DFlow: Inside Solana's $88 Billion Aggregator, Measured On-Chain (2026)

DFlow is the Solana trading API that quietly became the chain's second-largest DEX aggregator, took $88.45 billion of cumulative volume, and got bought by MoonPay in May 2026 without most traders noticing it existed. Its program ID is DF1ow4tspfHX9JwWJsAb9epbkA8hmpSEAtxXy1V27QBH, and on 1 November 2026 it appeared in 151 of 200 consecutive Solana blocks we pulled.

There is no review of DFlow on the first page of Google — just the vendor's homepage, its X and LinkedIn accounts, a Rajasthan state government portal that happens to share the name, and MoonPay's press release. So this is the measurement nobody has published: every number MoonPay claimed about DFlow, checked against DefiLlama's volume series, Solana mainnet, and DFlow's own keyless API. Two of those claims are accurate to within a percent. One is off by four months. And the headline feature — just-in-time routing — could not be made to engage once in fourteen attempts.

DFlow is a trading API, not an app. You do not open DFlow and swap tokens the way you open Jupiter's swap interface. Developers call GET /order with an input mint, an output mint and an amount; DFlow returns a fully constructed Solana transaction; the app's user signs it and the app submits it to an RPC. DFlow never holds the funds, and the docs say so plainly: "Funds stay in the user's wallet throughout the trade lifecycle."

That makes DFlow a peer of Jupiter's aggregator rather than of a Telegram trading bot. Its customers are other products. MoonPay's announcement names Coinbase, Phantom, Solflare and Kamino as platforms it powers, so if you have swapped a token inside Phantom or Solflare recently, there is a reasonable chance DFlow built the transaction and you never saw the name.

It sells four surfaces: spot trading through one endpoint for any SPL token with no protocol fee; WebSocket market data for top-of-book quotes and ten levels of depth; an identity service called Proof; and an agent stack of a CLI, an MCP server and prebuilt Claude Code skills so an AI trading agent can place orders. What is absent matters as much: no copy trading, no leaderboard, no portfolio tracking, no charts, and — per its own FAQ — no limit orders, "on the roadmap. No timeline yet."

MoonPay announced the acquisition of DFlow on 5 May 2026. The newsroom post calls DFlow "the fastest-growing trading infrastructure platform on Solana" and quotes both CEOs — Ivan Soto-Wright for MoonPay, Nitesh Nath for DFlow, who stays on. Nath is the answer to the "who created DFlow" question that Google surfaces in its People Also Ask box; the company has published no other founder.

The release makes five quantitative claims, and all five are checkable. Here they are against what we measured on 1 November 2026, six months later:

MoonPay's claim (May 2026)	What we measured	Verdict
"More than $50 billion in cumulative trading volume" since April 2025	$50,491,481,962 from 1 Apr 2025 to 5 May 2026	Accurate to 1.0%
"Over $12 billion in Q1 2026 alone"	$12,069,265,646 for 1 Jan – 31 Mar 2026	Accurate to 0.6%
"~10 million transactions per month"	~22.2 million/month at the 8.57 tx/s we clocked	Conservative; has roughly doubled
"More than 85% of Solana blocks" at peak	75.5% of 200 consecutive blocks, off-peak	Consistent with the rate
"First aggregator ever to surpass Jupiter in daily volume," November 2025	It happened — first on 24 July 2025, four months earlier	True, wrong date

That is a better hit rate than most acquisition releases survive. The rest of this article is the working.

DefiLlama's aggregator series puts DFlow's all-time Solana volume at $88,450,471,038 across 725 days, starting 6 November 2024. Almost none of that is early: only $97,220,085 landed before 1 April 2025, which is exactly why MoonPay dates the company's run from that month. Since the acquisition was announced, DFlow has added another $37,861,768,991.

