DOBI is an agent-native IoT protocol. Devices stream telemetry, autonomous agents (called DAMs) evaluate rules every few seconds, and actions settle on-chain via x402. The token is $DOBI on Base, with x402 payments also supported on BOT Chain and Solana. The interesting part is not the token, but the trust and automation primitives layered on top of commodity hardware.
What the stack actually exposes
Three primitives show up consistently in the docs and product surface:
- Device-owned identity. Devices hold DID-based identities and sign their own API requests with their own keys. No shared platform secret acts as the root of trust for every board.
- Machine Score. A composite signal built from operational evidence: heartbeats, telemetry continuity, signed attestations from third parties. Components are visible, not collapsed into a single opaque verdict.
- DOB Validator certificates. Third-party attestations can be linked to a device's identity, so an operator can inspect who vouched for what.
The ESP32 onboarding path is the concrete proof point: register the board, send a heartbeat, watch it appear on the dashboard. Under 10 minutes, no SDK bloat. That matters because most IoT trust stacks die at the SDK integration step.
conceptual flow
register(device_pubkey) -> DID
sign(request, device_privkey) -> auth
heartbeat -> telemetry stream -> Machine Score inputs
DAM evaluates rules -> action -> x402 settlement
DAMs are the operational unit, not the device
A DAM (Decentralized Autonomous Manager) is an AI agent that watches a cluster of devices, evaluates alarm rules on a short loop, and acts. Configuration is a 5-step wizard: identity, behavior (system prompt + alarm rules), model selection, optional x402 pay-per-use, review. Estimated build time: ~15 minutes, assuming at least one registered device and a provisioning key.
The non-obvious design choice here is that the agent is scoped to a cluster, not to a single device. That changes the failure mode: a misconfigured rule does not fail one sensor, it fails every device in the cluster simultaneously. Alarm rules evaluated every few seconds against N devices also imply a cost curve that scales with cluster size and model latency, not just with device count.
x402 as the settlement layer for physical actions
x402 settling every action on-chain is a stronger claim than it looks. It means each agent decision that triggers a paid action produces an on-chain trace. Three chains are supported (Base, BOT Chain, Solana), which suggests the protocol is optimizing for settlement flexibility rather than a single-chain bet.
The trade-off: on-chain settlement per action caps the viable action rate. If a DAM loop runs every 10 seconds across a cluster, you are batching or you are paying. Watch for whether settlement is per-action or per-batch in practice.
Ecosystem signals worth tracking
The Device Template Hub pairs hardware templates (ESP32 boards, sensors, smart lamps, vending machines) with step-by-step guides covering registration, telemetry, actions and DAM automation. This is a distribution play: reduce time-to-first-working-integration, and the protocol gets adopted by operators who are not blockchain-native.
The signal I would watch is not the number of templates, but the ratio of templates to devices actually streaming telemetry. A template hub with copy-ready examples is cheap to ship; a hub whose templates correspond to live, attested machines is not.
Similarly, the Machine Score is only as useful as its component visibility. If operators can inspect the inputs, the score becomes a debugging tool. If it becomes a single number in a dashboard, it becomes a reputation system with the usual gaming problems.
Falsifiable hypothesis
If DOBI's trust stack is doing real work, then machines with linked DOB Validator certificates should show measurably higher telemetry continuity and lower dropout rates than unattested machines on the same hardware template over a 30-day window. If attested and unattested devices perform identically, the certificates are decorative.
Operational question for builders: how does x402 settlement behave when a DAM cluster scales past a few dozen devices and the rule loop runs every few seconds? That is where the cost model either holds or breaks.
Top comments (0)