# MCP Server

<!-- -->

The Tigris MCP server lets you use Tigris in your agentic workflows without having to install additional software on your machine. It implements the [Model Context Protocol](https://modelcontextprotocol.io/) over streaming HTTP at `https://mcp.storage.dev/mcp` and authenticates with OAuth, so it works with desktop tools and with web agents such as ChatGPT.

## Features[​](#features "Direct link to Features")

The MCP server exposes Tigris to your agent as a set of tools:

* **Organizations**: list the organizations you belong to and pick the one to work in.
* **Buckets**: create, list, inspect, and delete buckets.
* **Objects**: list, read, write, and delete objects, and look up their metadata.
* **Sharing**: create presigned URLs so a browser, a person, or another system can download or upload an object without credentials.
* **Access keys**: list access keys and create keys scoped to a single bucket.
* **Agent workspaces**: give an agent a dedicated bucket with an optional expiry and scoped credentials, then tear it down when the work is done.
* **Checkpoints and forks**: snapshot a bucket, restore a snapshot into an independent bucket, and create copy-on-write forks so several agents can work on the same data in parallel.

## Example Prompts[​](#example-prompts "Direct link to Example Prompts")

Here are some prompts you can try once the server is connected:

* List my Tigris buckets
* Create a bucket called `my-bucket` with snapshots enabled
* Save this summary as `summary.md` in `my-bucket`
* Give me a link to share `summary.md`
* Create a workspace bucket for this task that expires in 7 days
* Checkpoint `my-bucket` before we start
* Create three forks of `dataset` so the agents can work in parallel
* Delete `old-file.txt` from `my-bucket`

## Installation[​](#installation "Direct link to Installation")

The quickest way to set up the server is to run [`tigris init`](https://www.tigrisdata.com/docs/cli/init.md). It detects your coding agent and writes the MCP server configuration for it. To set up a client by hand, pick it below.

* Claude Code
* Claude Desktop
* Cursor
* VS Code
* ChatGPT
* OpenAI Codex
* Goose

Run this command:

```
claude mcp add --scope user --transport http tigris https://mcp.storage.dev/mcp
```

Then run `claude` and use the `/mcp` command to authenticate.

1. Go to Settings > Connectors.
2. Click the **Add custom connector** button.
3. In the **Name** field, enter "Tigris".
4. In the **Remote MCP server URL** field, enter `https://mcp.storage.dev/mcp`.
5. Click the **Add** button.
6. Once the connector is added, click the **Connect** button and complete the authentication flow.

Click this button:

[Add to Cursor](cursor://anysphere.cursor-deeplink/mcp/install?name=tigris\&config=eyJ1cmwiOiJodHRwczovL21jcC5zdG9yYWdlLmRldi9tY3AifQ%3D%3D)

Or run this command:

```
npx @tigrisdata/tigris-mcp-server setup cursor
```

Click this button:

[Add to VS Code](vscode:mcp/install?%7B%22name%22%3A%22tigris%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fmcp.storage.dev%2Fmcp%22%7D)

Or do the following:

1. Press Control+Shift+P or Command+Shift+P to open the command palette.
2. Search for "MCP: Add Server".
3. Create an HTTP MCP server.
4. Enter the URL [`https://mcp.storage.dev/mcp`](https://mcp.storage.dev/mcp).
5. Name the MCP server `tigris`.
6. Authenticate with Tigris and get off to work!

1) Open [ChatGPT Settings](https://chatgpt.com/#settings/Connectors) and click **Create** in the top right corner of the Settings modal.
2) In the **Name** field, enter "Tigris".
3) In the **MCP Server URL** field, enter `https://mcp.storage.dev/mcp`.
4) In the **Authentication** field, select "OAuth".
5) Click **Create**.

Enable the `rmcp_client` feature in `~/.codex/config.toml`:

```
[features]

rmcp_client = true
```

Then run this command:

```
codex mcp add --url https://mcp.storage.dev/mcp tigris
```

Then run `codex` and use the `/mcp` command to authenticate.

Run `goose configure`:

```
goose configure
```

Choose "Add Extension":

```
┌   goose-configure

│

◆  What would you like to configure?

│  ○ Configure Providers

│  ○ Custom Providers

│  ● Add Extension (Connect to a new extension)

│  ○ Toggle Extensions

│  ○ Remove Extension

│  ○ goose settings

└
```

Choose "Remote Extension (Streamable HTTP)":

```
┌   goose-configure

...

◆  What type of extension would you like to add?

│  ○ Built-in Extension

│  ○ Command-line Extension

│  ● Remote Extension (Streamable HTTP) (Connect to a remote extension via MCP Streamable HTTP)

└
```

Enter the name "tigris":

```
┌   goose-configure

...

◆  What would you like to call this extension?

│  tigris

└
```

Enter the endpoint URI `https://mcp.storage.dev/mcp`:

```
┌   goose-configure

...

◆  What is the Streaming HTTP endpoint URI?

│  https://mcp.storage.dev/mcp

└
```

Set a timeout of `300` seconds:

```
┌   goose-configure

...

◆  Please set the timeout for this tool (in secs):

│  300

└
```

Enter a description:

```
┌   goose-configure

...

◆  Enter a description for this extension:

│  Tigris MCP server

└
```

Choose "No" when asked if you need to add custom headers:

```
┌   goose-configure

...

◆  Would you like to add custom headers?

│  ○ Yes  / ● No

└
```

Then run `goose` and authenticate with Tigris in your browser.

## Learn More[​](#learn-more "Direct link to Learn More")

* [How Do I Use Tigris with AI Coding Agents via MCP?](https://www.tigrisdata.com/docs/ai-agents/mcp-server.md)
* [Agent Skills](https://www.tigrisdata.com/docs/skills.md)
* [Agent Plugins](https://www.tigrisdata.com/docs/ai/agent-plugins.md)