beamdrop

name	beamdrop
description	Interact with a Beamdrop file storage server — upload, download, and manage files via the S3-compatible API. Use when the user wants to store files, create buckets, generate presigned/shareable URLs, manage API keys, set up webhooks, connect via MCP, or integrate Beamdrop into their project. Covers Go SDK, HTTP API, presigned URL strategies, webhooks, MCP server, and error handling.
license	MIT
metadata	{"author":"Tachera Sasi","repository":"https://github.com/ekilie/beamdrop-skills","version":"1.0.0","keywords":"ai, agent, skill, file-storage, s3, presigned-urls"}
references	["references/agent-instructions.md"]

Beamdrop File Storage Skill

When to Use This Skill

Use this skill when the user asks to:

• Upload, download, or manage files on a Beamdrop server
• Create, list, or delete storage buckets
• Generate presigned or shareable download URLs
• Manage Beamdrop API keys (create, list, delete, scope permissions)
• Store AI-generated artifacts (code, images, documents, build outputs) on Beamdrop
• Set up Beamdrop integration in their Go, Python, JavaScript, PHP, or any project
• Compare presigned URL types or decide on a file-sharing strategy
• Set up webhooks for real-time event notifications (object, bucket, share, presign events)
• Connect AI assistants via the built-in MCP server at /mcp
• Debug Beamdrop API errors (401, 404, 409, 429)
• Configure Beamdrop for production deployment

Prerequisites

The user needs a running Beamdrop instance with API authentication enabled:

beamdrop -dir /path/to/share -api-auth

They need these environment variables or config values:

• BEAMDROP_BASE_URL — The server URL (e.g., http://localhost:7777)
• BEAMDROP_ACCESS_KEY_ID — API access key (format: BDK_ + 16 hex chars)
• BEAMDROP_SECRET_KEY — API secret key (format: sk_ + 40 hex chars)

If the user doesn't have API keys yet, they can create them:

curl -X POST http://localhost:7777/api/v1/keys
# Response includes accessKeyId and secretKey (secret shown ONCE — save immediately)

Authentication Details

All S3 API requests at /api/v1/ use HMAC-SHA256 signing:

StringToSign = METHOD + "\n" + PATH + "\n" + RFC3339_TIMESTAMP
Signature = Base64(HMAC-SHA256(StringToSign, SecretKey))

• PATH is the URL path only — no query string. Example: /api/v1/buckets/my-bucket
• TIMESTAMP is RFC3339 UTC (e.g., 2024-01-15T10:30:00Z). Must be within ±15 minutes of server time.
• Base64 is standard encoding (not URL-safe)

Headers required on every request:

Authorization: Bearer BDK_xxxx:SIGNATURE
X-Beamdrop-Date: 2024-01-15T10:30:00Z

API key properties:

• permissions: "read", "write", or "read,write"
• bucketScope: Optional — restricts key to a single bucket
• expiresAt: Optional — key auto-expires after this time
• disabled: Can be set to temporarily disable without deleting

Go Client SDK

When generating Go code, always use the official client SDK at github.com/ekilie/beamdrop/pkg/client:

import "github.com/ekilie/beamdrop/pkg/client"

// Initialize client (validates BaseURL, creates HTTP client with 2-min timeout)
c, err := client.New(client.Config{
    BaseURL:     os.Getenv("BEAMDROP_BASE_URL"),   // Required, must have scheme+host
    AccessKeyID: os.Getenv("BEAMDROP_ACCESS_KEY_ID"),
    SecretKey:   os.Getenv("BEAMDROP_SECRET_KEY"),
    // HTTPClient: &http.Client{Timeout: 5*time.Minute},  // Optional custom client
    // UserAgent:  "my-app/1.0",                           // Optional, default "beamdrop-go-client/0.1"
})
if err != nil {
    log.Fatal(err)  // ErrInvalidBaseURL or ErrMissingCredentials
}

Bucket Operations

// List all buckets → *BucketList{Buckets []BucketInfo, Count int}
buckets, err := c.ListBuckets(ctx)

