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

# Set up TCR-Workbench

> Use the tcr-workbench-setup Skill to install, configure, and verify a local TCR-Workbench environment.

Use `tcr-workbench-setup` to install TCR-Workbench on a laptop, Mac, workstation, or GPU machine and get it to the point where analyses can run.
The Skill chooses the device, model size, and weight source for your machine, runs setup, and translates failures into fixes.

> [!NOTE]
>
> This Skill calls no Biohub tools.
> Your AI client runs the TCR-Workbench launcher in a local checkout of [TCR-Workbench](https://github.com/akdslab/tcr-workbench), so it needs a shell on your own computer, Python 3.9 or newer, and, for model setup, git and network access to GitHub and Hugging Face.

## Before you begin

* Connect Biohub MCP to a client that can run commands on your computer; see [Connect your client](https://docs.biohub.ai/biohub-mcp/connect-clients.md).
* Review the commands and downloads before approving them; see [Skills that run commands on your computer](https://docs.biohub.ai/biohub-mcp/security.md#skills-that-run-commands-on-your-computer).

## Run the Skill

Install TCR-Workbench on a Mac

Using the Biohub MCP tcr-workbench-setup Skill, set up TCR-Workbench on my M2 MacBook so I can score peptides.

## What the Skill guides

1. It starts read-only with `python3 tcr.py doctor` and stops when an install already fits the intended device and model.
2. It chooses CPU, Apple Silicon, or NVIDIA GPU from the machine, and sends Windows users to WSL.
3. It picks the weight source, Hugging Face by default or a shared registry when one is configured, and the DecoderTCR model size the hardware can hold.
4. It runs `python3 tcr.py setup`, which installs DecoderTCR from a pinned commit and configures the chosen runtime and weight source.
5. It diagnoses common failures, including missing git, Hugging Face authentication, and TLS-inspecting proxies, without disabling certificate verification.
6. It verifies the install with `python3 tcr.py doctor --deep` and hands off to the [`tcr-workbench`](https://docs.biohub.ai/biohub-mcp/skills/tcr-workbench) Skill.

## Expected result

The result includes:

* A managed `.tcr/` environment in the checkout with the chosen device, model, precision, and weight source.
* A `doctor --deep` report that shows a reachable weight source and healthy framework imports.
* The next command to run, such as the synthetic example that `python3 tcr.py example` writes.

> [!WARNING]
>
> Model workflows require weight downloads, about 1.3 GB for the default 300M model.
> Review the selected device, model, and downloads before approving setup or scoring.

## Skill files

| Resource | URI |
| - | - |
| Instructions | `skill://tcr-workbench-setup/SKILL.md` |
| File list | `skill://tcr-workbench-setup/_manifest` |
| License (MIT) | `skill://tcr-workbench-setup/LICENSE.md` |

## Related documentation

* [Biohub MCP Skills](https://docs.biohub.ai/biohub-mcp/skills/overview.md)
* [Analyze TCR data with TCR-Workbench](https://docs.biohub.ai/biohub-mcp/skills/tcr-workbench.md)
* [TCR-Workbench on GitHub](https://github.com/akdslab/tcr-workbench)


## Related topics

- [Analyze TCR data with TCR-Workbench](https://docs.biohub.ai/biohub-mcp/skills/tcr-workbench.md)
- [Biohub MCP Skills](https://docs.biohub.ai/biohub-mcp/skills/overview.md)
- [Get started with ESM models](https://docs.biohub.ai/learn/getting-started.md)
- [Codex, Claude, ChatGPT, and Cursor connection reference guide](https://docs.biohub.ai/biohub-mcp/connect-clients.md)
- [Connect to the Biohub MCP and run your first query](https://docs.biohub.ai/biohub-mcp/quickstart.md)