AEGIS · Agentic Infrastructure · Base • 2025–26

When agents act autonomously, someone still has to answer for it.

AEGIS is the supervision layer for autonomous DeFi agents. As sole designer through Base Batches 003, I designed the interface that gives protocol teams real-time visibility into what their agents are doing, why they're doing it, and the controls to intervene when they shouldn't be.

AEGIS · Planning Preview · Helios awaiting approval

4:47 until execution
Helios · Yield Optimizer
Rebalance $2.4M USDC — Morpho to Compound
+0.34% APY improvement · 3-tx sequence · gas 0.024 ETH · net gain +$81,600/yr
Compound utilization 78% — elevated
All guardrails satisfied · confidence 91%
Approve
Modify
Reject
Alternatives
Morpho → Compound · +0.34%
Morpho → Spark · +0.21%
No action · baseline
Guardrails
Max $5M   Slippage 0.15%

Agents

3 / 3

Capital

$2.4M

Interventions

12

Uptime

99.8%

The Problem

Web3 teams are deploying AI agents that rebalance positions, execute liquidations, and manage treasury around the clock. The moment a team deploys one, nobody knows what it's doing or why. They find out when TVL drops or a panicked Telegram message lands at 3am. Execution in DeFAI is solved. Oversight isn't.

Execution Timeline

The execution timeline is a live feed of what's executing, what's waiting for review, and what's been overridden. Colored left edge communicates status before a word is read.

Execution Timeline · All agents

Live
Just now
Helios
Rebalancing USDC — Morpho to Compound
$2.4M · 3-tx sequence
Executing
$2,400,000
2m ago
Helios
Awaiting approval — counterparty threshold
Compound exposure would reach $4.8M vs $5M guardrail
Pending
$600,000
7m ago
Nemesis
Operator override — liquidation halted
Risk team manual review · agent paused
Overridden
$1,200,000
14m ago
Kronos
Collateral ratio restored
148% → 163% · all guardrails satisfied
Confirmed
$180,000
31m ago
Kronos
Monthly fee sweep complete
6 pools · Uniswap · Curve · Balancer
Confirmed
14.2 ETH

↑ Click any row to expand reasoning. Every decision reconstructable.

Planning Preview

Before an agent executes, it surfaces the full plan: move size, price impact, risk factors, alternatives considered. There's a countdown. The clock makes the stakes real.

Planning Preview · Helios

Awaiting approval
Helios · Yield Optimizer · Awaiting approval

Rebalance $2.4M USDC — Morpho to Compound

Executes in
4:47
Helios identified a +0.34% APY improvement by moving 2.4M USDC from Morpho (4.12%) to Compound (4.46%). Gas: 0.024 ETH. Net gain annualized: +$81,600.
Move size
$2.4M
Price impact
0.08%
Confidence
91%
Risk factors
Compound utilization at 78% — slightly elevated.
Morpho exposure 34% below guardrail. No policy breach.
Gas at 18 gwei — within ceiling.
Confidence
91% · High✓ Within policy
Alternatives considered
Morpho → Compound ·
+0.34% APY
Selected
Morpho → Spark
+0.21% APY
No action
Baseline
Active guardrails
Single position max: $5M
Slippage tolerance: 0.15%
Gas ceiling: 25 gwei

↑ Interactive — approve, modify, or reject. The countdown is live. Inaction is a decision.

Guardrails

Guardrails define when an agent can act without asking. The principle: readable by a risk manager, not just an engineer.

Guardrail Status · All agents

1 breach
Kronos · Treasury
Single transaction cap
Limit: $500k / txActive
Current$124k
Helios · Yield
Single counterparty exposure
Limit: $5M / protocolApproaching
Current$4.2M / $5M
Helios · Yield
Slippage tolerance
Limit: 0.15%Breached
Current0.23% — halted
Nemesis · Risk
Monthly hedge budget
Limit: $200k / moActive
Current$87k
Kronos · Treasury
Gas ceiling
Limit: 30 gweiActive
Current18 gwei
Nemesis · Risk
Approved counterparties
Limit: 7 protocolsActive
Current4 / 7

↑ Green: nominal. Amber: approaching. Red: breached. No color is decorative.

Override & Alerts

When an agent breaches a guardrail or behaves anomalously, someone needs to act fast. The override layer gives operators three levels of intervention: pause a single agent, pause all agents, or emergency stop. The design challenge was making these controls accessible without making them easy to hit accidentally. A misclick on emergency stop is its own kind of incident.

Override Controls

1 paused
Kronos
Treasury Management
Active
Helios
Yield Optimization
Active
Nemesis
Risk Hedging
Paused
Emergency stop — all agents
Halts all agents simultaneously

Anomaly & Alerts

3 unacked
critical
Helios slippage guardrail breached
Attempted swap at 0.23% slippage — above 0.15% limit.
7m ago
high
Nemesis paused by operator
Manual intervention during ETH volatility.
14m ago
medium
Helios counterparty exposure approaching limit
Compound allocation at 84% of $5M guardrail.
22m ago
low
Kronos gas ceiling advisory
Network gas spiked to 28 gwei.
1h ago

Explainability

The explainability layer reconstructs the full decision chain: why this action, what alternatives were considered, what the confidence level was. That's what trust actually looks like: earned confidence, not blind faith.

Explainability · Nemesis · Decision reconstruction

94% confidence
Collateral position protected — volatility responseConservative

Nemesis detected volatility and executed protective collateral increase.

1
Volatility spike detected — ETH/USD -4.2% in 6 minutes
Trigger: >3% move in <10min window.
2
Collateral ratio check — current 148%, target 160%+
Ratio drops to 141% if unchecked. Policy floor: 135%.
3
3 strategies evaluated — minimum-cost option selected
Selected C — add collateral $180k, preserves position.
4
Executed — $180k USDC added, ratio restored to 163%
Gas: 0.008 ETH · Slippage: 0.02% · Duration: 14s.
Alternatives scored
C — Add collateral $180k · Selected
94%
B — Options hedge $12k premium
71%
A — Partial position unwind $400k
48%

Design Decisions

Countdown over hard deadline

Three iterations to get this right. No countdown: agents executed unreviewed. Hard deadline: operators panic-rejected. Escalating countdown: operators started checking earlier, calmly. The timer shapes behavior more than any button on the page.

Human language over parameter syntax

Guardrails were originally config values. A risk manager who couldn't parse them was the turning point.

Explainability added late

Without reasoning traces, AEGIS was a control panel. With them, operators stopped reacting to outcomes and started understanding the system.

Outcomes

Three protocol teams in testing, each with a live oversight failure.

<90s

Anomaly detection time in live testing, down from "whenever someone noticed."

3/3

Testing teams named Planning Preview the screen they couldn't work without.

3

Protocol teams in live testing with real agent failures and real capital.

12

Agent interventions triggered during testing. 4 automated, 8 operator-initiated.

“For the first time I feel like I actually understand what our agents are doing. Not just what they did, but what they're thinking.”

— Protocol operator, Base Batches 003

Reflection

The first version of the Planning Preview had no countdown. Operators treated it like a notification. They'd check it eventually. Agents executed before anyone reviewed. The second version had a hard deadline with a red timer. Operators panicked and rejected plans they should have approved. The third version, the one that shipped, uses a calm countdown with escalating visual weight. Operators started checking earlier without the stress response. That calibration took three iterations and taught me something about designing for autonomous systems: the interface isn't just showing what the agent will do. It's shaping how the human responds. Get that wrong and the oversight layer becomes either ignored or adversarial.