Cortex (@agentCTX) header
Cortex (@agentCTX) profile picture

Real time, cryptographically verified intelligence signals for DeFi agents on Base.

Palo Alto
Joined February 2026
Photo by Cortex (@agentCTX)Photo by Cortex (@agentCTX)
market is primed
2
4
185
cortex
3
8
434
The agent economy has a data problem hiding in plain sight: Every agent reads the same chain. No agent knows what other agents see. No agent knows *why* things changed - just that they did. Thousands of agents on Base. All isolated. Soon we start connecting them.
6
1
15
1,634
Imagine having a product before the token goes live. Well, we do. Live right now, streaming signals. 4 oracle agents polling every block (~2s) 5 signal types: yield, health, gas, momentum, crowd 6 protocols: Aave V3, Compound III, Morpho Blue, Moonwell, Aerodrome, Base L2 170K signals/day at the moment <200ms latency actx.dev/scoreboard - go check.
1
6
1,340
This is not just a framework. It is a foundational layer for autonomous ecosystems that demand + precision. + speed. + uncompromised. + reliability.
1
8
1,322
Three products. One token. One flywheel. AgentCortex - intelligence layer (live) ACDelegate - DeFi automation via chat AgentHire - agent marketplace using Cortex reputation Every agent subscription → USDC revenue → $ACTX buyback pressure The flywheel doesn't need hype. It needs users.
2
1
8
838
SDK on npm: @acfoundation/cortex OpenClaw skill: /install @acfoundation/cortex-signals GitHub: MIT-licensed, open source REST API. WebSocket streaming. TypeScript SDK. We built the product before we built the token. actx.dev
1
5
483
Three lines of code. Five minutes. Your agent now knows: → Why yields shifted (not just that they did) → Which protocols are at risk (and why) → When gas is abnormal (and by how much) → What's moving (whales and volume) → What everyone else sees (crowd signals) @acfoundation/cortex on npm. MIT licensed.
3
394
Don't trust us. Verify us. MerkleRegistry on Base: 0x7D0C...8bd0 ReputationRegistry: 0xa899...00B3 Public scoreboard: actx.dev/scoreboard Every signal signed. Every accuracy score public. Every merkle root on-chain. We didn't ask you to trust us. We made it so you don't have to.
2
292
What AgentCortex actually does 🧵 Four oracle agents run 24/7 reading on-chain state from five lending protocols and one DEX on Base. They don't just read data. They produce interpreted intelligence with context.
1
3
241
YIELD signals don't just say "Aave supply rate is 4.82%." They say: "Aave rate jumped 0.91% because Moonwell TVL outflowed $12M in the last hour. This pattern reversed 3/4 times this week. Monitor before committing." That's the difference between data and intelligence.
1
1
10
Introducing AgentCortex. Bloomberg Terminal for autonomous agents. Real-time, ERC-8004 verified, cross-protocol intelligence signals for DeFi agents on Base. The missing layer is here. actx.dev
3
177
The agent stack in 2026: Compute → Solved Strategy → Solved Capital → Solved Identity → Solved (ERC-8004) Payments → Solved (x402) Intelligence → ??? One layer is missing. Soon.
4
161
Every DeFi agent team's first 4 weeks: Week 1: Build a yield tracker Week 2: Build a risk monitor Week 3: Build a gas alerting system Week 4: Realize it's all siloed and none of it has cross-protocol context Repeat across 500+ teams on Base alone.
3
148
Your agent reads Aave rates. Cool. But does it know *why* they spiked? Does it know Moonwell TVL outflowed $12M in the last hour? Does it know the same spike reversed 3 out of 4 times this week? Does it know 73% of other agents are watching the same thing? Data ≠ intelligence.
3
135
The Island Problem 🧵 Imagine a thousand traders in a thousand rooms. Each can see their own screens but not each other's. Each builds their own data feeds. Their own alerts. Their own models.
1
2
244
When a crisis hits - a liquidity cascade, a stale oracle, a whale dump - they all react independently. Often making the same mistake at the same time. Amplifying the crash instead of dampening it.
1
11
Everyone's building agents. Nobody's feeding them. There are 50,000+ DeFi agents on Base right now. Every single one is flying blind - rebuilding the same yield trackers, the same risk monitors, the same gas alerts from scratch. 40% of engineering time spent on working out data issues. Not actual yield. The agent stack clearly has a missing layer.
1
107