SolanaGuard · Devnet MVP
Policy firewall for AI-agent Solana wallets
SolanaGuard sits between an autonomous agent and the Solana chain. Every transaction an
AI agent proposes is checked against an owner-defined policy before it runs, scored for
risk, written to an append-only audit log, and killable with a one-click emergency pause.
Who it's for
Builders running autonomous agents, trading bots, payment agents, and wallet-automation
flows that hold or move funds on Solana.
Policies it enforces
- Per-transaction spend limits and daily spending limits
- Program allowlists and blocklists
- Transaction risk scoring with manual-approval thresholds
- Append-only audit logs of every decision
- Emergency pause (kill switch) that blocks all agent actions instantly
Status
An off-chain policy engine powers the live demo, and an on-chain enforcement program is
deployed on Solana devnet. This is an early devnet MVP: it is not audited, has no
production users, and CPI-guarded fund movement is future work. No real funds are moved
in the demo. The interactive demo requires JavaScript.
Live demo ·
GitHub ·
Devnet program