The shape of the growth is the part worth reading slowly. DFlow's share of all Solana aggregator volume, month by month:

Month	DFlow volume	Jupiter volume	DFlow share
Apr 2025	$305.5M	$47.10B	0.6%
Aug 2025	$5.28B	$25.09B	17.4%
Nov 2025 (peak)	$10.31B	$47.40B	17.9%
Mar 2026	$1.48B	$13.71B	9.7%
Sep 2026	$9.33B	$15.09B	38.2%
Oct 2026	$8.08B	$13.88B	36.8%

DFlow's record single day is still $1,873,786,300, on 4 November 2025. But the share column is the real story, and it moved for two reasons at once. DFlow grew 26.4× from April 2025 to October 2026. Over the same stretch, Jupiter's Solana aggregator volume fell from a January 2025 peak of $183,718,828,867 a month to $13,882,253,311 — down 92.4%. A share going from 0.6% to 36.8% is partly one company arriving and partly another company's market shrinking under it.

Across the trailing 30 days, the 41 aggregators DefiLlama tracks on Solana moved $26,242,164,179 between them: Jupiter 51.20%, DFlow 29.65%, OKX Swap 10.64%, Titan Exchange 4.74%, and nothing else above 2%. It is a duopoly with a long tail.

Yes, on 12 of 725 days — and the first was 24 July 2025, not November 2025. Joining the two DefiLlama daily series day by day gives the complete list of days DFlow's Solana volume exceeded Jupiter's:

• 24 Jul 2025: $122.0M vs $51.7M (+136.0%) — the first.
• 13 Sep 2025: $233.0M vs $54.5M (+327.7%) — the widest margin.
• 10 and 14 Oct 2025: +15.1% and +44.1%.
• 15 Nov 2025: $1.226B vs $939.3M (+30.6%) — the only November 2025 day, and the biggest absolute figure.
• 27 Jan 2026: $1.640B vs $736.4M (+122.7%).
• 9 and 11 May 2026, 26 and 27 Sep 2026, 3 and 4 Oct 2026: six more, all by 4–19%.

So the substance of MoonPay's claim holds — a Solana aggregator did out-trade Jupiter on a full day, and DFlow was the one that did it — but the date is off by four months. The November 2025 day was the first time DFlow beat Jupiter at billion-dollar scale, which is plausibly the version that got noticed internally, though the release does not say that.

What it does not mean: twelve days out of 725 is 1.7%, DFlow has never led Jupiter over a full month, and the six most recent overtakes were by 4–19%. Calling DFlow "the aggregator that beat Jupiter" is accurate on the technicality and misleading on the trend. Calling it Solana's clear number two is accurate on both.

The DFlow aggregator program is DF1ow4tspfHX9JwWJsAb9epbkA8hmpSEAtxXy1V27QBH — a vanity address that spells the brand. It is not in the documentation; we pulled it out of a transaction the API built for us and confirmed it against the front-end bundle.

With the program ID in hand, the "85% of Solana blocks" claim stops being marketing and becomes arithmetic. On Sunday 1 November 2026 we fetched 200 consecutive finalized blocks (slots 464,138,926 to 464,139,125) with account keys resolved through address lookup tables. All 200 came back, none skipped, spanning 43 seconds of wall time — 215.0 ms per slot — and carrying 83,083 transactions at 415.4 per block.

Measured over 200 blocks	DFlow	Jupiter v6
Transactions	371	3,015
Share of all Solana transactions	0.447%	3.629%
Mean / median / max per block	1.85 / 1 / 13	15.07 / 7 / 160
Blocks containing at least one	151/200 = 75.5%	198/200 = 99.0%
Failed transactions	23 = 6.2%	2,110 = 70.0%

75.5% against a claimed 85%, measured on a Sunday afternoon when the chain was running at 110,000–191,000 transactions a minute, well below its weekday peaks. MoonPay's claim is explicitly about "peak periods," so these numbers do not contradict it — and the Poisson arithmetic says they actively support it. At the measured rate of 1.855 DFlow transactions per block, the chance a given block contains at least one is 84.4%; we observed 75.5%, the gap being the clustering you would expect from bursty traffic. Double the rate at peak and the figure goes past 90%. The claim is the kind that sounds inflated and is not.

We also measured the same thing a second, independent way, because a single method proving a vendor right is worth less than two agreeing. Pulling 1,000-signature windows for the DFlow program returned 8.55 to 8.62 transactions per second across three samples spanning 116–117 seconds each. At 215 ms per slot that is 1.84 transactions per block — against the 1.85 we counted directly from blocks. Two methods, one number. Extended out, 8.57 tx/s is roughly 22.2 million transactions a month, more than double the "~10 million" MoonPay cited in May.

