Brokerage Account Aggregation APIs

One question removes most of this market before any other: do you need to place orders, or only read data? Almost everything here is read-only and will never be otherwise — it is a different regulatory posture, not a missing feature. After that, coverage is decided by how each broker is connected rather than by how many brokers a vendor counts, and refresh is a daily batch rather than a live feed, whatever the tier is called.

• Reading positions and placing orders — SnapTrade, a $100 monthly platform fee plus $1–2 per connected user, and the only published production price in the category.
• The broadest US coverage, read-only — Plaid Investments, 3,213 institutions with investments in its own coverage file.
• Official broker APIs only, nothing scraped — Akoya.
• Canadian accounts, bank-owned brokerages and group plans included — Wealthica Business, 125 published providers, keys from sales.
• Holdings with CUSIP, ISIN and SEDOL — Yodlee, the old Envestnet | Yodlee.
• A sandbox key this afternoon and an institution list to check — Mastercard Open Finance, the old Finicity API.
• The aggregator a bank already runs — MX, with investment holdings billed as a product of their own.

The layer nobody goes looking for by name

This category sits between the two it connects. A market data API tells you what a security is worth. A portfolio tracker tells a user what they own and what it did. The thing that gets the second one its holdings — without asking anyone to upload a CSV every Sunday — is an account aggregation API, and most people arrive here having wanted it for weeks without knowing the layer had a name.

That is the honest description of this listing. If you are shopping for a tracker, this is what the tracker you pick is built on, and it explains why two trackers support wildly different broker lists. If you are building one, this is the vendor decision that determines your broker coverage, your support load and a line item that scales with every user you add.

There is one job here — aggregation — and no sub-facets worth inventing. Everything that varies varies along the axes below.

Read-only or trading — the split that removes most of the field

Almost everything in this market returns data and cannot place an order. That is not a gap the vendors are working on; it is a different regulatory posture and, in most cases, a different set of agreements with each broker. Plaid, Akoya, Yodlee, Mastercard Open Finance, MX and Wealthica Business are read products. SnapTrade is the conspicuous exception that submits orders as well.

The line matters more than it looks, because it is not a feature you bolt on later. If you want execution, most of this market is not a candidate and never will be, and you are choosing among a handful of vendors on their terms. If you only need to read, you are in a large competitive market where the interesting differences are coverage and price.

Even inside a trading-capable vendor the default is read. At SnapTrade, read access and trade access are separate grants; its $1 daily plan is read-only whatever the broker allows; and 24 of its 40 integrations are read-only on any plan — Fidelity, Vanguard US, Robinhood, Interactive Brokers and Questrade among them. Check the broker list against the write side before you design an order flow around it.

How the connection is actually made, and why vendor counts are not comparable

Every vendor publishes a number of supported institutions. Those numbers mix three mechanisms with completely different failure rates, and no vendor breaks the count down on the marketing page.

Credential-based access. The user hands over their brokerage username and password, and the vendor logs in as them and reads the screen or an internal endpoint. It reaches institutions that have no API at all, which is why it still exists. It also breaks when the site is redesigned, when the password is rotated, and — increasingly — when the broker enforces multi-factor authentication on every login, which is the change that has been quietly killing this method institution by institution. Some brokers' terms prohibit it outright.

OAuth token access. The user authenticates on the broker's own domain and grants a scoped, revocable token; the vendor never sees a password. Far more stable, and the only method that survives an MFA mandate cleanly. It does not settle the write side: at SnapTrade, the eleven brokers that need per-app approval — Chase, Fidelity, Schwab and Wells Fargo among them — are read-only without exception.

An official API partnership with the broker. A negotiated agreement and a real broker-operated API behind it. This is the most reliable and the narrowest. Akoya is built entirely this way and says so as a selling point — nothing scraped, every connection permissioned through the institution. The cost is reach: a provider that has not joined is unreachable, with no scraping fallback, and Fidelity is its anchor brokerage while most of the rest of the network is banks and their wealth arms.

The practical test when you evaluate coverage — take the ten brokers your actual users hold accounts at and find out which mechanism each one is on. Ten brokers on OAuth beats three hundred on scraped sessions for anything you intend to support. Only some vendors let you do that before signing. SnapTrade and Wealthica Business publish the mechanism per institution; Plaid publishes a coverage file; Mastercard Open Finance has a Get Institutions endpoint a sandbox key can call. MX and Yodlee publish no institution list, and Akoya's provider directory sits behind its console login.

Canada is its own market. Wealthica Business is the only aggregator here built for it: its Supported Providers page listed 125 entries — 21 banks, 84 brokerages and 20 others — including the bank-owned brokerages, group retirement plans and advisor platforms, and it is a credential network, 117 of the 125 by its own classification. The alternatives are thin there: Plaid's coverage file lists 15 Canadian institutions with investments, and SnapTrade has ten Canadian integrations, with RBC Direct Investing and Scotia iTRADE absent.

What comes back, and how stale it is

Positions versus a full picture. Current holdings with quantity and market value are the easy part and everyone has them. Cost basis, realised gains, transaction history and per-lot detail are where products separate, and where a tax or performance feature lives or dies. Plaid documents cost basis on holdings, up to 24 months of investment transactions, and tax lots where the institution provides them. Akoya's FDX-shaped endpoints cover holdings and a tax-lot endpoint with cost basis. MX returns cost basis without lots, and 90 days of transactions unless you buy Extended Transaction History. On a scraped connection the same fields quietly come back null.

