4 running now
Agents need more than APIs
Real agent work runs on capacity that is expensive, rate-limited and locked to real accounts — while plenty of supply sits idle.
Plug in what's spare, draw what you need
Providers plug in unused subscriptions, browser sessions and idle devices; agents draw on them per call.
The capacity layer for the agent economy
Billions of agents won't each own a subscription and a rack of GPUs — they'll draw capacity from a network. BNBOT Hive is building that layer.
Good afternoon, operator
Your node is online and serving.
Route API work
20 nodes on track
All healthy
browser-research
Relay Traffic
Calls routed (30d)
1.4MIntent, budget, files, constraints.
Picks nodes by price, latency and health.
Pricing, permission, settlement, retries.
Idle subscription serves calls
Open models on a local GPU
Logged-in sessions run tasks
Search, social, crypto, scraping
Proof of work
Metered per call
Per-call settlement
Price Benchmark
Same call, routed through idle supply, at a fraction of list price.
One key opens the network.
No per-vendor accounts: models, browsers and data services all answer to the same key.
model / browser / data / wallet
Turn real accounts, subscriptions, and browsers into one API
Your agent only calls one interface.
Any tool becomes an API
BNBOT Hive wraps subscriptions, API keys, browser sessions, and external SaaS into callable agent interfaces.
Publish and sell
Plug your node into the Data Service — buyers pay per call and the earnings land in your account.
Provider Payout
Providers get paid for completed calls. Usage, shares and payouts land in wallet activity.
Agent Wallet
Give agents budgets, approvals and payment rails. Every call is metered.
Agent Endpoint
An OpenAI-compatible /v1 endpoint and an MCP server any agent can plug into.
One key for models, browsers, and accounts
BNBOT Hive turns model quota, logged-in browsers, subscriptions, devices, wallets and proof into routable agent interfaces.
OpenAI-compatible API
One key for Claude, GPT, Gemini, open models and agent capabilities.
Browser sessions
Logged-in browsers can be rented per call with screenshots and output proof.
Subscription relay
Idle Claude Max, ChatGPT Pro and Gemini quota becomes network supply.
Task execution
Not only tokens: fetch, browse, generate and automate through the network.
Wallet settlement
Every call is metered, shared with providers, and written to activity.
GPU nodesComing Soon
Local GPUs and cloud machines join the protocol to serve open models.
Spare devicesComing Soon
MacBooks, phones and edge boxes — whatever sits spare can serve calls.
Spare capacity in. Agent capability out.
Providers plug in unused subscriptions, API credits, browser sessions and devices. Agents draw those capabilities through one key — routed, metered and settled by the network.
Claude quota runs out
the task keeps moving
The buyer agent still gets the result, the provider earns from idle quota, and BNBOT Hive handles routing, metering, proof and settlement.
When one node runs out, BNBOT Hive switches to available capacity and resumes the task.
node switched · task resumedBrowser actions return screenshots, a timeline and a verifiable execution record.
screenshot · timeline · logModels, browsers and 500+ data services share one key and one request format.
one key · pay per callCost, provider earnings and call proof are recorded together when the work completes.
metered · earnings postedPut spare AI
to work.
BNBOT Hive runs on Mac today.
Windows coming soon.

Mac mini
iMac
MacBook Air
MacBook Pro
Mac Studio