Jupiter v6 carried 8.1× DFlow's transaction count in our window and failed 11.3× as often. 70.0% of Jupiter v6 transactions in those 200 blocks reverted. 6.2% of DFlow's did. The signature-window samples told the same story from a different angle: Jupiter v6 at 55.6–66.7 transactions per second with 60.6–68.5% failing, DFlow at 8.55–8.62 per second with 10.5–10.6% failing.

MoonPay's release claims JIT routing produces "fewer failed transactions." The direction is right and the magnitude is larger than they bothered to say. But the honest reading is not "DFlow's router is eleven times better."

Jupiter v6 is the default integration target for every Solana sniper bot, arbitrage searcher and memecoin bot on the chain, and a large share of that traffic is speculative by design: fire a hundred, expect most to revert on a slippage check, profit on the few that land. That inflates a failure rate without saying anything about routing quality. DFlow's flow comes through partner apps where a human pressed a button and expects it to work. Different populations, not just different routers.

The number is still useful. If you are deciding where to send retail swap flow, a 6.2% revert rate against a 70.0% one is the difference between a support queue and a quiet one — and failures are not free, since the base fee and priority fee are spent whether the trade lands or not.

Just-in-time routing is DFlow's entire technical pitch, and we could not get it to engage once.

The idea is genuinely good. Every other aggregator computes a route off-chain, bakes it into the transaction, and hopes nothing moves before it lands. On Solana, where proprietary market makers requote faster than a block, something usually does. DFlow's answer is to build the transaction so the route can be rewritten during on-chain execution: at each leg it re-checks prop AMM prices and reroutes inside the same transaction if a better path appeared. Lower realized slippage, fewer reverts, tighter slippage limits.

The Trading API exposes a documented onlyJitRoutes parameter that forces a JIT-only route. We sent it fourteen times against the keyless developer endpoint: six sizes of SOL to USDC from 0.01 up to 1,000 SOL, then SOL into BONK, JUP, WIF, JTO and jitoSOL, then USDC into USDT, SOL and BONK. Every one returned HTTP 400 with {"code":"route_not_found"}, while the identical request without the flag filled instantly.

Before reading anything into that, we checked whether the parameter was honoured at all. An invented parameter — definitelyNotARealParam=true — is silently ignored and returns a normal 200 with a route. So onlyJitRoutes is real, the server acts on it, and it found no JIT route for any pair or size we tried. That sits awkwardly against DFlow's own docs, which say JIT routing needs "no builder configuration or extra parameters" and engages whenever "the route includes a prop AMM leg" — while our quotes routed through BisonFi, Tessera V, AlphaQ, Manifest and Deriverse, exactly the venue class the doc names.

Two explanations survive and we cannot separate them from outside. Either onlyJitRoutes means something narrower than its name and the doc's trigger condition is loose, or the free developer endpoint does not serve JIT routes at all. The production host returns a bare HTTP 403 without an API key, so that comparison was not available to us. We are reporting the measurement and declining the conclusion.

There is a deeper limit that applies to anyone who tries this. A quote endpoint, by construction, cannot show you a benefit that only materialises after the quote. If JIT routing works as described, every price comparison done at quote time understates DFlow — and the only honest way to measure it is to execute thousands of real trades on both routers and compare fills.

With that caveat stated, the quote-time comparison is still worth having, because it is what a builder evaluating the two would run on day one. We queried both APIs within the same second for ten sizes on SOL/USDC, Solana's deepest pair, at 50 bps slippage.

