· bitget integration

AlgoVault has 89.4%+ PFE Win Rate across 56,375+ signal calls, each Merkle-anchored on Base L2 (verifiable at ).

AlgoVault × Bitget — Build Verifiable AI Trading Agents

AlgoVault MCP gives your agent a composite verdict in one call — direction, confidence, regime, and cross-venue funding/sentiment context — backed by a publicly auditable record anchored to Base L2. Pair it with Bitget’s Agent Hub (and its agent-native AI account, GetClaw) and your agent has both the analytics brain and an execution venue purpose-built for autonomous AI trading.

Provenance: Official Bitget [email protected] (npm; description “Official Bitget MCP Server”, maintained by [email protected]) + github.com/BitgetLimited/agent_hub (default branch main). Demo execution requires a separate GetClaw demo account — Bitget’s per-account demo balance — because the Bitget MCP server has no env-var-level demo flag. This tutorial wraps the demo with BITGET_DEMO=true + MAINNET_BLOCKED=true constants + a hard 0.0001 BTC order-size cap. Verified 2026-04-25.

What you’ll build (90s read)

A runnable Node.js agent that:

  1. Calls AlgoVault MCP for get_trade_signal on BTC at the 1h timeframe.
  2. Reads the verdict’s signal and confidence and renders them as a natural-language instruction (e.g. “BUY 0.0001 BTC because confidence is 78% and the regime is TRENDING_UP”).
  3. Hands the instruction to Bitget’s GetClaw agent — which, when run against a GetClaw demo account, executes the order autonomously inside its dedicated AI account.

The whole loop runs in GetClaw demo mode with our wrapper guards — zero real-money risk in any code path.

Prerequisites (4 items)

  1. Node.js ≥ 22 (node --version to check).
  2. AlgoVault skills plugin installed:
    claude plugin install AlgoVaultLabs/algovault-skills
  3. Bitget GetClaw demo account — per Bitget’s Agent Hub announcement, GetClaw runs inside a dedicated AI account with its own balance. Sign up for a Bitget account at https://www.bitget.com, then enable GetClaw + a demo balance via the Agent Hub dashboard. The demo’s hard guards refuse to run unless you’ve set BITGET_DEMO=true and confirmed the keys belong to a demo account, not your mainnet wallet.
  4. Bitget Agent Hub MCP installed (recommended for the recipes; the demo runs without it via the public REST surface):
    claude mcp add -s user \
  --env BITGET_API_KEY=<your_demo_api_key> \
  --env BITGET_SECRET_KEY=<your_demo_secret> \
  --env BITGET_PASSPHRASE=<your_demo_passphrase> \
  bitget -- npx -y [email protected]

Demo: Natural-language verdict + GetClaw execution (≤80 lines)

// examples/bitget/demo.mjs (excerpt — see file for full source)
import { getAlgoVaultVerdict } from '../_shared/algovault-helper.mjs';

const MAINNET_BLOCKED = true;
const MAX_QTY_BTC = 0.0001;
const BITGET_BASE = 'https://api.bitget.com/api/v2';

if (process.env.BITGET_DEMO !== 'true') {
  throw new Error('BITGET_DEMO=true required — demo refuses to run against mainnet.');
}

console.log('=== DEMO MODE ===');
console.log('[BITGET_DEMO=true | wrapper-enforced; MCP server has no built-in demo flag]');

// 1. AlgoVault verdict
const verdict = await getAlgoVaultVerdict({ coin: 'BTC', timeframe: '1h' });

// 2. Natural-language instruction render
const instruction = verdict.signal === 'BUY' && verdict.confidence > 70
  ? `BUY ${MAX_QTY_BTC} BTC because confidence is ${verdict.confidence}% and regime is ${verdict.regime}`
  : `HOLD — ${verdict.signal}@${verdict.confidence}% does not meet policy threshold`;

// 3. Probe Bitget contract specs to prove the BITGET_DEMO path

console.log('=== NO REAL ORDERS PLACED ===');

Run it:

BITGET_DEMO=true node examples/bitget/demo.mjs

Expected output (last 4 lines):

[verdict | signal=... confidence=... regime=...]
[instruction | "BUY 0.0001 BTC because ..."]
[bitget | contracts HTTP 200 | host=api.bitget.com BTCUSDT minTradeNum=0.0001]
[getclaw | SKIPPED — see README for authenticated POST /v2/mix/order/place-order with BITGET_API_KEY/SECRET/PASSPHRASE]
=== NO REAL ORDERS PLACED ===

Walkthrough (line-by-line — neutral narration)

The script does three things in order:

  1. Fetch the AlgoVault verdict. The shared helper opens an MCP session against https://api.algovault.com/mcp, calls get_trade_signal for BTC on 1h, and returns the parsed signal / confidence / regime / factors payload. Free tier covers the BTC + 1h combination.

