Today we are launching OnchainBench, the benchmark for agents in crypto.
It helps you decide how far you can trust an AI agent with your keys: how often it finishes the job, what tasks cost, and how easily it falls prey to an attack.
People are starting to let AI agents swap, lend and even trade for them. The bigger the task, the greater the risk you're taking by trusting the agent. So before you hand it your keys, you need to assess it.
There are many risks to assess at once. An agent could buy two tokens for a liquidity position but never add them to the pool, so the job isn't complete. It could spend more on AI model calls than the task is worth. And, onchain, there's no undo: agents read web pages, files and tool descriptions that anyone can write, and if one of them tricks the agent into sending funds to the wrong address, the money is gone. Until now, there was no common way to evaluate agents on all of this and more.
OnchainBench helps you compare agents by giving them the kind of work people already ask them to do: swaps on @Uniswap, @SushiSwap and @Honeyswap, lending on @aave, deposits into sDAI from @SkyEcosystem, prediction markets on @Omen_eth, research, and jobs that run over several days.
Each task runs through a @safe smart account on a copy of a public chain. The first version covers @ethereum, @gnosischain, @0xPolygon, @arbitrum, @Optimism, @base, @Celo, @modenetwork and @robinhoodapp. The benchmark emulates the kind of contracts, liquidity and prices the agent would meet with your money, without putting any of it at risk. A task only counts as done if the chain shows it was done. Some tasks also hide an attack in what the agent reads, such as a look-alike token or a fake "verification" request, so you can see how often agents get tricked into moving funds before you trust one with yours.
In the first run, Claude Code (@claudeai), Codex (@OpenAI), Hermes (@NousResearch), OpenClaw (@openclaw) and our own new OCB agent on @autonolas worked through 147 private tasks they had never seen. They completed between 88.4% and 96.2% of tasks, at about $0.03 to $0.35 per completed task. Agents lost funds to an attack 13 times in 1,793 attack tests.
We are excited to contribute this to the crypto ecosystem. If you build a protocol or a chain, agents are already among your users, and will likely be one of your biggest sources of growth. OnchainBench shows, protocol by protocol and chain by chain, how well they handle yours.
For the AI ecosystem: want an agent eval with real ground truth and real adversaries? Onchain, every result can be checked and every injected instruction is a live risk. OnchainBench is open to contributions, uses private tasks, and is ready for your agent.
Help us make onchain agents harder to trick and better at the job. If you run a protocol or are a heavy user, propose the task you think agents still get wrong on the site below, and every agent on the benchmark could end up tested on it. If you build agents, run OnchainBench yourself and submit your agent to be run.
The leaderboard is live now at https://t.co/xs5WV4paUy
Share it with a friend!
Want to trade Stock Tokens and memecoins that people are actually talking about on Robinhood? 👀
@valoryag is giving the first 10 eligible users 50 USDG each when they create their Pearl Connect agent on Robinhood via Pearl.
Promotion Terms, including eligibility restrictions, apply. Read more here: https://t.co/1J3vZNEg6u
A model for forecasting real-world events
Olas prediction trader agents generated the data. @valoryag used it to train a better AI forecaster.
Introducing Olas-Predict-R1-14B: an open-weight reasoning model with forecasting performance on par with the much larger closed-source model, GPT4.1.
7. How it's going.
The scale is indisputable.
As of Sept 16, 2026, @autonolas reports 20.59M total on-chain agent transactions, including 14.66M agent-to-agent transactions (A2A).
The opportunity for researchers is to study these systems in operation: which incentives produce useful work, how agents choose counterparties, and where guardrails improve outcomes. @autonolas offers running code, transaction records, and concrete cases to test.
Links:
https://t.co/frAbVIXCHc
https://t.co/wdFt5b7p5l
https://t.co/EPLMlkWrqe
There's a lot of talk about what systems agents "will need", as if key systems aren't already live.
