Five hosted adapters covering the price-data stack agents actually need: CoinGecko spot prices, DefiLlama TVL + yields, DexScreener DEX pair liquidity, Pyth pull-oracle (free Hermes + paid Pyth Pro at 50ms / 200ms / 1s), and a verified Chainlink mainnet feed registry. Drop the URL, get the MCP tools. Free for every public tier.
All endpoints verified live 2026-05-27 · Chainlink addresses confirmed via on-chain eth_call · Pyth Pro endpoints reverse-engineered from openapi.json
Each box is one wmcp.sh adapter. Drop the canonical URL into /api/v1/tools?url=… and you get every tool listed — typed input schemas, agent-ready, no SDK.
Live crypto prices, market caps, trending coins, global cap, free-text search. ~30 req/min on free tier; pass Pro key via _auth for higher.
get_coin_price · get_coin_market_data · get_trending_coins · get_global_market_cap · search_coins
https://wmcp.sh/api/v1/tools?url=https://api.coingecko.com/api/v3
DeFi TVL across 7,500+ protocols, current prices for ~10k tokens, yield-pool APYs, historical chain TVL series, stablecoin breakdown by peg type.
list_protocols · get_protocol · get_current_prices · list_yield_pools · get_chain_tvl · list_stablecoins
https://wmcp.sh/api/v1/tools?url=https://api.llama.fi/
Real-time DEX pair data across 50+ chains: price, liquidity, 24h volume, transaction counts. Token-address lookup, pair detail, free-text search, trending boosts.
get_token_pairs · get_pair · search_pairs · get_token_profiles_latest · get_token_boosts_latest
https://wmcp.sh/api/v1/tools?url=https://api.dexscreener.com/
DefiLlama's official 23-tool MCP server (TVL, fees, yields, stablecoins, bridges, ETFs, hacks, raises, institutional holdings, token unlocks, oracle coverage, income statements). wmcp.sh handles the OAuth + token refresh so agents like Claude.ai / Cursor / Codex (which can't drive arbitrary OAuth flows) can connect with just a bearer token. Bring your own DefiLlama subscription.
get_market_totals · get_protocol_metrics · get_chain_metrics · get_token_prices · get_yield_pools · get_stablecoin_supply · get_etf_flows · get_bridge_flows · get_events · get_dat_holdings · get_token_unlocks · get_oracle_metrics · resolve_entity · ...+10 more
Connect at /dashboard, then point your agent at:
https://wmcp.sh/mcp/defillama
13 tools across the full Pyth surface. Hermes (free, ~400ms): list + latest + historical + publisher stakes. Benchmarks (free, historical OHLCV with intervals). Pyth Pro / Lazer (paid, Bearer token via _auth): 50ms / 200ms / 1s channels, payload-reduce.
list_price_feeds · get_latest_price · get_price_at_time · get_publisher_stake_caps · list_benchmarks_feeds · get_benchmarks_feed · get_historical_price · get_historical_price_interval · get_price_differences · list_lazer_symbols · lazer_latest_price · lazer_price_at_timestamp · lazer_reduce_price
https://wmcp.sh/api/v1/tools?url=https://hermes.pyth.network/
17 verified Ethereum mainnet price feed aggregator addresses (BTC, ETH, stablecoins, top L1/L2 tokens, FX, gold). All addresses eth_call'd 2026-05-27 — confirmed live, decimals match, recent updatedAt. Plus the canonical eth_call template so your agent reads live prices through any RPC.
list_feeds_ethereum · get_feed_address · get_read_call_template
https://wmcp.sh/api/v1/tools?url=https://data.chain.link/
Most "AI tool platforms" ship connectors that wrap docs without testing. We ran live curl against every documented endpoint and eth_call against every Chainlink address before publishing.
asset_type filter (Hermes rejects it)use Benchmarks instead ✓Python — Claude agent gets live BTC + ETH prices
from wmcp import WmcpClient
from wmcp.anthropic import to_anthropic_tools, execute_tool_use
from anthropic import Anthropic
client = WmcpClient(api_key="webmcp_live_…")
tools = client.tools("https://api.coingecko.com/api/v3")
anthropic = Anthropic()
msg = anthropic.messages.create(
model="claude-opus-4-7",
max_tokens=1024,
tools=to_anthropic_tools(tools),
messages=[{"role":"user","content":"What are BTC and ETH prices in USD right now?"}],
)
for block in msg.content:
if block.type == "tool_use":
print(execute_tool_use(client, "https://api.coingecko.com/api/v3", block.model_dump()))
cURL — direct execute against Pyth Hermes
curl -X POST 'https://wmcp.sh/api/v1/tools/execute' \
-H 'content-type: application/json' \
-d '{
"url": "https://hermes.pyth.network/",
"tool": "get_latest_price",
"args": {
"ids": ["e62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43"],
"parsed": true
}
}'
cURL — Chainlink BTC/USD lookup + read template
# 1. Get the feed address
curl -X POST 'https://wmcp.sh/api/v1/tools/execute' \
-H 'content-type: application/json' \
-d '{ "url":"https://data.chain.link/", "tool":"get_feed_address", "args":{"pair":"BTC/USD"} }'
# 2. Read the live price via your favorite RPC (cloudflare-eth.com etc.)
curl -X POST 'https://ethereum-rpc.publicnode.com' \
-H 'content-type: application/json' \
-d '{ "jsonrpc":"2.0", "method":"eth_call", "params":[{"to":"0xF4030086…","data":"0xfeaf968c"},"latest"], "id":1 }'
| Approach | Pros | Cons |
|---|---|---|
| Roll your own per-API SDK | Full control | 5 SDKs to maintain · auth-token plumbing · MCP-shape wrappers · ~2 days of work |
| Composio / Pipedream | Curated commerce + SaaS connectors | No oracle / price-feed / DeFi connectors exist — these vendors are owner-side |
| Smithery / dev-mcp | Some MCP servers cataloged | Each is a separately maintained third-party server, varying quality + uptime |
| wmcp.sh /price-data | Single endpoint, 28 tools, all verified, all free for public tiers | Free CoinGecko rate-limited; agent UX for choosing across 5 sources is on you |
args._auth. CoinGecko Pro key also supported via _auth for higher limits.channel arg on lazer_latest_price). CoinGecko + DefiLlama + DexScreener: standard REST round-trip, plus wmcp.sh adds a 60s edge cache that hides repeat queries. Chainlink: bounded by whatever RPC your agent uses to eth_call — Cloudflare typically ~80-150ms._auth on the Hermes tools will continue to work — same pattern as CoinGecko Pro. Anonymous Hermes calls may rate-limit or 401. Our adapter is forward-compatible: _auth is already wired on every Pyth tool, even Hermes./api/v1/tools?url=… is shaped as both MCP tool_use and OpenAI function_call JSON. Drop it into whichever agent framework you're using. Examples for Anthropic SDK + OpenAI SDK at /integration/openapi.