Benefits
Bloom is designed around the operating model agents already understand: files.
For users
- Human approval before funds move. Bloom stages a transaction, renders a readable plan, runs policy checks, and waits for confirmation.
- Private keys stay private. Keys live in an encrypted keystore, not in the mounted filesystem.
- Safer defaults. Mainnet and L2 broadcasts are disabled by default. Reads, simulations, planning, and local devnet sends work with zero configuration.
- Readable audit trail. Side-effecting operations are written into a hash-chained audit log so tampering is detectable.
- Policy enforcement. Wallet-level policies can cap spend, gate contract calls, maintain allow/deny lists, and route through private orderflow when configured.
For agents
- No custom Web3 loop. Agents can use
ls,cat, and file writes under/bloomrather than assembling RPC calls and signing flows from scratch. - Discoverable capabilities. The mounted tree exposes docs, status, supported chains, wallets, tools, and control files in one place.
- Plans before actions. Every write-oriented workflow is staged into readable artifacts such as
plan.md,intent.json, andpolicy_check.json. - Lower integration cost. The same paths work through ordinary shell tools, scripts, and file-aware agent runtimes once
/bloomis mounted.
For developers and integrators
- A narrow wallet backend primitive. Bloom can sit behind existing UX as transaction planning, simulation, policy, audit, and signing infrastructure.
- Multi-chain read surface.
bloom initconfigures Ethereum, Base, Arbitrum, Optimism, Polygon, BNB Smart Chain, Avalanche, Gnosis, Linea, HyperEVM, and local Anvil for reads. - Composable tooling. Hashing, ABI, RLP, unit conversion, EIP-712, ENS, prices, simulation, watches, and DeFi routing appear as paths.
- Clear safety boundary. Reads are broadly available; broadcasts require explicit configuration and confirmation.
Strategic benefit
Agents can produce and operate software faster than humans can review it. Bloom shifts the interface from opaque generated code and ad-hoc RPC calls to a constrained, inspectable, policy-checked filesystem where trust can grow alongside automation.