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

# OneBucket Connector for Claude

> Connect Claude to your OneBucket storage over MCP — browse, read, and manage objects and query storage activity, using your own credentials.

The OneBucket connector is a Model Context Protocol (MCP) server that lets Claude browse, read, and manage the objects in your buckets — and query your storage activity — using your own credentials, with your explicit consent.

## What you can do

Ask Claude things like:

* "List my buckets, then show what's in `reports/` in the `analytics` bucket."
* "Download `notes/meeting.md` from `team-docs` and summarize it."
* "Upload this text as `drafts/summary.md` in `team-docs`."
* "Copy `logo.png` to `archive/logo.png`, then delete the original."
* "Give me a presigned download link for `builds/app.zip`."
* "What happened in my storage in the last 24 hours?"

## Tools

Read tools run without per-call confirmation. Write tools prompt for confirmation before running.

### Reading

| Tool      | Title                         | What it does                                                        |
| --------- | ----------------------------- | ------------------------------------------------------------------- |
| `list`    | List buckets & objects        | List buckets, or objects in a bucket (prefix/delimiter/paging)      |
| `head`    | Get object metadata           | Object metadata without downloading content                         |
| `get`     | Download object               | Object content inline (≤700 KB) or a presigned URL for larger files |
| `presign` | Generate presigned URL        | Time-limited GET or PUT URL (\~1 hour)                              |
| `locate`  | Locate object across backends | Which configured backends hold an object                            |

### Writing

| Tool       | Title                      | What it does                                            |
| ---------- | -------------------------- | ------------------------------------------------------- |
| `put`      | Upload object              | Upload inline text (≤700 KB) or get a presigned PUT URL |
| `create`   | Create bucket              | Create a bucket                                         |
| `copy`     | Copy object                | Copy an object to a new key (async)                     |
| `move`     | Move (rename) object       | Move/rename within a bucket (async, copy-then-delete)   |
| `migrate`  | Migrate object to backend  | Copy an object to another backend (async)               |
| `prefetch` | Prefetch object into cache | Warm the local cache for faster next read               |
| `delete`   | Delete object or bucket    | Delete an object, or a bucket when no key is given      |

### Storage events

Served by a separate **events** endpoint (see [Connect](#connect-it-to-claude)):

| Tool          | Title               | What it does                                            |
| ------------- | ------------------- | ------------------------------------------------------- |
| `list_events` | List storage events | Recent storage events, filterable by time/bucket/action |
| `get_event`   | Get storage event   | A single storage event by ID                            |
| `tail_events` | Tail storage events | Long-poll for new storage events                        |

## Requirements

* A OneBucket / Attimis account with at least one access key. [Request an account](https://app.onebucket.io/sign-up).
* Your OneBucket cluster host, shown in your [OneBucket console](https://app.onebucket.io) — the MCP endpoints live on it (see below).

## Connect it to Claude

<Steps>
  <Step title="Add the connector">
    In Claude, go to **Settings → Connectors → Add custom connector** (or find "OneBucket" in the directory once listed).
  </Step>

  <Step title="Enter the server URL">
    Provide your OneBucket **storage** MCP URL — `https://mcp-core.<your-cluster>.onebucket.io` — copying the exact host from your OneBucket console. To also query storage activity, add the **events** endpoint `https://mcp-synapse.<your-cluster>.onebucket.io` as a second connector.
  </Step>

  <Step title="Sign in">
    Claude discovers the authorization server automatically (RFC 9728 / RFC 8414) and opens the OneBucket sign-in page. Sign in and approve the requested access — Claude completes an OAuth 2.1 authorization-code flow with PKCE, and your secret key never leaves OneBucket.
  </Step>

  <Step title="Try it">
    Ask: *"List my OneBucket buckets."*
  </Step>
</Steps>

## How authentication works

The connector uses standard MCP OAuth. It advertises protected-resource metadata at `/.well-known/oauth-protected-resource`; the OneBucket authorization server ("axon") exposes `/.well-known/oauth-authorization-server` with authorization, token, and dynamic-registration endpoints. Clients register dynamically (RFC 7591) and authenticate as public clients using PKCE (S256). Every tool call is scoped to your organization and access key; cross-organization access is not possible. You can revoke access at any time from your [OneBucket console](https://app.onebucket.io).

## Limits & behavior

<Note>
  Inline reads/writes are capped at 700 KB; larger transfers use presigned URLs (valid \~1 hour) so big files bypass the model.
</Note>

* `copy`, `move`, `migrate`, and `prefetch` are asynchronous — they return once queued, so a destination may not be readable immediately.
* Small PDFs/binaries (≤10 KB) are returned inline as blobs; larger ones return a presigned URL.

## Support

* Support: [onebucket.io/support](https://www.onebucket.io/support)
* Privacy policy: *coming soon*