> ## 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.

# Strand AI in ChatGPT

> What the Strand AI MCP server can and cannot do from ChatGPT: run markers on slides that are already ingested, with slide upload handled elsewhere because ChatGPT's sandbox has no outbound network access.

ChatGPT can connect to the Strand AI MCP server as a custom connector and
operate samples your organization already has. It cannot bring a new slide in.

<Warning>
  **You cannot upload a slide from ChatGPT.** Sending a slide to storage
  requires the client to make an outbound network request, and as of August 2026
  ChatGPT's sandbox does not allow one. This is a property of ChatGPT's sandbox,
  not a Strand AI restriction, and unlike claude.ai there is no host to add to
  an allowlist. Ingest the slide first at
  [app.strandai.com/samples/upload](https://app.strandai.com/samples/upload),
  from [Claude Code](/mcp/claude), or with the [Python SDK](/sdks/python), then
  work on it from ChatGPT.
</Warning>

Once a sample exists, ChatGPT can list and read samples, run cell segmentation,
estimate or submit a Lattice run, track the job, and return a result download
link.

## Before you start

Adding a custom MCP connector requires ChatGPT's **developer mode**. OpenAI
marks this mode as elevated risk and disables it by default. In Business and
Enterprise workspaces, an administrator must enable it before members can see
the add-connector UI. Check OpenAI's connector documentation for the current
setting and contact your workspace administrator if necessary.

You also need a Strand AI account, membership in a Strand AI organization, and
enough credits in that organization for the runs you intend to submit.

## Connect

Add a custom connector pointing at the Strand AI server URL:

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

Strand AI publishes OAuth metadata and a client-registration endpoint at that
URL. Compatible clients discover the authorization flow from this metadata and
open Strand AI's sign-in and consent screen. Sign in, then select the Strand AI organization this connection
may act on. The connection is bound to your user and the selected organization.
Runs use that organization's credits.

<Note>
  If connecting fails, **delete the connector and add it again**. MCP clients
  generally cache their initial OAuth client registration. Reconnecting reuses the
  cached registration.
</Note>

You can review or revoke a connection at any time from [Settings, Connected
Apps](https://app.strandai.com/settings/connected-apps) in the Strand AI
dashboard. Revoking takes effect immediately.

## What works from ChatGPT

The connector exposes ten tools. Tools other than `upload_slide` call Strand AI
directly and do not require outbound network access from the client.
`upload_slide` requires outbound access because the client sends the slide bytes
to storage.

| Tool                                                         | Needs the client's own network access                                                    |
| ------------------------------------------------------------ | ---------------------------------------------------------------------------------------- |
| `list_markers`, `list_samples`, `get_sample`, `patch_sample` | No                                                                                       |
| `segment_sample`, `submit_run`, `cancel_job`, `wait_for_job` | No                                                                                       |
| `download_results`                                           | No. It returns a link; you fetch the bytes yourself                                      |
| `upload_slide`                                               | **Yes.** It creates the upload session, but the client has to send the slide bytes to it |

`upload_slide` returns a presigned session for sending slide bytes directly to
storage over HTTP. Without outbound network access, the transfer cannot finish
and the sample remains in `uploading` until the organization's retention policy
removes it. [Claude Code](/mcp/claude) and the [SDKs](/sdks/quickstart) run on
your machine and can complete the transfer.

See the [MCP server overview](/mcp/overview) for what each tool does and the
[Pricing page](/pricing) for what a run costs.

## Example prompts

> List my Strand AI samples and summarize the job history of the most recent
> one.

> Dry-run CD3e, CD8, and Ki67 on sample `SAMPLE_ID`, show me the credit
> estimate, and ask before submitting.

> Job `JOB_ID` finished. Give me an OME-TIFF download link for it.

## Data handling

<Warning>
  Research use only (RUO). Not for use in clinical or diagnostic procedures.
</Warning>

The service is for de-identified data only. Do not upload protected health
information or other regulated health data, and upload only data you are
authorized to process. Organization retention settings and your Strand AI
agreement govern uploaded data and results. See the [Strand AI Privacy
Policy](https://strandai.com/privacy) and [Security and data
handling](/security-data-handling). Report suspected security issues to
[security@strandai.com](mailto:security@strandai.com).

For connector problems, email
[support@strandai.com](mailto:support@strandai.com) with the job or sample ID.
Do not include slide data or credentials.