Zolva Agent

Open trading-agent infrastructure on Robinhood Chain. Build, backtest, and inspect agents. API · MCP · x402.

0x8e04549484d0612be8373a0263b11928a5f7681f

[Zolva Agent @zolva_agent]

$ZOLVA is live | 0x8e04549484d0612be8373a0263b11928a5f7681f

A platform for programmable trading agents and open market data on Robinhood Chain.

Build an agent, test its strategy, and follow its activity or explore the public data through an API or an AI client. The pieces are connected, so an agent’s market, rules, decisions, and recorded results can be examined together.

[Zolva Agent @zolva_agent]

Follow both, every decision with its reason:

Rules only: https://t.co/ow3tUBXcBP
Rules + AI: https://t.co/3gl3UowD7l

[Zolva Agent @zolva_agent]

A public test: two Zolva agents, same rule, same stocks, started the same hour.

One trades on rules alone. The other has the AI co-pilot, which can pass on a trade but never resize one.

Tied so far: the AI approved every signal. The first pass is where they split 👁️

[Zolva Agent @zolva_agent]

See how Prompt Mode works: https://t.co/EPjNeZjsUH

[Zolva Agent @zolva_agent]

Turn a trading idea into agent rules.

In Zolva Studio, open Custom → Prompt and describe when you want the agent to buy and sell. Prompt Mode turns your description into a first draft you can review.

[Zolva Agent @zolva_agent]

Your agent’s AI co-pilot, your model:

Lite: Claude Haiku 5.5, DeepSeek V4.1 Flash, GPT-5.6 Luna
Standard: Gemini 3.8 Flash, GPT-5.4 mini
Pro: Claude Sonnet 5.5, GPT-5.6 Terra, Grok 4.7
Max: Claude Opus 5.5

Whichever you pick, it only approves or passes.

[Zolva Agent @zolva_agent]

The rules, co-pilot note, and backtest appear together in the Studio. Review and edit the rules, then see the backtest update as you make changes.

Use the backtest to examine historical behavior, not predict future results.

[Zolva Agent @zolva_agent]

@NousResearch Congrats on almost 1,000 entries!

This week, Zolva joined the community plugin catalog: read-only data on trading agents on Robinhood Chain, plus a cron that sends every new trade with its reason.

https://t.co/4GWTmt4OkK

[Zolva Agent @zolva_agent]

Zolva is now live in Hermes Agent by @NousResearch.

Ask your own agent which Zolva agents beat holding, and why each trade happened. Or set one cron job and get every new trade in Telegram, with its reason.

One URL. Read only. Nothing to sign.

https://t.co/VXhyWo8YV3

[Zolva Agent @zolva_agent]

@chiragnarang23 One data point from running this live: our AI co-pilot can only pass on a rule’s signal, nothing else. Over 48 reviews it passed 10. 6 of those fell further, 4 didn’t.

Narrow beats smart so far, but 10 is far too few to call it.

[Zolva Agent @zolva_agent]

Setup in one line:

hermes plugins install zolvaagent/hermes-zolva hermes plugins enable zolva

Then ask: /zolva-agents which agents are beating holding this week?

[Zolva Agent @zolva_agent]

Zolva is now live in Hermes Agent by @NousResearch.

Ask your own agent which Zolva agents beat holding, and why each trade happened. Or set one cron job and get every new trade in Telegram, with its reason.

One URL. Read only. Nothing to sign.

https://t.co/VXhyWo8YV3

[Zolva Agent @zolva_agent]

@vladtenev @Stanford someone has to stay in the trenches while you’re at Stanford. our agents took the night shift on Robinhood Chain 🫡

[Zolva Agent @zolva_agent]

Fixed both, live now:

• Once an agent is published, its record stays public. Owners can take it off the leaderboard, but its page, trades and losses stay up.
• Real money orders wait when a Chainlink feed is stale or unreadable. No pool-price fallback.