Trade	DFlow route	Jupiter route	DFlow vs Jupiter
0.1 SOL → USDC	Tessera V	GoonFi V2	−0.0010%
10 SOL → USDC	Tessera V	GoonFi V2	−0.0021%
1,000 SOL → USDC	BisonFi ×2 + Deriverse	ZeroFi + TesseraV + GoonFi V2	−0.0019%
10,000 SOL → USDC	Tessera V + BisonFi	GoonFi V2 + TesseraV + BisonFi	−0.0063%
100,000 USDC → SOL	BisonFi + Manifest + BisonFi	GoonFi V2 + BisonFi + TesseraV	+0.0039%
1,000,000 USDC → SOL	Tessera V + BisonFi	4 legs incl. Raydium CLMM	−0.0161%

DFlow was behind on nine of the ten sizes and ahead on one, by margins of 0.001% to 0.016%. Every single delta is under two basis points. On a $1,000 swap the worst case is sixteen cents. For practical purposes the two routers are tied at quote time on the deepest pair on Solana, and anyone claiming a meaningful price edge in either direction on SOL/USDC is reading noise.

One incidental cross-check: DFlow quoted 1 SOL at 94.97 USDC while CoinGecko had SOL at $94.98 the same minute, so the pricing is sane and the aggregator is not quietly skimming.

DFlow's /venues endpoint returns 44 liquidity sources. Jupiter's program-to-label map returns 105. That is the single clearest structural difference between the two, and it is one call each to verify.

DFlow's list covers what matters on volume: Whirlpools, three Raydium programs plus Raydium Launchlab, four Meteora programs, Pump.fun and its AMM, Phoenix, Manifest, Sanctum and Sanctum Infinity, PancakeSwap, Saros, Zora, MetaDAO, and the prop-AMM cohort that now wins most SOL/USDC flow — HumidiFi, SolFi and SolFi V2, ZeroFi, Tessera V, AlphaQ, JanusFi, Aquifer, BisonFi, Obric V2, Byreal, Deriverse, KDEX, Gamma, Ghost, Triangle, GatorSwap and Scorch. There is even a "BisonFi Predictions" venue sitting in the spot router.

Normalising the naming differences (DFlow writes "Whirlpools" where Jupiter writes "Whirlpool," "Tessera V" where Jupiter writes "TesseraV"), 35 venues appear on both sides and only seven are genuinely DFlow-only: Doppler CPMM, Doppler Launch, Ghost, JanusFi, KDEX, Triangle and Zora. Around 70 are Jupiter-only, including Orca V1 and V2, OpenBook V2, Invariant, Saber, Crema, Perena, Heaven, Woofi, Solayer, Hylo and Jupiter's own RFQ and lending venues.

The practical reading: DFlow is built to win on the pairs where the money is, not to find a route for everything. For SOL, major stables and liquid memecoins that is the right trade-off; for the long tail of obscure launchpad tokens and half-abandoned pools, Jupiter will find routes DFlow does not. The token list is not the constraint — /tokens returned a 239 MB JSON array of 5,111,430 mints, the largest routable list we have measured on any Solana API. Having the mint is not the same as having a venue for it.

DFlow charges no protocol fee on spot trades, and it does not take positive slippage. Both are stated flatly in the FAQ, and the /order responses we pulled came back with "platformFee": null every time.

What you actually pay is four things: the Solana base fee of 5,000 lamports per signature (about $0.0005 at today's SOL price); an optional priority fee, for which DFlow's /priority-fees endpoint returned 50,000 / 100,000 / 200,000 microlamports per compute unit at its medium, high and very-high tiers; whatever Raydium, Meteora, PumpSwap or a prop AMM bakes into the leg; and whatever the app in front of you decided to charge.

That last one is the thing to understand about any white-label aggregator. DFlow exposes platformFeeBps, platformFeeMode and feeAccount so the integrating app can take a cut in either the input or output token, plus positiveSlippageFeeAccount with a positiveSlippageLimitPct cap so the app can keep favourable price movement too. DFlow taking nothing does not mean your swap was free — it means the fee decision belongs to whoever built the app, and you will not find it in DFlow's docs.

Credit where it is due on one detail: the FAQ actively tells builders not to set platformFeeBps if they are not collecting a fee, because the API folds the declared fee into the slippage tolerance and "declaring a fee you don't collect wastes slippage budget and gives users worse pricing." A vendor volunteering something that costs it nothing and saves users money is rare enough to mention.