Identifiers. If the holdings have to join to a securities master, the identifier on each position decides how much reconciling you do. Yodlee carries CUSIP, ISIN and SEDOL as fields on every holding. Wealthica Business has CUSIP but no ISIN. Mastercard Open Finance returns a security ID with its type beside it, which is how a CUSIP arrives. Plaid nulls CUSIP and ISIN unless you hold a CUSIP Global Services licence, and MX has neither — you match on symbol and a description string.

Refresh is a daily batch, not a live feed. This is the expectation most often wrong on arrival. Plaid checks investment accounts at least once per market day and up to two to four times depending on the institution, with most changes landing overnight after the close. That is normal for the category, because the brokers themselves post holdings on an end-of-day cycle. An app that promises intraday portfolio value is pricing positions from a market data feed and getting only the position list from here. Any vendor tier sold as "real-time" means a fresher pull on request, not a stream.

A stale connection is your support ticket, not the vendor's. Connections break — a password change, an MFA prompt, a broker redesign — and the user finds out through your app. Whatever the reconnection flow looks like, that is what your inbox looks like.

The rules are mid-rewrite, and that is the load-bearing fact

Do not build on the assumption that US brokers are legally obliged to expose your users' data. As of September 2026 they are not.

The CFPB finalised the Section 1033 Personal Financial Data Rights rule — the American open banking rule — in late 2024. It was challenged by Forcht Bank, the Kentucky Bankers Association and the Bank Policy Institute in the Eastern District of Kentucky. In February and March 2025 the court paused the case for 90 days and pushed the compliance dates back by the same period, moving the first from 1 April to 30 June 2026. On 29 July 2025, after the CFPB announced it would reconsider the rule, the court stayed the case pending the new rulemaking; that stay left the deadlines alone. On 29 October 2025 the court postponed the rule itself and enjoined the CFPB from enforcing it until the reconsideration is complete, so the 30 June 2026 date bound nobody. An advance notice on 22 August 2025 reopened four questions, including whether data providers may charge for access, and a replacement proposal reached OIRA for review on 4 August 2026. It had not been published by 27 September 2026.

Two consequences you can act on:

Access is a commercial relationship, not a right. The provision that would have made data access free is the one being rewritten. JPMorgan Chase began billing aggregators for data pulls, and Plaid signed a paid data-access agreement with it rather than litigate. Expect the cost of reaching large institutions to be inside your vendor's price, and expect that price to move.

The technical standard is settling even while the law is not. The FDX API is the de facto format for this data in North America — the CFPB recognised Financial Data Exchange as the first standard-setting body under the 1033 framework, a five-year recognition running to January 2030, and FDX's Spring 2025 survey counted roughly 114 million customer connections running through FDX-aligned APIs in the US and Canada — connections, not accounts, and the latest figure FDX has published. Its 6.0 release in December 2023 updated the tax-document structures. A vendor exposing FDX-shaped responses is betting on the same standard, and that bet looks safe regardless of how the rule lands.

Pricing, and how thin the free tier really is

This is the rare category where almost nobody publishes a number. The shapes you will meet are per connected user per month, per successful API call, or an annual contract with a floor — and which one you are quoted depends on your volume more than on the product.

SnapTrade is the exception that lists prices — a free Build plan capped at 5 connected accounts for development, then Launch: a $100 monthly platform fee that includes no users, plus $1 per connected user per month for daily data or $2 for real-time, billed per user rather than per connection, so a user who links three brokers still counts once. Plaid bills Investments as a per-Item subscription with holdings and transactions as separate line items, and shows you the rate only after you apply for production access. Akoya splits customers at 10,000 monthly connections and prints no figure on either side. Mastercard Open Finance publishes its model without its rates: Data Access Tiers billed per customer per month, where Tier 1 — account name, type, status and currency — is at "no additional cost" but exists only inside a paid contract, and holdings arrive at Tier 2. MX bills per product and per enrolled user, and a user is enrolled in Investment Data the moment you call a holdings endpoint for them. Yodlee and Wealthica Business publish nothing, though both document that parts of the payload — asset classification at Yodlee, webhooks at Wealthica — sit on a premium plan.

For someone building alone, the sandbox is the first filter. Five of the seven let you create one yourself without a sales call: SnapTrade, Plaid, Akoya, Mastercard Open Finance and Yodlee. Two go further and reach real accounts for free — Plaid's Trial plan, for new US and Canadian teams, allows 10 production Items for life, and SnapTrade's Build plan five connected accounts. Yodlee's sandbox lacks webhooks and several services that appear only in the environment provisioned after a contract. MX and Wealthica Business have no self-serve route at all: MX's development credentials come with becoming a customer and reach only its own test institutions, and Wealthica's keys are issued by email from sales, with a Demo provider to test against only once you have them. Past the sandbox, the obstacle is moving to production against real institutions, where the price appears, the contract appears, and in some cases a review of what you intend to do with the data appears. Write the integration in the sandbox by all means. Do not commit to a launch date until you have the production number in writing.