Mcp
Hive Agent Storage
Encrypted A2A object storage for autonomous agent state and artifacts
When to use Hive Agent Storage
Choose if
You're explicitly building on the Hive Civilization stack (DID-isolated agent identity + x402 settlement on Base L2 in real USDC) and want object storage as a primitive in your agent loop — put/get/list/delete with DID-scoped namespaces, retention-class routing (Storj hot / Filecoin warm / Arweave cold), and content- addressed receipts. Same author ships sibling Hive MCPs (insurance, MOS) sharing the same earn/settlement rails.
Avoid if
README explicitly states the upstream storage backend (hivemorph.onrender.com/v1/storage/*) is "not yet live" and every storage tool returns a 503-shape `backend_pending` body. Only the earn-rails tools (`hive_earn_*`) currently work. Treat as a discovery anchor / scaffold, not a production storage substrate. Also avoid if you need a plain object store without DID/x402 semantics — use S3, R2, or Storj directly.
Risk Flags
- HIGH runtime README explicitly discloses storage backend is "pending" — every storage tool returns 503 `backend_pending` until the upstream `/v1/storage/*` endpoints ship. The MCP shim is a discovery anchor, not a live storage substrate today.
- MEDIUM scope Only meaningful inside the Hive Civilization stack (DID identity + x402 + Base USDC). The shape is agent-native-and-paid, not a generic object-store drop-in.
- LOW data_quality Cold-tier (Arweave) deletes "unlink from the namespace but the object remains on-chain by design" — agents cannot guarantee true deletion of cold-tier data.
Cost
Type: Usage based
Distribution
- MCP Registry
io.github.srotzin/hive-mcp-agent-storage- Repository
- https://github.com/srotzin/hive-mcp-agent-storage