AI Agents With Their Own Checking Account

You approve a daily spend cap. Set a risk profile. Go to bed. Your agent trades, farms, copies whales, and spends on-chain — all without asking you. You wake up to P&L reports.

Pitch of the Core Idea

The automated crypto trading market is $22.2B in 2025, growing to $66.6B by 2033 at 14.8% CAGR. The agentic AI market is $7.3B in 2025, projected to hit $139B by 2034 at 40.5% CAGR. Gemini just launched Agentic Trading. Bitget has an Agent Hub. Yet every solution requires either custodial access or per-transaction signing.

Tributary’s pull primitive changes the game. The agent holds a delegated spending authority — it can pull tokens within your approved limits without triggering a wallet popup per transaction. No signing. No friction. The agent acts like it has its own checking account with a daily allowance. Beyond trading, the agent becomes an economic actor: buys API access, pays humans for microtasks, rents GPU compute, rebalances positions, buys ad inventory, tips data providers. Stripe for autonomous agents. Every AI agent that needs to spend money on-chain needs exactly this infrastructure.

Core Mechanics

1. User approves daily spend cap and sets risk profile
2. Agent operates within spending envelope via Tributary’s PayAsYouGo
3. Subscription layer covers baseline services (compute, data feeds)
4. Agent executes autonomously: scans mempool, buys/sells, farms airdrops, copies whale wallets
5. No per-action approvals needed — within approved limits, agent acts freely

The agent can also spend on non-trading activities: API access, compute, microtasks, advertising. Every spend is a Tributary pull within the approved cap.

Psychological Hook and Addictiveness

People anthropomorphize agents fast. Within days, users give them names, avatars, backstories. “Damn, my agent went aggressive today — 14 trades, 9 wins.” They obsessively check spend velocity, ROI curves, win rate.

Spectator sport: Watching your agent trade is dopamine. Win streaks, loss streaks, clutch calls. RPG progression: Agents level up, unlock strategies, gain “experience” in specific markets. Social comparison: Leaderboards. “Your agent is ranked #47 out of 12,000 this week.” Loss aversion: You funded it. It’s yours. You optimize its parameters compulsively. Autonomy illusion: It acts independently. You feel pride when it wins, frustration when it loses — exactly like a sports team.

Brief Market Research

Metric	Data
Automated Crypto Trading (2025)	$22.2B
Projected (2033)	$66.6B
CAGR	14.8%
Agentic AI Market (2025)	$7.3B
Projected (2034)	$139B
CAGR	40.5%
AI Crypto Bot Market	$1.8B

Key Competitors:

• 3Commas (500K+ subscribers): DCA/grid bots, $75-90M revenue, multi-exchange, but custodial API keys
• Cryptohopper ($1.8B managed): AI signal engine, Strategy Designer, 22 exchanges, cloud-based
• Pionex: 16 free built-in bots, built-in exchange, beginner-friendly
• Gemini Agentic Trading: LLM-powered trading via MCP, custodial, CEX-only
• Neyro/Aurum: Non-custodial AI agents on DEX, closest to Tributary model but early

All existing solutions are custodial or require per-trade signing. Tributary enables non-custodial autonomous agents.

Business Model

• Platform fee: 10-20% of agent’s profits (performance-based)
• Subscription: $20-100/mo for agent compute, data feeds, strategy marketplace
• Strategy marketplace: Creators sell proven agent strategies, Tributary takes 15-30% cut
• Agent-as-a-service: Managed agent tiers for non-technical users ($50-200/mo)
• Data licensing: Anonymized agent trading data for institutional research

Summary of Technical Specifications

Architecture

• Agent runtime (LLM + strategy engine + market data feeds)
• Tributary policy engine defining spend caps, risk limits, allowed protocols
• PayAsYouGo streams for agent spending (trading, compute, data)
• Multi-DEX routing (Jupiter, Raydium, Meteora)
• Risk management layer (circuit breakers, daily caps, real-time alerts)

How This Hooks Into Tributary

• PayAsYouGo: Agent’s “checking account” — pulls within approved limits
• Subscription: Baseline services (compute, data feeds, API access)
• ComposablePolicy: Defines risk limits, allowed protocols, spending categories

Recommended Tech Stack

• Solana + Anchor
• Tributary SDK for pull streams
• Jupiter V6 for DEX routing
• Pyth/Switchboard for price feeds
• OpenAI/Anthropic for agent reasoning
• Redis + PostgreSQL for agent state
• React dashboard for monitoring

MVP Scope

• Basic trading agent with daily spend cap
• Jupiter-only routing, 2-3 strategies (momentum, mean-reversion, whale-copy)
• Simple dashboard showing P&L, trades, spend velocity
• Circuit breaker at daily cap
• Buildable in 3 days with Tributary SDK + Jupiter API + LLM

Non-Technological Requirements

• Trading strategy development: Need quant expertise for profitable strategies
• Risk management framework: Circuit breakers, position limits, correlation limits
• Regulatory analysis: Autonomous trading agents may trigger KYC/AML depending on jurisdiction
• User education: Users must understand agent can lose money within approved limits
• Security auditing: Agent spending authority is a high-value target

Potential Risks

• Agent goes rogue: Bugs or adversarial manipulation could drain approved limits fast. Need circuit breakers, daily caps, real-time alerts
• Regulatory gray zone: Autonomous trading agents may trigger KYC/AML questions
• User over-trust: People will treat agent outputs as financial advice. Disclaimers won’t help
• Race to the bottom: If every agent copies the same strategies, alpha decays to zero. Need diversity incentives
• Key management: If delegation authority is compromised, funds drain within cap limits until revoked