// Create bucket — errors with 409 BUCKET_EXISTS if already exists
created, err := c.CreateBucket(ctx, "my-bucket")   // → *BucketCreated{Bucket, Created, Location}

// Create bucket (IDEMPOTENT — preferred for automation)
// Returns 201 if new, 200 with Exists=true if already existed
created, err := c.CreateBucketIfNotExists(ctx, "my-bucket")

// Check existence via HEAD (returns false on 404, not an error)
exists, err := c.BucketExists(ctx, "my-bucket")   // → bool

// Delete (must be empty — 409 BUCKET_NOT_EMPTY if objects remain)
err = c.DeleteBucket(ctx, "my-bucket")

Bucket name rules: 3-63 chars, regex ^[a-z0-9][a-z0-9.-]{1,61}[a-z0-9]$, no IP-like names.

Object Operations

// Upload from bytes → *ObjectCreated{Bucket, Key, ETag, Size, URL}
uploaded, err := c.PutObject(ctx, "bucket", "path/to/file.txt", []byte("content"))

// Upload from io.Reader (streaming — use for large files, avoids loading into memory)
f, _ := os.Open("large-file.bin")
defer f.Close()
uploaded, err = c.PutObjectReader(ctx, "bucket", "large-file.bin", f)

// Download → *ObjectBody{ObjectMetadata, Body []byte}
obj, err := c.GetObject(ctx, "bucket", "path/to/file.txt")
fmt.Println(string(obj.Body))                    // file content
fmt.Println(obj.ContentType)                     // "text/plain"
fmt.Println(obj.ETag)                            // MD5 hex hash (quotes trimmed)

// Metadata only (no body download) → *ObjectMetadata
meta, err := c.HeadObject(ctx, "bucket", "path/to/file.txt")
fmt.Printf("Size: %d, Type: %s\n", meta.ContentLength, meta.ContentType)

// Check existence (false on 404, not an error)
exists, err := c.ObjectExists(ctx, "bucket", "key")

// Delete
err = c.DeleteObject(ctx, "bucket", "path/to/file.txt")

// List objects with S3-style prefix/delimiter → *ObjectList
list, err := c.ListObjects(ctx, "bucket", client.ListObjectsOptions{
    Prefix:    "folder/",    // only keys starting with "folder/"
    Delimiter: "/",          // group into "directories" at "/"
    MaxKeys:   100,          // default 1000
})
// list.Contents → []ObjectInfo{Key, Size, LastModified, ETag, ContentType}
// list.CommonPrefixes → []CommonPrefix{Prefix} (virtual subdirectories)
// list.IsTruncated → bool (more results available)

Object key rules: max 1024 bytes, no .., no leading /. Forward slashes create virtual directory hierarchies. Max upload size: 5GB. Writes are atomic (crash-safe). ETag = MD5 hex hash of content.

Presigned URLs — Choosing the Right Type

Beamdrop has two presigned URL mechanisms. Always choose deliberately — they have different trade-offs:

Client-side HMAC presigned URLs

Generated locally using your secret key. No server API call needed. Self-contained URL.

url, err := c.PresignObjectURL("GET", "bucket", "file.txt", time.Now().Add(24*time.Hour))
// Returns: "http://server/api/v1/buckets/bucket/file.txt?access_key=BDK_xxx&expires=...&token=..."

Use when:

• You need zero server overhead (URL computed locally)
• Generating many links in a batch
• Embedding in automated emails or notifications
• Temporary access with no tracking needed

Limitations:

• Cannot be revoked (valid until expiry)
• No download counting or limits
• Key rotation invalidates ALL outstanding client-side URLs
• URL is long (contains bucket/key path + query params)

Server-side presigned URLs (RECOMMENDED for most use cases)

Created via API call, stored in database with a short 32-char hex token. Clean /dl/{token} URLs.

presigned, err := c.CreatePresignedURL(ctx, client.CreatePresignedURLRequest{
    Bucket:       "bucket",
    Key:          "file.txt",
    Method:       "GET",
    ExpiresIn:    int64Ptr(3600),    // 1 hour (seconds). Omit for no expiry.
    MaxDownloads: intPtr(10),         // Max 10 downloads. Omit for unlimited.
})
// presigned.URL = "https://server/dl/a1b2c3d4..."  (short, clean)
// presigned.Token = "a1b2c3d4..."
// presigned.DownloadCount = 0