Plus CI on every push (242 tests) and a “Workers only” note in the README. Thanks for the review 🫡

https://t.co/4zUXtFdiOi

[slime @whoiskn0why]

I tested @zolva_agent
Cloned the repo, installed dependencies and ran everything locally.

What’s good:

• Real paper-trading engine with substantial strategy logic
• 239 tests passed across 24 files
• Local build + Cloudflare build both passed
• REST API and MCP server functional
• Real agents, decision histories, markets and trade records
• Hourly autonomous decisions with written explanations and optional AI trade veto

The engine actually works. This isn’t vaporware or a landing page pretending to be a trading product.

What’s bad:

• Owners can hide losing agents by unpublishing them, undermining performance transparency
• Stale or missing Chainlink feeds fall back to pool prices without sufficient validation, creating a serious safety concern for live trading
• Normal Node server deployment is broken
• Documentation and environment configuration are partly outdated
• No CI pipeline, live trading security remains untested

Verdict: 7/10

Two architectural problems need addressing: performance integrity and price-feed safety.

The foundation is legitimate. The issues are fixable, and paper trading works properly.

Backend looks good. $zolva

[Zolva Agent @zolva_agent]

This is the best kind of review: you ran it instead of reading the landing page. Thank you.

slime, You’re right on both architecture points:

• Unpublishing shouldn’t erase a record. A published agent’s history should stay public, losses included.
• A stale or missing Chainlink feed shouldn’t let a live order through on the pool price alone. Live trading should hold until the check passes.

CI and a clear “Cloudflare Workers only” note are on the list too.

Fixes will land on https://t.co/4zUXtFdiOi with the commits! Once again, Thanks.

[Zolva Agent @zolva_agent]

The transaction, if you want to check it:
https://t.co/m2ZMSSZzDC

Other pairs pay the fee to a wallet that buys back ZOLVA and burns it. Stock tokens pay no Zolva fee.

[Zolva Agent @zolva_agent]

Every $ZOLVA buy on Zolva Swap burns a little supply.

The 0.25% swap fee on a ZOLVA buy is taken in ZOLVA and sent to 0x…dEaD, in the same transaction as the swap.

The first one is on chain: 0.003 ETH in, 85,364 ZOLVA bought, 213.41 burnt.

Small start.

[Zolva Agent @zolva_agent]

Zolva now has a public dev log.

Everything that shipped since we went live on Oct 15, newest first. 48 entries, including the research that didn’t work.

Every entry links to its source: the commit, the page or the transaction. Check it yourself.

https://t.co/SdA0LN0CyA

[Zolva Agent @zolva_agent]

Your agent’s AI co-pilot, your model:

Lite: Claude Haiku 5.5, DeepSeek V4.1 Flash, GPT-5.6 Luna
Standard: Gemini 3.8 Flash, GPT-5.4 mini
Pro: Claude Sonnet 5.5, GPT-5.6 Terra, Grok 4.7
Max: Claude Opus 5.5

Whichever you pick, it only approves or passes.

[Zolva Agent @zolva_agent]

So far on paper agents: 43 trade reviews, 33 approved, 10 passed.

6 of the 10 passed buys fell further in the next 24 hours. One didn't: on Nov 9 it passed on $MU and $MU rose 6.4% the next day.

A second opinion, not an oracle. Small sample.

All 41 model calls cost $0.006 in total, about $0.00015 per review (DeepSeek V4.1 Flash).

You pick from 9 models, set a spending limit and an end date. New accounts get $0.50 free.

Turn it on in the Studio: https://t.co/OmrOXF7GTC

[Zolva Agent @zolva_agent]

How does the AI co-pilot actually take part in a Zolva agent's decision?

Every hour, the rules go first. The AI only speaks when the rules want to trade, and then it has one choice: approve or pass.

Here's one real decision, start to finish 🧵

[Zolva Agent @zolva_agent]

What it can't do:

· change the size or the price · skip or delay a stop loss (forced exits never go to it) · trade on its own

If it's off, out of credits or the model fails, the rules decide alone. The rules are always the default.