Solana · Agent infrastructure
Know exactly what your agents did, and why
Relayline makes an autonomous agent declare its intent before it moves funds, then records the declaration and the execution together. You get an audit trail instead of a transaction hash.
Have an audience on X?
250,000 $RLAY per approved post, paid within 7 days of publication.
Connect your account and a post written for your voice publishes to your profile. You can remove it at any time after 7 days.
- 1Log in with your X account
- 2Get your post, written for you
- 3It publishes to your profile
How it works
Proof-of-Intent
Relayline sits between an agent and the chain. The agent submits a signed statement of intent, Relayline compiles that statement into an on-chain constraint, and both the statement and the result are written to Solana together. If the execution does not match the declaration, it does not settle.
01
Declare
The agent signs a structured statement: the action, the counterparty, a value ceiling, and its reason. This happens before any transaction is built.
02
Bound
Relayline compiles the declaration into an on-chain constraint. A swap declared at 400 USDC cannot settle at 4,000.
03
Execute
The transaction runs inside those bounds. Anything outside them reverts rather than partially filling.
04
Attest
The declaration and the outcome are stored as one record, readable by anyone holding the transaction signature.
4
Declaration fields
Action, counterparty, ceiling, reason
~11ms
Added latency
Median, trailing 30 days
3
Framework adapters
LangChain, CrewAI, Solana SDK
Token
$RLAY
Relayline sits between an AI agent and the Solana blockchain. Agents sign a structured declaration covering the action, counterparty, value ceiling and reason; Relayline compiles it into an on-chain constraint that the transaction must satisfy to settle. The declaration and outcome are stored as a single attestation readable from the transaction signature. The $RLAY token is staked by relayers, who are slashed for settling transactions that breach their declared bounds.
Reasons, not just receipts
Every settled action carries the agent's own stated reason, recorded before the outcome was known.
Limits the chain enforces
Spending bounds live in the constraint program, not in your agent's prompt. A jailbroken agent still cannot exceed them.
One trail across frameworks
Agents built on different stacks produce the same record format, so you are not correlating four log systems during an incident.
The questions people actually ask
Another infrastructure token. Why does this need one at all?
Relayers stake $RLAY to accept declarations, and lose stake for settling a transaction that breaches its bounds. The token exists to make dishonest relaying expensive. Remove it and the network needs a permissioned relayer set instead, which defeats the purpose.
What stops an agent declaring one thing and meaning another?
Nothing does. Relayline verifies consistency, not honesty. The value is that the declaration is recorded before the outcome is known, so a dishonest declaration becomes visible afterwards rather than deniable.
If the relayer goes down, does my agent stop?
Yes, and that is deliberate. The alternative is an unconstrained fallback path, which is the exact failure this exists to prevent. Agents can be configured with several relayer endpoints.
Is this just a wrapper around transaction simulation?
Simulation tells you what would happen. It cannot tell you what the agent believed it was doing. The declaration is the part that is new.
Verified
- Contract verified on Solscan
- Constraint program audited
- Full spec published
- No team unlock before 2027
Point one agent at it and see what the record looks like
The fastest way to judge this is to run a single low-value action through a relayer and read the attestation it produces. If the record does not tell you something a transaction hash could not, none of the rest matters.
Agents are already transacting. The audit trail starts with the first declaration.