// List all presigned URLs
urls, err := c.ListPresignedURLs(ctx)   // → *PresignedURLList{URLs, Count}

// Check download stats
details, err := c.GetPresignedURL(ctx, "token")
fmt.Printf("Downloaded %d/%d times\n", details.DownloadCount, *details.MaxDownloads)

// Revoke — /dl/token immediately returns 404
err = c.DeletePresignedURL(ctx, "token")

Use when:

• You need to revoke access after sharing
• You want download counting or limits
• You need an audit trail (createdBy, createdAt tracked)
• Link must survive API key rotation
• You want short, clean URLs for users
• Sharing sensitive files

Decision matrix:

Scenario	Client-side	Server-side
Quick temporary link, no tracking	✅
Need to revoke after sharing		✅
Limit number of downloads		✅
Track download count		✅
Batch-generate 1000 links	✅
Share in download portal		✅
Must survive key rotation		✅
Sensitive files with audit trail		✅
Embed in automated emails	✅

HTTP API Quick Reference

When generating code in other languages, use the HTTP API directly with HMAC signing:

Buckets

• GET /api/v1/buckets → {"buckets":[{name, createdAt}], "count":N}
• PUT /api/v1/buckets/{name} → 201 {bucket, created, location} | 409 BUCKET_EXISTS
• PUT /api/v1/buckets/{name}?createIfNotExists=true → 201 (new) or 200 {exists:true} (existed)
• HEAD /api/v1/buckets/{name} → 200 or 404
• DELETE /api/v1/buckets/{name} → 204 | 409 BUCKET_NOT_EMPTY

Objects

• PUT /api/v1/buckets/{bucket}/{key} (body = raw bytes) → 200 {bucket, key, etag, size, url}
• GET /api/v1/buckets/{bucket}/{key} → raw file (headers: Content-Type, Content-Length, ETag, Last-Modified). Supports Range header
• HEAD /api/v1/buckets/{bucket}/{key} → headers only, no body
• DELETE /api/v1/buckets/{bucket}/{key} → 204
• GET /api/v1/buckets/{bucket}?list=true&prefix=X&delimiter=/&max-keys=N → {contents, commonPrefixes, isTruncated}

Presigned URLs

• POST /api/v1/presign (JSON: {bucket, key, method, expiresIn, maxDownloads}) → 201 {token, url, ...}
• GET /api/v1/presign → {urls, count}
• GET /api/v1/presign/{token} → presigned URL details with current downloadCount
• DELETE /api/v1/presign/{token} → 200 (immediate revocation)
• GET /dl/{token} → public download (no auth needed). 404 if expired/revoked/max-reached

API Keys

• POST /api/v1/keys (JSON: {name, permissions, bucketScope}) → 201 {accessKeyId, secretKey, ...} — secret shown ONCE
• GET /api/v1/keys → {keys, count} — no secrets
• DELETE /api/v1/keys?accessKeyId=BDK_xxxx → 204

Error Handling

All errors return consistent JSON: {"error":{"code":"CODE","category":"CATEGORY","message":"...","details":{}}}

Handle these errors in generated code:

HTTP	Code	What to Do
400	INVALID_BUCKET_NAME	Fix bucket name: 3-63 chars, lowercase a-z0-9 + hyphens/dots, start/end with letter/digit
400	INVALID_OBJECT_KEY	Fix key: no .., no leading /, max 1024 bytes
401	UNAUTHORIZED	Check: (1) API key exists and not disabled, (2) timestamp within ±15 min, (3) signing path matches request path
403	PERMISSION_DENIED	API key lacks required permission or bucket scope doesn't match
404	BUCKET_NOT_FOUND	Create bucket first with CreateBucketIfNotExists
404	OBJECT_NOT_FOUND	Object doesn't exist — check key spelling, verify bucket
409	BUCKET_EXISTS	Use ?createIfNotExists=true to avoid this
409	BUCKET_NOT_EMPTY	Delete all objects before deleting bucket
413	FILE_TOO_LARGE	File exceeds 5GB limit — split or compress
423	OBJECT_LOCKED	Another operation holds the lock — retry after brief delay (lock timeout is 30s)
429	RATE_LIMIT_EXCEEDED	Retry after Retry-After header seconds. Response has X-Retryable: true. General: 100/min, upload: 10/min
507	STORAGE_FULL	Server reached -max-storage limit — cannot upload

