Onchain automation is here, and it's free to use.
5,000 onchain executions/month. $1 gas credits. No credit card needed.
20+ protocols, every major EVM chain, retry logic, MEV protection, and a full audit trail built in.
This is what reliable onchain automation looks like.
Base MCP lets AI agents execute DeFi transactions onchain.
With the MICA and Clarity act, who logs what the agent did, when it submitted, whether the parameters matched intent, and what happened when it failed?
Make sure you have an audit trail available!
Base MCP requires user sign-off before every transaction, which is smart for a consumer interface.
But autonomous agents can't pause for approval, I guess that's the whole point of autonomous.
There is this gap that exists between "agent-assisted" and "agent-executed".
@base shipped an MCP that lets ChatGPT and Claude manage wallets, swap tokens, and execute DeFi transactions onchain.
You can use KeeperHub to make sure that your agent transacting actually does so in a secure and reliable fashion.
KeeperHub: 5,000 free executions per month.
Connect your agent via MCP, CLI, x402, or MPP.
Run keepers for protocol monitoring, DeFi automation, liquidation protection, treasury management.
Retry logic, gas optimization, full audit trail included.
Start at https://t.co/rQAPltYWBH
The Web3 agentic stack is forming.
LLM for reasoning.
ERC-8004 for identity.
x402 and MPP for payments.
An execution layer for reliable onchain settlement.
The first three have serious institutional backing. The execution layer is still mostly fragile custom scripts and hope.
ERC-8004 gives AI agents verifiable onchain identity.
Three registries: who the agent is, what it has done, who vouches for it. Without it, your agent is an anonymous address. With it, it has a reputation.
KeeperHub publishes ERC-8004 agent cards for every workflow in the marketplace.
x402 and MPP both let AI agents pay for services without human approval.
1. x402 is per-request, works across any chain, any service.
2. MPP batches thousands of micropayments into single settlements on Tempo.
KeeperHub supports both. Your agent picks based on its wallet.
THORChain's post-mortem will say GG20 TSS.
however, would you have caught the node churn before the vault drained?
validator participation rates and threshold completion are actually very easily monitorable.
most teams just aren't watching them.
Agents shouldn't be hardcoded with execution logic.
KeeperHub workflows are now published as ERC-8004 agent cards on IPFS: discoverable, verified, x402-payable.
An agent finds the right execution workflow, pays for it, runs it. No human in the loop.
The CLARITY Act cleared Senate committee 15-9 Wednesday.
agents operating onchain will need provenance records, which means, every execution and every decision.
Audit trails are no longer optional, they are now the law.
Ouch... Another exploit, this time $10.8M gone from @THORChain.
Root cause: a churned validator node in the GG20 TSS scheme was leaking key material. The vault emptied before anyone detected it.
The monitoring gap strikes again.
DAO vote passes. parameter change approved. then what? someone has to execute it onchain.
Agents can handle this. but execution without simulation, retry logic, and an audit trail is how governance transactions fail without anyone noticing.
We intend on closing that gap.
Your agent approved a contributor payment via @Superfluid_HQ.
the stream opens.
nobody is watching the buffer.
no retry if it drains.
no record if it stops unexpectedly.
@KeeperHubApp runs 16 Superfluid actions with SLA-backed execution. nothing goes silent.
40+ DeFi protocols shut down in 2026.
$770M extracted.
if we were to generalise the common thread... nobody was watching the configurations that failed.
CLARITY Act Senate markup is expected this month.
When an AI agent executes a payment autonomously, who holds proof of what it was authorized to do?
KeeperHub logs trigger, simulation, submission, outcome, and timestamp. Every execution.
Autonomous agents are making real financial decisions. When a $ 2M rebalance goes wrong
"the agent did it"
is not an answer regulators or your ops team will accept.
Every KeeperHub execution logs: trigger, simulation, submissio$´2Moutcome, timestamp.
Huge congratulations to the winners of the
@KeeperHubApp <> @ETHGlobal Open Agents Hackathon track:
01 Tradewise Agentlab
02 Keeper-Gate
03 ZW.ARM
180 submissions across the board, and we were blown away by the quality.
MASSIVE kudos to all the builders! 💚
Shipped some good updates, during the busy week of a hackathon:
- Superfluid: 16 native actions live
- Per-workflow MCP at /mcp/w/[slug]
- ERC-8004 A2A + MCP cards out
- Native ETH on Uniswap V3
- Billing is now live on KeeperHub