liquidpad-agent-kit

Build agents on Liquid Protocol in 50 lines.

Composable wrapper around liquid-sdk + the public LiquidPad API + ERC-8004 reads. Read-only without a private key. Add a key to enable deploys.

npm install liquidpad-agent-kit

What it gives you

• draftConcept({ vibe }) — turn a free-form vibe into { name, symbol, theme, tagline } via LiquidPad's curator endpoint.
• deployToken({ name, symbol, ... }) — deploy a token via liquid-sdk. Auto-stamps your ERC-8004 agent id into deploy context for provenance.
• shipFromVibe('matcha latte ghosts') — end-to-end pipeline: vibe → concept → onchain. One call, full lifecycle.
• getProofOfBurn(tokenAddress) — pending fees + burn transfers + LIQ buybacks for any token. Sourced directly from Base, no backend needed.
• verifyToken(tokenAddress) — provenance check via LiquidPad's public verify endpoint (kind, deployer, ERC-8004 agent id).
• listAgents() — query the cross-ecosystem agent registry (ERC-8004 verified agents shipping on Liquid Protocol).
• inference({ model, messages }) — OpenAI-compatible chat completions via the LiquidPad LLM Gateway, gated by your $LPAD tier. One credential covers deploy + inference.
• listModels() — show which gateway models your tier can call.
• getUsage() — current quota + rate-limit state.

The shape is intentionally tiny so you can drop it into any TypeScript / Node.js project — agents, bots, CLIs, MCP servers, dashboards.

Quick start

import { createLiquidAgent } from "liquidpad-agent-kit";

const agent = createLiquidAgent({
  privateKey: process.env.AGENT_PK as `0x${string}`,
  liquidpadApiKey: process.env.LIQUIDPAD_API_KEY,
  erc8004AgentId: 50962,
});

const result = await agent.shipFromVibe("matcha latte coded by ghosts at 3am");

console.log(result.concept);       // { name: "Ghost Matcha", symbol: "GMTC", ... }
console.log(result.tokenAddress);  // 0x...
console.log(result.txHash);        // 0x...

Runnable examples

Six self-contained scripts in examples/ — copy .env.example, install, run.

| File | What it shows | Needs | |---|---|---| | 01-verify.mjs | Provenance check on any Base address | nothing | | 02-proof-of-burn.mjs | Pending fees + burn transfers + LIQ buybacks | nothing | | 03-list-agents.mjs | Cross-ecosystem agent registry | nothing | | 04-draft-concept.mjs | Vibe → concept | API key | | 05-inference.mjs | Chat completion through LLM Gateway | API key | | 06-ship-from-vibe.mjs | Full pipeline: vibe → concept → onchain | API key + private key + ETH for gas |

Read-only mode

Everything except deployToken and shipFromVibe works without a wallet:

const agent = createLiquidAgent();

const proof = await agent.getProofOfBurn("0xBF0775cBCA2744549cD016DAb8D3b3403De58bBF");
console.log(proof.burns.count, "burn cycles");
console.log("pending fees:", proof.pendingFees);

const verify = await agent.verifyToken("0xBF0775cBCA2744549cD016DAb8D3b3403De58bBF");
console.log(verify.kind, verify.deployer);

API reference

createLiquidAgent(config?)

Returns a LiquidAgent instance. All config fields are optional.

| Field | Type | Default | Notes | |---|---|---|---| | privateKey | 0x{hex} | — | Required for deployToken / shipFromVibe. Read-only without it. | | rpcUrl | string | viem default | Base mainnet RPC. | | liquidpadApiKey | string | — | From @liquidpadbot on Telegram (/apikey). Required for draftConcept / shipFromVibe. | | liquidpadApi | string | https://api.liquidpad.site | API base URL. | | liquidpadSite | string | https://www.liquidpad.site | Site URL (used for verifyToken, listAgents). | | erc8004AgentId | number | — | Stamped into deploy context for provenance. |

Methods

draftConcept({ vibe, previous?, feedback? }) → Concept

Turn a vibe into a shippable concept. Returns { name, symbol, theme, tagline }.

deployToken({ name, symbol, image?, metadata?, context? }) → DeployResult

Deploy a token via liquid-sdk. ERC-8004 agent id is auto-merged into context if configured.

Returns { tokenAddress, txHash, poolId, event }.

shipFromVibe(vibe, options?) → DeployResult & { concept }

draftConcept(vibe) → deployToken(concept) in one call.

getProofOfBurn(tokenAddress, { fromBlock?, treasuryAddress? }) → ProofOfBurn

Pending fees + burn transfers + LIQ buybacks. Reads directly from Base. No API key needed.

verifyToken(tokenAddress) → VerifyResult

{ verified, kind, deployer, deployedAt, txHash, erc8004AgentId }. Calls LiquidPad's public /api/verify/<addr>.

listAgents({ verified? }) → AgentEntry[]

Cross-ecosystem agent registry. Currently returns ERC-8004 verified agents that have deployed via Liquid Protocol.

inference(req) → ChatResponse

OpenAI-compatible chat completion via the LiquidPad LLM Gateway. Quotas, rate limits, and model access scale with your $LPAD tier. The same liquidpadApiKey you used for draftConcept works here too.

const reply = await agent.inference({
  model: "liquidpad-trial",
  messages: [
    { role: "system", content: "You're a token concept curator." },
    { role: "user", content: "Pitch me a vibe for a Base meme token." },
  ],
});

console.log(reply.choices[0].message.content);
console.log("tier:", reply.liquidpad?.tierName, "tokens:", reply.liquidpad?.tokensCharged);

Errors follow the OpenAI shape ({ error: { type, code, message } }). Tier-locked models throw model_not_available_for_tier. Quota exhaustion throws monthly_quota_exceeded.

listModels() → ModelEntry[]

Returns the models the caller's tier can use, OpenAI-shaped ({ id, object, owned_by, tier_required }).

getUsage() → UsageReport

Current quota counter, rate-limit state, and key status. Useful for budget meters in your UI.

Design

• Tiny surface: 9 methods. Composable. Each one does one thing.
• Zero lock-in: drop it in, take it out, no opinions about your runtime.
• Read-only friendly: most surface works without a private key. Pull data into dashboards, audits, indexers without spinning up a wallet.
• Attribution-aware: ERC-8004 agent id stamps into deploy context automatically — your token's provenance is queryable onchain forever.

Built on

• liquid-sdk by @m00npapi — the actual onchain primitive.
• viem — Base mainnet client.
• LiquidPad public API — concept curator, verify, agent registry.

See also

• LiquidPad ecosystem dashboard — live attribution metrics.
• liquidpad-mcp — MCP server for AI assistants (Claude, Cursor, Kiro).
• aeon framework — autonomous agent runtime that LiquidPad runs on.

License

MIT. See LICENSE.