graph-uniswap

Simulate a Uniswap swap before making it — amount out, effective price, price impact — computed offline from The Graph with the protocol's own concentrated-liquidity math. Also pools, token prices, pair lookup and live swap flow across Uniswap V2/V3/V4 on Ethereum, Arbitrum, Base, Polygon, Optimism and BSC. Trigger keywords: uniswap, swap, quote, price impact, slippage, how much would I get, pool, liquidity, fee tier, LP, DEX, token price, V3, V4.

Answer Uniswap questions with live subgraph data — and, uniquely, answer "what would this trade actually cost me?" without an RPC, a node, or a private key.

Setup

Needs a free Graph API key (https://thegraph.com/studio, 100k queries/month free) in GRAPH_API_KEY. The server runs over stdio via npx -y graph-uniswap-mcp.

Picking a tool

The user is asking…	Use
"how much would I get for X", "what's the slippage/price impact", "is this pool deep enough"	quote_swap
"what's WETH worth", "price of X"	get_token_price
"best pools on ", "most active pools"	top_pools
"the USDC/WETH pool", "which fee tier"	find_pool
"stats for this pool"	pool_info
"recent trades", "who is trading"	recent_swaps
"what chains/versions are supported"	list_markets
anything the above cannot express	raw_query

Typical flow for a trade question: find_pool → pick the pool → quote_swap.

quote_swap — the one to reach for

Simulates an exact-input swap against the pool's real tick liquidity.

text

quote_swap  pool=0x8ad5…e6d8  tokenIn=WETH  amountIn=100  chain=ethereum  version=v3
→ amount_out 223661.14 USDC · effective_price 2236.61 · price_impact_pct 0.6291 · ticks_crossed 2

price_impact_pct is measured against the fee-adjusted spot, so it isolates depth rather than restating the fee. fee_pct is reported separately.

Always check quotable. When it is false, report the reason — do not substitute a guess or fall back to a TVL figure. It refuses on:

• V4 pools with a hook — hooks can override fee and curve, so vanilla math would be wrong. On Base the top V4 pools by volume are hook-driven.
• Trades bigger than the visible liquidity — you get lower_bound_amount_out, not a quote. Suggest splitting the trade, or quoting on-chain.
• Zero in-range liquidity, or a deployment exposing no ticks (V3 on Base).

The number simulates the most recently indexed block. Real execution depends on state at inclusion — say so when the amount is material. It is single-pool: not a router, no multi-hop, no splitting.

TVL is null on purpose — do not work around it

tvl_usd is null for V3/V4. Uniswap's native subgraphs accumulate totalValueLockedUSD from per-event deltas and it drifts: on the canonical USDC/WETH 0.3% pool it read 138.9M USDC against 6.1M actually held on chain (22.7x high, on a blue-chip pool). V2 reserve0/reserve1 are real balances and are still returned.

If a user asks how deep a pool is, run quote_swap at their size — that is the truthful answer. Never reconstruct a TVL estimate from raw_query and present it as fact.

Other things worth knowing

• Rank pools by volume, never TVL — one spam-token pool can claim trillions in fake liquidity, so a TVL sort returns junk at the top.
• Omitting version picks the highest-query-volume version on that chain, which is not always the most useful one. Pin version when the user means a specific one.
• Chain aliases (eth, arb, matic, bnb) are accepted.
• A stablecoin price should read ≈ $1 — a quick sanity check that pricing resolved correctly.