  2. Render the natural-language instruction. When the verdict satisfies the agent’s pre-configured policy (signal === 'BUY' AND confidence > 70), the script generates a one-line natural-language instruction GetClaw can interpret directly. When the policy does not fire, the instruction is HOLD with the failing reason. This recipe is the canonical “AlgoVault as the analytics brain, GetClaw as the natural-language execution agent” pattern.

  3. Probe Bitget contract specs. The script issues GET /api/v2/mix/market/contracts?productType=USDT-FUTURES to confirm Bitget’s BTCUSDT perpetual is available + records the exchange’s minimum trade unit (minTradeNum: 0.0001). The agent’s pre-configured policy uses this minimum as the order size cap. The === NO REAL ORDERS PLACED === banner closes every run regardless of the policy outcome. The script aborts immediately if BITGET_DEMO is not set to true, if MAINNET_BLOCKED has been edited away from true, or if the requested order size exceeds the MAX_QTY_BTC = 0.0001 cap — three independent guards against accidental mainnet execution.

3 Recipes

Recipe 1 — Natural-language verdict + GetClaw execution

This is the recipe examples/bitget/demo.mjs implements. The agent calls AlgoVault get_trade_signal for BTC on 1h, renders the verdict as a natural-language instruction, and hands the instruction to GetClaw running inside a dedicated demo account. GetClaw interprets the instruction, applies its own pre-configured risk policy (size cap, slippage tolerance), and either executes via POST /v2/mix/order/place-order or skips with a logged reason. GetClaw’s autonomy is bounded by the demo account’s balance + your wrapper’s BITGET_DEMO=true env contract — never by code-level checks alone.

Recipe 2 — Five Bitget AI Skills + AlgoVault composite verdict (complementary)

Bitget Agent Hub ships five analytical AI Skills out of the box: macro, technical, sentiment, intel, and news. Each Skill looks at one dimension of the market in isolation. AlgoVault’s composite verdict is complementary — it fuses the underlying primitives (RSI, EMA, funding rate, OI momentum, volume, plus cross-venue regime) into a single direction-and-confidence call. The agent runs both: Bitget Skills give per-dimension narrative context the agent can show to the user; AlgoVault gives the cross-dimensional decision. Use them together — they answer different questions.

Recipe 3 — Agent-native portfolio rebalance via GetClaw

The agent runs a daily regime scan via AlgoVault get_market_regime (or reads regime off get_trade_signal) across the assets in the GetClaw demo account’s portfolio. For any asset whose regime has shifted (e.g. moved from RANGING to TRENDING_DOWN), the agent’s pre-configured rebalance policy issues a natural-language rebalance instruction to GetClaw — inside the demo account’s walled garden. GetClaw’s account-isolation guarantees the rebalance can’t touch any other Bitget account funds. Multi-asset position state lives in the GetClaw demo account, not in agent memory — the agent reads, decides, and instructs; GetClaw owns and executes.

⚠️ Production setup (real-money)

The demo above runs in a Bitget GetClaw demo account only. Real-money setup requires:

  • KYC completion on Bitget.
  • Production API keys with only the permissions your agent needs (no withdrawals, IP-allowlisted) — distinct from your demo account’s keys.
  • Risk controls: per-order size cap, daily loss limit, kill switch.
  • Position monitoring: a separate agent or watchdog that tracks open positions independently. With GetClaw’s account-isolation model, the watchdog reads from the GetClaw account specifically — not the master Bitget account.

See examples/bitget/README.md for the full real-money checklist + the Bitget Agent Hub install via claude mcp add ... -- npx -y [email protected]. AlgoVault provides analytics; your agent (and GetClaw) decide what to execute inside the policy you set.

Why AlgoVault? (closing — MOAT recap)

  • Composite verdict, not raw indicators. One JSON response replaces 26-indicator vote-counting.
  • Cross-venue intelligence. Funding spreads, regime, and sentiment fused across 5 exchanges — not derivable from any single-venue API.
  • Publicly verified. Every signal anchored to Base L2 via Merkle proof. Verify before you subscribe.

89.4% PFE Win Rate · 56,375+ calls · 16+ on-chain batches → view live track record

Install

claude plugin install AlgoVaultLabs/algovault-skills

Once installed, every Skill in the pack is one-line invokable from Claude Code, Cowork, or any MCP-compatible client.

Tutorial © AlgoVault Labs · MIT licensed · Provenance verified 2026-04-25 · Bitget [email protected] (npm) + BitgetLimited/agent_hub (GitHub)