How Broker API Comparator works

Why these five brokers

The comparator covers five retail broker APIs that together represent the overwhelming majority of AI-driven retail trading setups among US and EU operators in 2026:

• Alpaca — default choice for solo operators thanks to a free paper tier with live-parity and an official MCP server.
• Interactive Brokers — the only realistic option for multi-asset, international, or scale use cases.
• Tradier — options-first retail niche with a genuine API focus.
• Schwab — the API successor to TD Ameritrade's developer program; still maturing but too large to ignore.
• Robinhood — included because readers repeatedly ask, even though its public API story remains weak.

What we don't cover

• Futures-first brokers (NinjaTrader, AMP Futures) — different audience, different feature surface.
• Crypto-native venues (Coinbase, Kraken) — covered separately in exchange-focused tools.
• Institutional and professional-only platforms (Bloomberg Terminal, Refinitiv) — out of scope for retail decisions.
• tastytrade — API exists and is improving, but we currently don't have enough first-hand usage to grade it fairly.

Data sources

All facts reflect published developer documentation as of 2026-04-22.

• Alpaca Markets documentation
• Interactive Brokers API
• Tradier developer portal
• Schwab developer portal
• Robinhood documentation

MCP availability is cross-referenced with the Finance MCP Directory rubric.

Fit-score formula

The score is deliberately simple: it's the count of active filters that the broker satisfies, out of the total number of active filters. The "Max auth complexity" slider is always an active criterion; the other "Must support" filters only count when the user has turned them on.

fit_score = count(active_filters_satisfied)
max_score = count(active_filters)
fits      = (fit_score == max_score)

No weighting. No hidden tier-bias. If two brokers tie on fit-score, the one with lower auth complexity wins — on the assumption that, all else equal, a simpler integration is better. Reading the full feature matrix is still recommended before committing, which is why the "Compare full table" toggle is one click away.

Auth complexity rubric (1–5)

• 1 — static API key in a header, no token refresh.
• 2 — API key + per-request signing or HMAC.
• 3 — OAuth 2.0 with a simple refresh-token loop.
• 4 — OAuth 2.0 with 3-legged flow, strict redirect rules, or app-review gating.
• 5 — requires a persistent gateway process running on the client machine (e.g. IB Gateway / TWS).

Rate-limit caveat

Displayed rate limits are the vendor-published tier for a typical retail account. Real-world effective limits are frequently lower under concurrent load, burst behavior, or per-endpoint throttling that the docs don't explicitly surface. For high-frequency strategies, measure the actual limits on a paper/sandbox account before relying on the published figure.

MCP status

Three-level classification:

• Official — first-party MCP server maintained by the broker.
• Community — third-party MCP server actively maintained; audit before production use.
• None — no MCP server currently listed or maintained beyond experimental forks.

Refresh cadence

Data is refreshed quarterly, or immediately when a broker announces a material API change (new auth flow, MCP launch, rate-limit change, new asset class). The most recent refresh date is shown at the top of the comparator and on this page. Corrections are logged at /corrections/.

Limitations

1. Published data only. Private API tiers, enterprise quotas, and partner-only features are not modeled.
2. Regional eligibility not modeled. Several brokers restrict account opening by country; verify eligibility directly.
3. Order-type nuance lost. "Supports OCO" does not distinguish between true multi-leg OCO groups and simple paired cancellation.
4. Fee disclosure is coarse. PFOF-model commissions are shown as "$0" even though effective spread costs are non-zero.
5. MCP grade is a snapshot. Community servers rise and fall quickly; the status here is re-verified each refresh cycle.

Editorial independence

The comparator may reference brokers with current or future affiliate relationships with AI Fin Hub. The fit-score algorithm is deterministic and applied uniformly. No affiliate or sponsor payment alters rank, score, or feature presentation. See /sponsor-disclosure/.

Changelog

• 2026-04-22 — Initial release with 5 brokers (Alpaca, IBKR, Tradier, Schwab, Robinhood).