI connector docs.

Before Connecting

  • A Strand AI Platform account and membership in a Strand AI organization
  • A supported, de-identified H&E whole-slide image
  • Enough credits in the selected organization for the requested inference job

There is no separate connector credit cap or free tier. Before submission, Strand AI shows an estimate. Submission reserves that amount against the selected organization's balance. Completion settles and debits the delivered work, then releases any unused reservation. Eligible cancellation releases the reservation.

Authorization

Claude opens Strand AI's OAuth consent flow. Sign in to Strand AI, explicitly select the organization Claude may act in, review the requested read, write, results, and credit-spend access, then approve or deny it. Revoke access at any time from Platform Settings under Connected Apps.

Remote File Transfer

Upload: upload_sample creates the canonical owned sample and a presigned, resumable PUT transfer bound to the selected organization. Its returned upload ID is the sample ID for get_sample and submit_job. A capable client sends one supported H&E WSI directly to storage; committing the final byte automatically starts ingest processing. Automated de-identification is off by default and runs during ingest only when enabled for your organization. Transfer bytes bypass the conversational model context and MCP transport; Lattice still processes the stored slide to generate predictions.

Browser upload is an independent alternative. Upload at app.strandai.com/samples/upload, then discover and select the resulting owned sample with list_samples and get_sample. The browser path uses the selected organization's ordinary sample visibility; it is not a handoff from an upload session.

Download: download_results takes the format you want. Native OME-Zarr is returned without conversion. An OME-Zarr ZIP or OME-TIFF is generated on request, and repeating the call polls that export until an expiring Strand AI download link is ready. Optional include_original_upload separately requests the stored source file, not processed H&E. It defaults to false, is runtime-gated, and is currently off in production. The source can be multiple gigabytes; if ingest de-identification ran, the stored de-identified copy is returned. Its private bearer URL expires after 15 minutes, so do not share it.

Available Tools

  • upload_sample - create the owned sample and a direct-to-storage WSI transfer; finalization starts ingest
  • list_markers - list markers available to the selected organization
  • list_samples and get_sample - browse owned or public samples and capped owned job history
  • update_sample - update a sample's name, writable unified field values such as Tags, or microns per pixel; Public is read-only here
  • segment_sample - discover, start, or retry segmentation for an owned sample
  • submit_job - estimate without reserving credits, or reserve credits and start inference
  • get_job and cancel_job - read one exact status without waiting or polling, or cancel an in-flight job
  • download_results - retrieve native OME-Zarr or create an OME-Zarr ZIP or OME-TIFF export, with optional processed H&E, segmentation, or the gated stored original upload

Example Prompts

  • Upload my de-identified H&E slide to Strand AI, dry-run a CD3e, CD8, and Ki67 prediction, show me the credit estimate, and ask before submitting it.
  • Read sample SAMPLE_ID and summarize its job history. Read job JOB_ID; if it is complete, give me the OME-Zarr ZIP download link.
  • Update sample SAMPLE_ID to 0.5 microns per pixel and replace its tags, then dry-run CD4 and FOXP3 before asking me to submit.

Troubleshooting

  • If authorization fails, reconnect Strand AI and select an organization where your account is still a member.
  • If a presigned upload session expires, create a new upload_sample session and send the bytes again. Browser upload is a separate alternative; select its resulting owned sample with list_samples or get_sample.
  • If submission reports insufficient credits, choose a funded organization or ask its Strand AI administrator to add credits.
  • If a result link expires, ask Claude to run the corresponding download tool again.

For unresolved connector issues, include the job or sample ID—but no slide data or credentials—in an email to [email protected].

Data Handling and Support

Organization retention settings and your Strand AI agreement govern uploaded data and results. Slides must be de-identified before upload. You must be authorized to process them. Do not upload protected health information or other regulated identifiable health information. See the Strand AI Privacy Policy and Security and data handling. For connector support, email [email protected]; report suspected security issues to [email protected].

Research use only (RUO). Not for use in clinical or diagnostic procedures.