> ## Documentation Index
> Fetch the complete documentation index at: https://docs.strandai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Use Strand AI from your AI assistant

> Connect Claude, ChatGPT, or another AI assistant to Strand AI to browse your slides, check Lattice estimates, and view predicted spatial proteomics from H&E in the chat.

Strand AI works inside AI assistants such as Claude and ChatGPT. Once
connected, you can ask in plain language to find a slide, check what a run
would cost, and look at the results, without leaving the chat. In Claude you
can also start runs.

Lattice predicts spatially resolved protein-marker maps from routine H\&E,
delivered as virtual mIF channels. The assistant works with the same samples,
credits, and results you see at [app.strandai.com](https://app.strandai.com).

## What you can do

* **Find samples.** "Show my samples from last week" or "Find the TCGA colon
  slides in the public cohort."
* **Organize sample metadata.** Ask to list sample fields or create a field such as
  "Study cohort" with a description and select options. Creating a definition is
  free. Assigning its values is a separate sample update.
* **View a slide in the chat.** Ask the assistant to open a sample, and an
  interactive viewer appears in the conversation. Pan, zoom, and switch markers
  just like on the platform.
* **Check the cost before you run.** "How much would CD3e, CD20, and Ki67 cost
  on this slide?" The estimate is free.
* **Run Lattice (Claude).** Strand AI tells your assistant to show you the credit
  estimate and get your OK before it spends anything. Keep your assistant's
  "ask before using tools that make changes" setting on.
* **Check progress and download results.** Ask whether a run has finished, then
  get a download link for the predicted markers.
* **Segment cells.** Start cell segmentation on a slide (free) to get per-cell
  measurements with your results.

## Pick your assistant

| Assistant | What works | Setup guide |
| - | - | - |
| Claude (web and desktop) | Everything, including the in-chat viewer and slide upload | [Claude](https://docs.strandai.com/mcp/claude.md) |
| Claude Code | Everything except the in-chat viewer | [Claude](https://docs.strandai.com/mcp/claude.md#claude-code) |
| ChatGPT | Browsing, the in-chat viewer, estimates, and results. Upload slides and start runs on the platform. | [ChatGPT](https://docs.strandai.com/mcp/chatgpt.md) |
| Other assistants | Connect with the server address below | [Other assistants](#other-assistants) |

## Before you start

* **You need a Strand AI account** that belongs to an organization. Runs use
  that organization's credits. [Sign up free](https://app.strandai.com).
* **Slides must be de-identified before upload.** Never upload slides or file
  names that contain patient information. Automated de-identification is off
  by default and runs only when enabled for your organization. See [Security
  and data handling](https://docs.strandai.com/security-data-handling.md).
* **Large slides upload best in your browser.** For multi-GB whole-slide
  images, upload at
  [app.strandai.com/samples/upload](https://app.strandai.com/samples/upload),
  then ask your assistant to work with the slide.
* **Research use only.** Not for use in clinical diagnosis, treatment, or
  patient-care decisions. Assistants are instructed to decline clinical
  requests.

## Try these prompts

> Show my samples, then open the newest one in the viewer.

> Estimate CD3e, CD4, CD20, and Ki67 for sample "Biopsy 12". Don't run it yet.

> That estimate looks fine. Run it. (Claude)

> Has my last run finished? If so, give me the results as OME-TIFF.

> Which markers can I predict?

## Good to know

* **The viewer opens one sample at a time.** If you have several samples with
  similar names, the assistant may ask which one you mean.
* **Your slide images stay private.** The viewer shows your results to you, but
  the images themselves are never sent to the AI model.
* **Connections are per organization.** When you connect, you pick one Strand
  AI organization, and the assistant can only see that organization's samples
  and the public cohort. To switch organizations, reconnect.
* **You stay in control.** See and revoke connected assistants anytime in
  [Settings → Connected Apps](https://app.strandai.com/settings/connected-apps).
* **After an update, refresh the tool list.** If something described here is
  missing, refresh Strand AI's tools in your assistant. In Claude: **Customize →
  Connectors → Strand AI → ⋯ → Refresh tools list**. In ChatGPT: **Settings →
  Plugins → Strand AI → Refresh tools**.

## Other assistants

Any assistant that supports remote [MCP](https://modelcontextprotocol.io)
connectors can use Strand AI. Add a custom connector with this address, then
sign in with your Strand AI account when prompted:

```text theme={"dark"}
https://mcp.strandai.com/mcp
```

Assistants that can't sign in for you can use a Strand AI API key instead. See
[API keys](https://docs.strandai.com/mcp/claude.md#using-an-api-key).

For scripted or bulk work, the [Python SDK](https://docs.strandai.com/sdks/python.md) and [REST
API](https://docs.strandai.com/api/overview.md) are usually a better fit than an assistant.

## Help

Email [support@strandai.com](mailto:support@strandai.com) with the sample or run
ID. Don't include slide files, passwords, or API keys.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.