The documentation is unusually good and unusually honest, which makes the rough edges easy to find. pond.dflow.net/llms.txt indexes 55 pages and llms-full.txt ships the whole corpus as 146 KB of clean markdown — if you are pointing an agent at DFlow, start there rather than scraping.

Things we verified, or that the docs admit:

• The keyless developer endpoint works. dev-quote-api.dflow.net needs no API key and returned /order responses in 30–50 ms. Production needs a key via Google Form, 2–5 business days.
• You cannot call it from a browser. The Trading API sets no CORS headers, so every integration needs a backend or edge proxy. The docs say so.
• Stream access is a separate permission. A production key that trades fine on /order can still be refused on the quote and book streams until DFlow enables it on that specific key.
• Mint addresses only. No symbol resolver — "USDC" is not a valid input, EPjFWdd5… is.
• Default slippage is 20 bps. Omit slippageBps and the response comes back with it already applied.
• It has moved to Solana v1 transactions, with a migration guide and a transactionVersion parameter — a post-Alpenglow detail a lot of tooling has not caught up on.

The program itself is actively maintained rather than a frozen artifact. DF1ow4tsp… is owned by the BPF Upgradeable Loader and was last deployed at slot 462,900,625 — 29 October 2026, 13:38 UTC, three days before we looked. Its upgrade authority 9uo2iJKJkmiYwy5MrkrbGQXbPqwHXqvQaUqxhCkowgsh is still set rather than revoked, holding 1.2249 SOL in a system-owned account with no data — a shape indistinguishable from a Squads multisig vault from outside, so it proves the program can be upgraded and nothing about who can do it.

Two gaps worth knowing: DefiLlama lists DFlow's audit count as 0, and verify.osec.io returns is_verified: false with no repository on file, meaning no reproducible build has been published. That is not a claim the bytecode differs from any source — it is the absence of a way to check, on infrastructure handling $8 billion a month.

DFlow built the first API that tokenizes Kalshi's prediction markets on Solana, and the product has been close to dormant since July 2026.

MoonPay's release devotes a section to it: every Kalshi market position represented as a native Solana token, minted and settled through DFlow, so apps can embed forecasting markets and positions can move through standard token interfaces. It is genuinely novel plumbing — regulated off-chain liquidity exposed as SPL tokens.

DefiLlama tracks it as a separate protocol, and the volume curve is steep in both directions: $1,836,927 in December 2025, peaking at $6,111,897 in February 2026, down to $133,421 in July 2026, then exactly $0 every day from August onward — 122 consecutive days to the day we checked. All-time volume is $17,311,389 over 344 days, 220 of them non-zero, the last being 2 July 2026.

A clean $0 in a data adapter is usually a broken adapter rather than a dead product, so we went to the chain. The prediction program is pReDicTmksnPfkfiz33ndSdbe2dY43KYPg4U2dbvHvb — the adapter's own source names it, along with the CASH and USDC settlement vaults and a fee account. Paginating its full signature history gives a more precise picture than "zero": 7,637 transactions in June 2026, 3,868 in July, then 179 in August, 220 in September, 95 in October, and one on 1 November, timestamped 14:58:08 UTC — sixteen minutes before we pulled it.

So the product is alive and it is not at zero. It went from roughly 250 transactions a day in June to roughly three a day in October, a 98.8% collapse between June and August, and DefiLlama's exact $0 is wrong in the strict sense while being right about the substance. Why the residual transactions produce no measured volume we cannot resolve from outside: either they carry no settlement-vault transfers, or the adapter's data source stopped returning rows for them.

The other tell is documentation. DFlow's docs index covers spot trading, Proof, the agent tools and the full API reference across 55 pages, and prediction markets appear on none of them. There is a marketing post about the Kalshi API on dflow.net and a route for it in the front-end bundle, but the developer documentation a builder would need does not exist in the published corpus.

One checkable detail on the collateral. The CASH token these markets settle in (CASHx9KJUStyftLFWGvEVf59SGeG9sh5FfcnZMVPCASH) is a Token-2022 mint with a supply of 128,963,901.64 and metadata served from token-metadata.bridge.xyz. Its extensions include a permanent delegate and a confidential-transfer authority, meaning the issuer retains the ability to move CASH out of any holder's account. Normal design for regulated stablecoin collateral, abnormal if you assumed you were holding something like USDC.

