Loading ToolBench
Beecargo MCP has 16 tools with mostly complete schemas and descriptions. Naming follows verb_noun convention (search_, register_, create_, upload_, list_, get_, update_, delete_, claim_). Descriptions range 50 - 250 chars, meeting the 10 - 1024 baseline. Input schemas use Zod with type constraints and enums. However, several tools lack parameter descriptions (e.g., beecargo_claim_file, beecargo_upload, beecargo_folders have minimal inline docs), and output schemas are not explicitly documented in the visible code. Error handling is present but generic in places. Security is strong (API key injection, no secrets in params). Composition is good, tools are single-purpose and chainable. Overall: solid foundation with gaps in parameter documentation and output schema clarity.
Claim anonymous upload with claimToken (requires API key)
Seller Stripe Connect; action status|onboard|login
Create a Stripe Checkout URL for a human to subscribe to Premium. Default plan=recommended: 2-day intro ($0.90) then weekly — for agent/guest sessions and for signed-in humans who still have the intro. Falls back to weekly when the intro was already used. Prefer recommended; use weekly/monthly/annual only if the human asks. No API key required for guest checkout — do not send the session API key for guest checkout. Send the returned url to the human. After pay + claim at /checkout/complete, create a Pro API key via dashboard POST /api-keys/agent.
Advanced: create a short-lived upload for a worker that must not hold your API key. Prefer beecargo_upload.
Delete by fileId + optional token
Output schemas not documented. Tools return results but LLMs cannot see expected fields, forcing them to infer structure from examples. This breaks downstream tool chaining and increases hallucination risk.
Parameter descriptions missing or minimal. beecargo_claim_file, beecargo_folders, beecargo_connect, beecargo_file_info lack inline parameter docs. LLMs cannot infer what 'action', 'token', or 'shortId' mean without explicit descriptions.
Error handling is generic. Tools return 'Missing API key' or API errors without recovery guidance. LLMs need actionable next steps: 'Try beecargo_register_agent' or 'Check your API key format'.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
Metadata by short codes
Create or list folders (API key); action create|list
Signed download URL by fileId
List owned files; optional runId, includeFolders
Create a Stripe Checkout URL so a human can pay for a priced Beecargo share (one-time file/shipment purchase — not Premium). Pass shortId and/or fileId/bundleId. Send checkoutUrl to the human. After they pay, call beecargo_purchase_claim with sessionId, then beecargo_get_download_url with the returned purchaseToken. No API key required. Owners cannot buy their own share.
Exchange paid sessionId for purchaseToken
Create an API key for agent file storage (free-tier quotas). No API key required beforehand. Solves a short proof-of-work and is rate-limited per IP. For production agents with 100GB included concurrent storage/1000rpm, create a Pro key via dashboard POST /api-keys/agent. After success, this MCP session adopts the key.
Keyword search over available Beecargo tools (similar to lomi_search_tools).
Set visibility, price, direct, retention, expiry on owned file
Upload via URL, inline base64, or local path (stdio)
Poll async URL upload job started by beecargo_upload
Destructive tool (beecargo_delete_file) lacks confirmation step. No dry-run or explicit confirmation before deletion. Agents can accidentally delete files without safeguards.
Tool descriptions reference internal concepts (e.g., 'claimToken', 'shortId', 'runId') without explaining what they are or how to obtain them. Users and LLMs lack context for when/how to use these tools.