
An agent-first desk for @Enter_Elysium and @HyperliquidX. Ask in your own language, get quotes and best-price swaps, and trade perps from chat.
Joined November 2026
- Tweets81
- Following23
- Followers153
- Likes124
Pinned Tweet
Zypher Agent is live on @Enter_Elysium mainnet.
One app for HyperEVM, Elysium and Hyperliquid:
- Launchpad: browse new Elysium tokens live (token creation coming soon)
- Market: Activity, Holders and trading right on the token page
- Swap and Bridge across HyperEVM, Elysium and other chains
- Perps: all 299 Hyperliquid markets
- Portfolio across 13 networks
Or just ask the agent: "swap 0.05 HYPE to USDC", "long $10 BTC 5x", "bridge 20 USDC to Base".
The agent only acts on what you typed, and you can pause it anytime. The Agent Wallet is optional and custodial, so keep the balance small.
0xe32ceedff1e5834ede703f009f81454e9cc154e4
zypheragent.com
Zypher MCP just crossed 1,600+ users within its first 24 hours. And we're just getting started.
V0.7 is now live, while we're already working on V1.0.
Here's what's coming next:
- More MCP tools to expand what agents can do.
- New skills to make agents more capable and useful.
- x402 payments on @Enter_Elysium, enabling AI agents to pay for services and interact with payment-enabled infrastructure.
Zypher building an agent infrastructure where AI agents can access tools, execute actions, and make payments autonomously.
1,600+ users in the first 24 hours is a great start.
Now it's time to build the next version.
V1.0 is coming.
⚡ Boosted 10x ZYPHER on DEX Screener!
dexscreener.com/elysium/0xe3…
How Zypher Agent works, what it will and will not do with your wallet, and plain-language guides to the markets it reads. No hype, nothing you have to take on trust.
Zypher Agent is a web app and MCP server for Hyperliquid and HyperEVM mainnet. It brings market data, swaps, perpetual orders, token launches and an assistant into one place. A single rule shapes every part of it: the agent works inside hard limits, and every action is checked before it can run.
We started from a plain question. If an AI model can read markets and call tools, what is the least it needs to be useful, and the most it should ever be trusted with? Our answer is that it should never hold your main wallet key, and that anything it can execute on its own must stay inside limits you set.
zypheragent.com
Which networks:
Perpetuals trade on Hyperliquid mainnet and swaps run on HyperSwap V3 on HyperEVM, chain ID 999, with HYPE as the gas token. Token launches go through the Boxa launchpad on Elysium (chain ID 99901), a community-run chain and not the official mainnet. These are real funds: start small and read every proposal.
44
Where approval happens:
Every proposal shows its parameters and passes Zypher Check, which looks at route, price impact, pool depth and token safety, and is checked against your policy limits on the server. It expires after 10 minutes. With your own wallet, swaps are signed in the browser and the app never holds your key. The optional Agent Wallet is a separate small custodial wallet, held encrypted by the operator, that executes on its own within daily USD caps and a leverage cap. Pause is instant, a kill switch stops everything, and every execution is journaled with its rationale.
32
What the assistant can do:
Read tools cover Hyperliquid markets, tickers, candles and account data, HyperEVM tokens, quotes and balances, and a token health check. When you ask for a trade, the assistant returns a proposal: a swap, a perpetual order, a position close or a token launch. With your own wallet it executes nothing itself; with the optional Agent Wallet it can execute from chat.
It’s one shared $500/day API-execution budget per Agent Wallet, across all API tokens and clients not $500 per token. Risk-increasing actions share that budget; closes, cancels, and transfers back to the owner don’t. Separate transfer/launch budgets aren’t configurable yet, but wallet-level caps, allowlists, pause, and the kill switch still apply.
Replying to @Zypheragent
Does the $500/day API-token cap apply across all MCP wallet writes, or can clients set separate limits for transfers and token launches?
Watch this demo video to see how $ZYPHER MCP Works.... Your agent, inside hard limits!
Zypher Agent is now an MCP server (v0.7.0).
An AI agent can read markets, check tokens and trade perps with a real wallet, straight from your terminal. Works with any stdio MCP client (tested in omp).
Mainnet test, real funds, tiny size:
- One call shows the wallet across Arbitrum, HyperEVM, Elysium and Hyperliquid
- The agent flagged the #1 Elysium token as unsafe: the index said 138k HYPE volume, on-chain showed zero swaps
- BTC long, 2x isolated, filled at 73,528 with stop-loss and take-profit resting on Hyperliquid
- Bad orders (no limit price, empty TP/SL, invalid cancel) were refused before signing
- My first $10 order was rejected by Hyperliquid's minimum. The agent didn't retry blindly: it checked status, confirmed nothing changed and asked again. Root cause fixed the same day.
Install in 2 minutes:
1. zypheragent.com/agent → Agent → Wallet → API tokens → Create token
2. Run in your terminal, never in chat:
claude mcp add -s user -e ZYPHER_API_TOKEN=<your_token> zypher-agent -- npx -y @zypheragent/[email protected]
No token: 17 read-only tools (markets, tokens, wallet).
With a token: 31 tools, adding swap, perps, TP/SL and launch.
Caps, kill switch and your key stay server-side. Tools move real funds with no confirmation step, so keep caps small.
Docs + demo video below.
