Read GMGN aggregate bundle, sniper, fresh-wallet, and concentration signals through Noesis without treating them as wallet-level proof.

Bot /bundleREST GET /api/v1/token/{addr}/bundlesMCP token_bundles

How to read Noesis bundle and bot signals

TL;DR. GET /api/v1/token/{addr}/bundles combines token metadata with GMGN's token-level and holder-category statistics. It reports aggregate signals such as bundler-classified supply, fresh-wallet rate, bot/degen share, holder concentration, and category counts. It does not return bundle groups, individual bundle wallets, same-slot transactions, SOL spent, or evidence that particular wallets sold.

The endpoint supports Solana, Ethereum, Base, BNB Chain, and Robinhood Chain when GMGN has data for the requested token.

Data sources and response contract

Noesis makes four independent reads:

Unified token metadata (token)
GMGN token statistics (token_stat)
GMGN holder-category statistics (holder_stat, nullable when unavailable)
Detailed bundle telemetry (bundler_stat, nullable when unavailable)

Representative response:

{
  "token": {
    "address": "EPuZ1X...Fpump",
    "symbol": "TOKEN",
    "source": "gmgn"
  },
  "token_stat": {
    "top_bundler_trader_percentage": 0.125,
    "fresh_wallet_rate": 0.081,
    "top_bot_degen_percentage": 0.032,
    "top_rat_trader_percentage": 0.018,
    "top_entrapment_trader_percentage": 0.009,
    "top_10_holder_rate": 0.278,
    "dev_team_hold_rate": 0.018
  },
  "holder_stat": {
    "bundler_count": 28,
    "sniper_count": 14,
    "fresh_wallet_count": 35,
    "insider_count": 12
  },
  "bundler_stat": {
    "bundler_count": 63,
    "bundler_swaps": 148,
    "bundler_hold_ratio_raw": "12.5",
    "bundler_max_hold_ratio_raw": "3.1",
    "bundler_swap_quote_amount": 48123.5
  }
}

Numeric availability depends on GMGN. Treat missing or zero-valued fields conservatively; they do not prove that the behavior is absent.

Run the analysis

Telegram bot

/bundle <token_address>

REST API

curl -H "X-API-Key: $NOESIS_API_KEY" \
  "https://noesisapi.dev/api/v1/token/<TOKEN_ADDRESS>/bundles?chain=sol"

For EVM tokens, pass chain=eth, base, bsc, or robinhood when automatic detection would be ambiguous.

MCP

token_bundles(address="<TOKEN_ADDRESS>", chain="sol")

Field interpretation

top_bundler_trader_percentage: GMGN's aggregate share associated with its bundler-trader classification.
fresh_wallet_rate: aggregate fresh-wallet signal from GMGN.
top_bot_degen_percentage: aggregate bot/degen signal.
top_rat_trader_percentage: aggregate rat-trader signal.
top_entrapment_trader_percentage: aggregate entrapment/insider-style signal.
top_10_holder_rate: top-ten holder concentration.
dev_team_hold_rate and creator_hold_rate: creator/team concentration signals.
bundler_count, sniper_count, fresh_wallet_count, and insider_count: upstream category counts when holder statistics are available.
bundler_swaps and bundler_swap_quote_amount: detailed bundle activity totals when indexed.
fields ending in _raw: deliberately unscaled provider strings. Do not render them as percentages until their response-specific unit is established.

These are provider-defined aggregates. GMGN does not expose its full classification methodology through these endpoints, so Noesis must not present the values as independently verified wallet relationships.

What this endpoint cannot answer

The response cannot establish:

which wallets formed a bundle;
whether wallets bought in the same slot;
how many SOL a group spent;
the group's current combined holding;
whether the alleged bundle wallets sold;
whether wallets share a common funder.

Use /fresh for the wallet-age-≤7-day trading cohort and /topholders for current large-holder positions. Noesis preserves the upstream Fresh tag but does not independently recompute wallet age. Even when several signals agree, present the result as risk evidence rather than proof of common ownership.

Caveats

The endpoint depends on GMGN availability and classification coverage.
Existing token_stat percentage fields use 0-1 ratios. Detailed bundler_stat ratio fields remain raw strings because live responses have used inconsistent scales.
holder_stat is nullable because Noesis treats that supplementary GMGN request as fail-soft.
The REST endpoint is currently one credit, although its request-throttle class is Heavy (one request per five seconds). Credit price and throttle class are separate concepts.