Two side products say where MoonPay is pointing this.

Proof is an identity service. A partner app sends a user to dflow.net/proof, the user completes KYC through Stripe Identity, and the app can then ask DFlow whether a given wallet belongs to a verified individual. The docs say there is no fee, and spot trading explicitly does not require it — it exists so apps can gate features by identity or jurisdiction without building verification themselves. One gotcha the docs volunteer: Proof only redirects back to https:, moz-extension:// or chrome-extension: URLs, so a native app using myapp://callback fails silently and needs universal links instead.

The agent stack is the other half: a one-line-install CLI that provisions a Solana wallet and trading capability for an agent in a single step, a hosted MCP server that plugs DFlow's docs and API into Claude Code, prebuilt skills, and operator guardrails for trade size limits, approved assets and rate controls. Helius documents a DFlow trading skill of its own, which is why a Helius docs page ranks in the top ten for the brand name. If you are building an open-source trading bot or an agent that needs Solana execution, this is the most agent-native aggregator API currently shipping, and the guardrails are the part most comparable products skip.

DFlow does not do copy trading, and nothing in its 146 KB documentation corpus suggests it ever will. It is a router: it answers "what is the best way to execute this swap right now," and it answers it well. It does not answer "which trade should I make," which is what most people searching for a Solana trading tool are actually asking.

Those are complementary layers. An aggregator optimises a decision you already made; the decision still has to come from somewhere — your own research, a wallet tracker, or a trader whose results you can verify. That is where copy trading sits.

uwuu operates at that decision layer. You pick a trader from a verified on-chain leaderboard and uwuu mirrors their trades into your own wallet in real time at sub-400ms execution, non-custodially, with smart trade filtering so you are not forced to copy everything. The fee is performance-based — you pay when you profit. How to copy trade on Solana walks through the setup, and the best Solana trading bot comparison covers where every category of tool fits.

What is DFlow on Solana?

DFlow is a trading API and DEX aggregator built natively for Solana. Apps call its /order endpoint, get back a signed-ready Solana transaction, and the user's wallet signs it — DFlow never custodies funds. Its on-chain program is DF1ow4tspfHX9JwWJsAb9epbkA8hmpSEAtxXy1V27QBH and it has routed $88.45 billion of cumulative volume.

What is DFlow used for?

Developers use it to add token swapping to an app without building a router. It aggregates 44 Solana liquidity venues, charges no protocol fee on spot trades, streams live quotes and order book depth over WebSocket, and lets the integrating app set its own platform fee. It also offers an identity service called Proof and a CLI and MCP server for AI trading agents.

Who created DFlow?

Nitesh Nath founded DFlow and remains its CEO. MoonPay acquired the company and announced it on 5 May 2026; Nath stayed on to run the execution layer inside MoonPay. No other founder has been publicly named.

Is DFlow better than Jupiter?

On price, they are tied: across ten SOL/USDC sizes quoted in the same second, every difference was under two basis points, with Jupiter marginally ahead on nine of ten. Jupiter aggregates 105 venues to DFlow's 44, so it finds more long-tail routes. DFlow's measured transaction failure rate was 6.2% against Jupiter v6's 70.0%, though Jupiter's figure is inflated by bot traffic that expects to fail.

Can I use DFlow on mobile?

Not as a standalone mobile app — DFlow does not publish one. You use it indirectly, inside apps that integrate it, including mobile wallets. The API also cannot be called from a browser at all, because it sets no CORS headers, so every integration routes through a backend.

Does DFlow have a token or an airdrop?

No. DFlow has never launched a token, and as a MoonPay subsidiary it has no stated plan to. Search demand for "dflow token" and "dflow airdrop" exists, but there is nothing behind it — treat any token claiming to be DFlow as unaffiliated.

Is DFlow audited?

DefiLlama lists DFlow's audit count as 0, and verify.osec.io reports the aggregator program as unverified with no source repository on file, meaning no reproducible build has been published. The program's upgrade authority is still active rather than revoked. None of that implies a problem with the code; it means the usual external checks are unavailable.