NPM: @zypheragent/mcp
npmjs.com/package/@zypherage…
Great milestones for us. $ZYPHER MCP Just surprassed 1600+ user today!
Give your AI agent real market data and a wallet it can trade from. This is the Model Context Protocol server of Zypher Agent. It works with Claude Code, Codex, Cursor, Claude Desktop, omp and any client that speaks stdio MCP.
npmjs.com/package/@zypherage…
When an agent interacts with real financial markets, knowing what happened is just as important as knowing what it intended to do.
Zypher journals executions together with their rationale and makes the activity visible through the application.
This provides users with a record of agent activity and helps them understand how actions were initiated and executed.
Combined with policy checks and execution controls, the audit trail becomes an important part of the overall system.
The objective is to make agent activity easier to inspect rather than leaving users guessing about what happened behind the scenes.
Give your AI agent real market data and a wallet it can trade from. This is the Model Context Protocol server of Zypher Agent. It works with Claude Code, Codex, Cursor, Claude Desktop, omp and any client that speaks stdio MCP.
Read, no setup. Hyperliquid perps, HyperEVM quotes, the Elysium market, holders, activity and token checks. No account, no key, no token.
Act, when you allow it. Add one API token from your Agent Wallet and the agent can swap, trade perps with TP/SL, deposit, launch a token and bridge, inside caps you set.
Your key stays on the server. The package holds no keys and cannot sign anything.
e/acc
npmjs.com/package/@zypherage…
Zypher Agent is now an MCP server (v0.7.0).
An AI agent can read markets, check tokens and trade perps with a real wallet, straight from your terminal. Works with any stdio MCP client (tested in omp).
Mainnet test, real funds, tiny size:
- One call shows the wallet across Arbitrum, HyperEVM, Elysium and Hyperliquid
- The agent flagged the #1 Elysium token as unsafe: the index said 138k HYPE volume, on-chain showed zero swaps
- BTC long, 2x isolated, filled at 73,528 with stop-loss and take-profit resting on Hyperliquid
- Bad orders (no limit price, empty TP/SL, invalid cancel) were refused before signing
- My first $10 order was rejected by Hyperliquid's minimum. The agent didn't retry blindly: it checked status, confirmed nothing changed and asked again. Root cause fixed the same day.
Install in 2 minutes:
1. zypheragent.com/agent → Agent → Wallet → API tokens → Create token
2. Run in your terminal, never in chat:
claude mcp add -s user -e ZYPHER_API_TOKEN=<your_token> zypher-agent -- npx -y @zypheragent/[email protected]
No token: 17 read-only tools (markets, tokens, wallet).
With a token: 31 tools, adding swap, perps, TP/SL and launch.
Caps, kill switch and your key stay server-side. Tools move real funds with no confirmation step, so keep caps small.
Docs + demo video below.
NPM: @zypheragent/mcp
npmjs.com/package/@zypherage…
Zypher Agent is now an MCP server (v0.7.0).
An AI agent can read markets, check tokens and trade perps with a real wallet, straight from your terminal. Works with any stdio MCP client (tested in omp).
Mainnet test, real funds, tiny size:
- One call shows the wallet across Arbitrum, HyperEVM, Elysium and Hyperliquid
- The agent flagged the #1 Elysium token as unsafe: the index said 138k HYPE volume, on-chain showed zero swaps
- BTC long, 2x isolated, filled at 73,528 with stop-loss and take-profit resting on Hyperliquid
- Bad orders (no limit price, empty TP/SL, invalid cancel) were refused before signing
- My first $10 order was rejected by Hyperliquid's minimum. The agent didn't retry blindly: it checked status, confirmed nothing changed and asked again. Root cause fixed the same day.
Install in 2 minutes:
1. zypheragent.com/agent → Agent → Wallet → API tokens → Create token
2. Run in your terminal, never in chat:
claude mcp add -s user -e ZYPHER_API_TOKEN=<your_token> zypher-agent -- npx -y @zypheragent/[email protected]
No token: 17 read-only tools (markets, tokens, wallet).
With a token: 31 tools, adding swap, perps, TP/SL and launch.
Caps, kill switch and your key stay server-side. Tools move real funds with no confirmation step, so keep caps small.
Docs + demo video below.
NPM: @zypheragent/mcp
npmjs.com/package/@zypherage…
Another step forward for Zypher Agents.
We've submitted our application for the Elysium $1M Builder Credits Program with @Alchemy.
Access to reliable RPC, WebSockets, Data APIs, and webhooks will be valuable as we continue building and scaling Zypher on Elysium.
Our focus is still the same: making AI-powered trading more accessible through natural language, real-time market data, and policy-controlled execution.
Appreciate the @Alchemy & @Enter_Elysium team for supporting builders and creating more opportunities to build within the ecosystem.
Let's keep building. e/acc
Every analysis tool is read only: it reads markets and token contracts, and it never places an order. Execution lives in a separate, later tier, behind your policy limits and a one-time confirmation that expires.
Checks before exposure: Three further tools look beyond price at contract risk, social signal and new pairs. Their output is data, never an instruction.
> Token Health
analyze_token_health scores bytecode for mint, pause, blacklist and owner controls. A heuristic, not an audit.
> Social
get_social_info searches X and needs a bearer token. Every tweet is wrapped as untrusted external text.
> Discovery
get_new_pairs lists newly created pairs. In development: it needs an indexer that is not yet built.
22




