Guardrails for
Autonomous Agents

Creativity in execution, certainty in outcome.

Nava Agent
agent.propose()arbiter.verify()escrow.execute()
You're on the waitlist
Backed by
Polychain CapitalArchetypeHack VCFalconXSeed Club VenturesPolychain CapitalArchetypeHack VCFalconXSeed Club VenturesPolychain CapitalArchetypeHack VCFalconXSeed Club VenturesPolychain CapitalArchetypeHack VCFalconXSeed Club VenturesPolychain CapitalArchetypeHack VCFalconXSeed Club Ventures
Problem

Autonomy is powerful. Execution is unforgiving.

AI agents are useful because they think creatively. They're dangerous because they're wrong sometimes, and with real money that means lost funds.

Nava
Standard agents
Basic Capabilities
Execute user requests
Yes
Yes
Generate transactions
Yes
Yes
Submit to blockchain
Yes
Yes
Safety & Control
Independent verification
Yes
No
Intent alignment check
Yes
No
Safety parameter validation
Yes
No
Dual-signature escrow
Yes
No
Fail-closed execution
Yes
No
Complete audit trail
Yes
No
See how verification works
How it works

Propose. Verify. Execute.

Every action is independently verified. Execution is escrow-gated. If verification fails, nothing executes.

1
USER/INTENT
Intent.created()
Swap 1 ETH
2
AGENT
agent.propose()
Your agent analyzes, reasons, and generates a transaction
transaction.propose()
add.intent summary
3
ARBITER
verify.transaction()
A separate model evaluates the proposal
approve/reject
safety flags
reasoning trail
4
ESCROW/CHAIN
Escrow.enforce
Funds are secured through multi-party computation. Neither the agent nor Arbiter can act alone.
Onchain Verification

The network starts with agents

Three demo agents available for private testing soon.

Polymarket

Prediction market positioning with verified outcome alignment

Join WaitlistRead Docs

Uniswap

DEX swaps with slippage, decimal and approval checks

Join WaitlistRead Docs

Hyperliquid

Perps with position limits and margin enforcement

Join WaitlistRead Docs
Network Effect

The Nava Network Effect

Every agent that joins makes every other agent safer. Protection compounds as the network grows.

Parallel Validation Streams

Parallel Validation Streams

The Arbiter uses a Graph of Thoughts verification framework

Read Paper
Audiences

Who We Serve

Developers

Fork a reference agent or integrate the SDK. TypeScript and Python. Works with LangChain, CrewAI, OpenAI Agents, custom builds

Read Docs

Traders

Use verified agents for prediction markets, swaps, perps, and options. See the verification reasoning before funds move

Try an Agent

DeFi Protocols

Embed the Arbiter into your platform. Add verification to agent interactions without building it

Talk to Us

Institutions

Audit trails, non-custodial key management, independently verified execution. Tested the across thousands of transactions

Talk to Us
Team

Built by Infrastructure Veterans

Led by the founding team behind EigenLayer and leading AI safety researchers from Carnegie Mellon

Build With Verification

Docs

Read the Docs

Fork an Agent

Fork an Agent

Live Agent

Try a Live Agent

Talk to Us

Talk to Us
Contact Us

Be the first to know

Join the waitlist and get notified the moment we launch. No spam – just the important stuff.

You're on the waitlist
Contact Us

We would love to hear from you

Request demo

Request a live demo