Go client error handling pattern:

result, err := c.GetObject(ctx, "bucket", "key")
if err != nil {
    var apiErr *client.APIError
    if errors.As(err, &apiErr) {
        switch apiErr.Code {
        case "OBJECT_NOT_FOUND":
            // handle missing object
        case "RATE_LIMIT_EXCEEDED":
            time.Sleep(time.Duration(apiErr.RetryAfter) * time.Second)
            // retry
        }
    }
}

Common Workflows

Store AI-generated artifacts

c.CreateBucketIfNotExists(ctx, "ai-artifacts")

// Use structured key paths: {purpose}/{session}/{filename}
key := fmt.Sprintf("generations/%s/%s", sessionID, "output.json")
c.PutObject(ctx, "ai-artifacts", key, resultBytes)

// Generate a shareable link (server-side, revocable, 24h expiry)
presigned, _ := c.CreatePresignedURL(ctx, client.CreatePresignedURLRequest{
    Bucket:    "ai-artifacts",
    Key:       key,
    Method:    "GET",
    ExpiresIn: int64Ptr(86400),
})
fmt.Println("Download:", presigned.URL)

Upload and share with download limits

c.CreateBucketIfNotExists(ctx, "shared")
c.PutObject(ctx, "shared", "report.pdf", pdfBytes)

// Server-side URL — max 5 downloads, expires in 7 days
presigned, _ := c.CreatePresignedURL(ctx, client.CreatePresignedURLRequest{
    Bucket:       "shared",
    Key:          "report.pdf",
    Method:       "GET",
    ExpiresIn:    int64Ptr(7 * 24 * 3600),
    MaxDownloads: intPtr(5),
})

List and organize files

// List top-level "directories" in a bucket
list, _ := c.ListObjects(ctx, "bucket", client.ListObjectsOptions{
    Delimiter: "/",
})
// list.CommonPrefixes → ["folder1/", "folder2/"]

// List files within a "directory"
list, _ = c.ListObjects(ctx, "bucket", client.ListObjectsOptions{
    Prefix:    "folder1/",
    Delimiter: "/",
})
// list.Contents → [{Key:"folder1/file.txt", ...}]
// list.CommonPrefixes → ["folder1/sub/"]

Deduplicate with ETags

// ETags are MD5 hashes — use for deduplication
meta, _ := c.HeadObject(ctx, "bucket", "file.txt")
if meta.ETag == computeMD5Hex(newContent) {
    fmt.Println("Content unchanged, skipping upload")
} else {
    c.PutObject(ctx, "bucket", "file.txt", newContent)
}

Scoped API keys for limited access

# Create a read-only key scoped to one bucket
curl -X POST http://localhost:7777/api/v1/keys \
  -H "Content-Type: application/json" \
  -d '{"name":"readonly-reports","permissions":"read","bucketScope":"reports"}'

Validation Rules Summary

Rule	Constraint
Bucket name length	3-63 characters
Bucket name regex	^[a-z0-9][a-z0-9.-]{1,61}[a-z0-9]$
Bucket name chars	Lowercase a-z, digits 0-9, hyphens, dots
Bucket name start/end	Must be letter or digit
IP-like bucket names	Rejected (e.g., 192.168.1.1)
Object key max length	1024 bytes
Object key no-go patterns	Empty, contains .., starts with /
Max upload size	5GB (5,242,880,000 bytes)
HMAC clock skew tolerance	±15 minutes
Object lock timeout	30 seconds
Rate limit — general	100 req/min per IP
Rate limit — upload	10 req/min per IP
Rate limit — auth	5 req/min per IP