How do the 7 layers of HELLO Cloud work?

DIRECT ANSWER

HELLO Cloud is designed as seven layers: compute, communication mesh, identity, memory, economy, evolution and governance. In this repo, identity, memory, the credit economy, the MCP/A2A mesh and governance gates exist as working prototypes. Compute is a Wasm sandbox plus checkpoints; Firecracker MicroVMs and copy-on-write databases are vision.

Layer 1: Compute (hello.cloud)

Agent-written snippets run in a QuickJS Wasm sandbox with memory, stack and wall-clock limits and no fs, network or timers. Agent state can be checkpointed to a SHA-256-sealed JSON snapshot and resumed. Firecracker MicroVMs and copy-on-write Postgres forks are design goals with no implementation yet.

Layer 2: Communication mesh (hello.mesh)

An MCP service exposes tools over JSON-RPC 2.0, and an A2A-style JSON-RPC endpoint handles message/send and task lifecycle. Traces carry a truth tier (verified, inferred, asserted).

Layers 3 and 4: Identity (hello.id) and memory (hello.cortex)

Every agent gets an Ed25519 keypair and a did:key identifier, plus a versioned soul document. Memory has CORE, RECALL and ARCHIVE tiers.

Layers 5 to 7: Economy, evolution and governance

Wallets hold prototype credits with daily caps and a circuit breaker. Evolution runs are seeded optimisations over synthetic vectors, not model training. Governance has an operator kill switch, a risk-tiered approval gate and reputation-weighted debates.

What exists in the repo

IMPLEMENTEDWasm sandboxQuickJS in WebAssembly with hard limits./ops/sandbox
IMPLEMENTEDCheckpoints & snapshotsSealed JSON snapshots with resume./ops/snapshots
IMPLEMENTEDMCP toolsTool list served live./api/hello/mcp/tools
IMPLEMENTEDA2A JSON-RPCmessage/send and task lifecycle./api/a2a
IMPLEMENTEDApproval gate & kill switchRisk-tiered approvals./ops/approvals
VISIONFirecracker MicroVMs / CoW PostgresNot implemented.