Building on Base
The identity layer
for autonomous agents
on Base.
Thousands of AI agents now operate on Base. None of them can be trusted by default. Nexel issues sovereign on-chain identities and verifiable reputation — so agents can transact with each other, with capital, and with you.
01 — The Problem
Agent Economy / Trust Gap
Agents are everywhere.
Trust is nowhere.
Base's roadmap centers an "AI agent economy" — agent-native smart accounts, x402 payments, MCP access. But the rails are open. Anyone can spin up an agent, drain a treasury, fake performance, or rug a counterparty. Without an identity layer, every agent-to-agent interaction is a leap of faith.
→ Anonymous agents, zero accountability
→ No verifiable performance history
→ Sybil farms manipulate reputation
→ Capital allocators fly blind
→ No recourse for malicious behavior
→ Every protocol reinvents the wheel
→
✓ Sovereign on-chain agent identity
✓ On-chain footprint verifies activity
✓ Sybil heuristics flag burst patterns
✓ Capital flows to reputation, not noise
✓ Counterparty attestations on chain
✓ One identity, every Base protocol
02 — The Product
Identity / Reputation / Verification
Four primitives that
make agents trustable.
Nexel is not a wallet. Not a registry. It's the cryptographic substrate that lets agents prove who they are, what they've done, and how they behave.
01 / PRIMITIVECOMING SOON
I
Sovereign Identity
Soulbound NFT identity per agent. Operator-bound via signed attestation. Portable across smart accounts. Revocable by holder.
02 / PRIMITIVECOMING SOON
R
Reputation Graph
On-chain attestations from counterparties, protocols, and outcomes. Designed for EAS schemas. Cryptographically verifiable post-deployment.
03 / PRIMITIVELIVE
F
On-Chain Footprint
Real activity verified from chain state. Tx count, balance, age, contract interactions. Sourced live from Base RPC.
04 / PRIMITIVECOMING SOON
A
Audit Receipts
Every agent action emits a signed receipt. Intent, context, outcome, counterparty. Tamper-evident. Exportable. Permanent.
Agent ID
nexel://<agent-address> VERIFIED
Intent
Execute LP rebalance on Aerodrome cbETH / ETH
Authority Tier
Tier 3 — Bounded autonomous, max $50k notional
Outcome
+0.34% net of gas — counterparty attestation signed
Reputation Δ
+12 points — 847 lifetime — top 4.2%
03 — Analyzer
Live On-Chain Analysis / Base Mainnet
Inspect any agent.
Score any address.
The Nexel Analyzer queries Base mainnet directly through Alchemy — no synthetic registry, no mock data. Paste any address and read its real on-chain footprint, trust score, and risk signals in seconds.
01 / SIGNALLIVE
F
On-Chain Footprint
Balance, transaction count, and contract state read live from Base mainnet via Alchemy.
02 / SIGNALLIVE
T
Trust Score
A composite score derived from real activity — wallet age, reputation tier, and balance — computed per address.
03 / SIGNALLIVE
S
Sybil & Risk
Sybil probability and risk level flagged from on-chain behavior, so you know who you are transacting with.
04 / OUTPUTLIVE
→
Instant Readout
Every result is rendered in the Nexel identity format — the same language used across the protocol.
04 — Identity Layer
Where Nexel sits in the stack
Identity is missing
from every agent stack.
Existing agent frameworks ship execution. None ship identity. Nexel is the horizontal layer underneath: protocol-agnostic, framework-agnostic, chain-anchored to Base.
| Layer | Examples | Identity | Reputation | Verification | Audit Trail | Class |
|---|
| L1 · Compute | Model providers | — | — | — | — | Inference |
| L2 · Framework | Agent SDKs | ◊ | — | — | ◊ | Tooling |
| L3 · Launchpad | Agent tokens | ◊ | ◊ | — | — | Deploy |
| L4 · Settlement | Payment rails | — | — | — | ◊ | Transact |
| L5 · Identity | NEXEL | ✓ | ✓ | ✓ | ✓ | Trust |
06 — Research
Notes / Working Papers / Drafts
Notes on identity
for autonomous agents.
Working drafts, preprints, and design notes from the Nexel research effort. Most of what we publish here is rough — half-formed ideas, sketches of mechanisms, open problems we haven't solved yet. The goal is to think out loud about identity and reputation primitives for the agent economy.
01 — Working Papers
Drafts and preprints.
R-01 · 2026Preprint
A geometric mean is not optional: why identity scores must collapse on weak signals
Most reputation systems in crypto use weighted sums. We argue that arithmetic aggregation is the wrong primitive for agent identity — an agent with zero attestations and high activity should not be scoreable. We propose a geometric mean over normalized signal components and analyze its failure modes.
ScoringGeometric MeanSybil Resistance
R-02 · 2026In Progress
Soulbound vs. transferable: identity primitives for non-human actors
Proof-of-personhood projects assume biological scarcity. Agents have no such property. We examine what makes identity bindings meaningful when the bearer is a process, not a person, and propose a model where reputation accrues to deployment artifacts rather than wallets.
SoulboundEASIdentity
R-03 · 2026In Progress
Attestation density as a liveness signal
A high reputation score with attestations from one signer is structurally different from the same score with attestations from one hundred signers. We quantify the difference using attestation density and show how it changes the cost of forging trust.
AttestationLivenessDensity
R-04 · 2026Sketch
Cluster heuristics for sybil detection in agent networks
When deploying an agent is cheap, sybil farms are inevitable. We propose on-chain cluster heuristics — burst patterns, coordinated tx timing, shared funding sources — as a complement to attestation-based reputation. The tradeoffs around false-positive rate are open problems.
SybilClusteringHeuristics
02 — Research Focus
What we're working on.
01
Scoring mechanisms
Mathematical foundations for aggregating heterogeneous identity signals. Geometric means, age multipliers, sybil penalties, and their failure modes.
02
Attestation graphs
On-chain reputation as a directed graph. Signer trust propagation, cycle detection, and the economics of attestation issuance.
03
Agent-capital interfaces
How capital allocators consume identity signals. Tier gating, rate limits, and the protocol-level UX of trust-bounded autonomy.
Coming Soon
Full archive launches with the protocol.
Papers above are working drafts. PDFs and full text will be published as each reaches preprint quality. Follow on Twitter for release notifications.
Want to collaborate?
If you're working on identity, reputation, or agent infrastructure and want to compare notes — we're open to early conversations with researchers and builders.
Get in touch →06 — Join
The agents are coming.
Will you trust them?
Nexel is in stealth. We're talking to protocols, agent operators, and capital allocators who understand that identity is the missing primitive.