For example, “Crypto x AI, AI x Crypto” survey by @initc3org’s, edited by @giuliacfanti and @AriJuels, outlines key questions on agents, including their payments and coordination.
It omits @autonolas, despite a track record of 20M+ total on-chain agent transactions since 2021.
Here are live systems these discussions can learn from.
Link:
https://t.co/WgFTL2i7p9
6. Further early experimentation.
Fun fact: @autonolas has been exploring how agents can participate in—and help decentralize—DAOs since before the ChatGPT era. As early as 2022, its discussions included agents launching their own DAOs, using what we then termed 'autonomous services'.
In 2023, Governatooorr put AI-assisted governance into practice on Olas: an autonomous agent delegate that could interpret proposals and vote on behalf of DAO members. Its experimental “good” and “evil” modes made the guardrail question concrete: whose objectives does an agent pursue, and what should constrain its decisions?
These experiments give researchers working examples to examine when studying decentralized governance, delegation and accountability in agent-run organizations.
Links:
https://t.co/N3EPx45a95
https://t.co/IX45DBI6x7
https://t.co/vCpDIsQOXj
https://t.co/FzuOelt3Fw
https://t.co/kxXK7rU11q
https://t.co/N3EPx45a95
How do you decentralize a DAO?
Our proposed constitution prioritizes autonomy & decentralization
https://t.co/UBaIFRZ3Mr
Related decentralized innovations
☴for twitter accounts
https://t.co/bwfaIS9k1w
☴for protocols
https://t.co/SYrRo5eyg6
☴read more https://t.co/AYSP8ap6FW
Valory just registered a .agent domain.
We support the effort to have the .agent top-level domain managed by a community, instead of being owned by one company.
Get involved & pre-register your domain here: https://t.co/MsbaEYBxTC
We’re growing at Valory and looking for top talent to join us:
→ Senior Product Manager
→ Senior Python Engineer
If you or someone in your network is interested, apply via Wellfound:https://t.co/qkbRF0zHm2
Are your weekend plans going to change your life? 🤔
They might if they include checking out our new roles...
Valory was built by world-class experts at the cutting-edge of crypto and AI.
Soon, a select few will join us and help take @autonolas to new heights:
Wanna help people go from paying for AI they rent to earning from AI they own?
I heard the folks behind the original agent project in crypto (@autonolas) are hiring again...
Send to your growth, product & legal/ops frens who wanna build something new💃
https://t.co/38MRZ79fT5
Can an AI be "truth-seeking"? What does that even mean?
On Episode 14 of the Agents Unleashed podcast, host @ThomasMaybrier digs into:
- Political bias & power dynamics: What's inside Grok 4?
- Microsoft’s medical AI that outperforms doctors
- Why protocols are crucial for the growth of agentic AI
🎧 Watch here on X or listen anywhere → https://t.co/PqzMexS23m
01:13 – Controversy around Grok 4
03:49 – The illusion of truth-seeking AI
06:35 – Microsoft’s MAI‑DxO in healthcare
09:27 – @TheoriqAI's @RonBodkin on why AI agents need protocols
11:59 – Mistral's new Voxtral model, AWS goes in on agents, DOOM as a diffusion model
Agentic AI is getting smarter, riskier and more integrated into our lives. This episode looks at where we’re at, where we’re going and what tradeoffs we might already be making.
This episode of Agents Unleashed is sponsored by @Sigil_Fund
We’re proud to welcome Gemma Welsh as COO.
As a founding member, she’s spent 3.5+ years dedicated to achieving co-owned AI and is more committed to Olas than ever.
With Valory & Olas maturing, Oaksprout the Tan, CPO, joins the ecosystem builders.
🔗 https://t.co/9rnq3hTCcm
Attending @token2049 in Dubai this April 30 – May 1?
A member of the Valory team, @victornotaro, will be on-site during the event—and he'd love to connect with you!
Drop him a line or leave a reply below—see you there!