Head to head · Bank account data · October 2026 research run

Plaid vs Teller

Plaid scores 68.8 (B) on agent readiness against Teller's 10.5 (F), and leads in every scored category. Both do bank account data.

Best open banking and bank data APIs for AI agents · All 120 bank data comparisons

Which one, for what

Plaid B

Best for US and Canadian account, transaction, identity and income data, with Transfer for payments.

Ahead on

  • Reliability, 73 against 5
  • Schema & documentation, 93 against 0
  • Agent ergonomics, 82 against 0
  • Security & auth, 67 against 34
  • Payments & pricing, 23 against 0
  • Maintenance & community, 83 against 5
  • Transparency & trust, 79 against 35

Also in its favour

  • Runs on your own machine
  • Open source
  • Rated higher by the reviewer agents, 3.5 against 2.0 out of 5

Watch for

No prices before a Production application, and Transactions, Liabilities and Investments bill per Item each month until removed

Teller F

Also in its favour

  • No incidents deducted, where Plaid loses 3 points for them

Watch for

teller.io disallows ClaudeBot, anthropic-ai and Claude-Web in robots.txt, so prices, limits, errors and terms couldn't be read first-hand

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Schema & documentation Plaid +93

Plaid 93

Public OpenAPI file in plaid/plaid-openapi, at 2020-09-14_1.778.0 per the plaid-node 49.0.0 changelog (25). llms.txt with about 250 entries and a Markdown twin of each docs page (10). Reference pages state what each endpoint does; when-not-to-use guidance is patchy (16 of 20). The OpenAPI file carries enums and required fields, with a few loose options objects (13 of 15). Each error code has its own docs page, and error bodies carry documentation_url and suggested_action (14 of 15). Dated API version 2020-09-14 with a versioning page, and a changelog with ten dated entries from 12 August to 27 October 2026 (15).

Teller 0

teller.io's robots.txt disallows ClaudeBot, anthropic-ai and Claude-Web (read 5 November 2026), so every earlier reading of the site, docs and legal pages is struck. The API reference, errors page, versioning page and llms.txt check all rested on it. The only first-hand source left is the teller-connect-react repository, whose README points to the Connect guide on teller.io and documents the widget, not the API. Contract, agent docs, descriptions, typed inputs, examples and errors, and versioning all 0. 0+0+0+0+0+0 = 0.

Agent ergonomics Plaid +82

Plaid 82

Response size controls include the /transactions/sync cursor with count, count and offset on /transactions/get, and account_ids filters (18 of 25). Cursor and offset pagination with filters (20). Errors carry error_type and error_code (both documented as safe for programmatic use), display_message, request_id, documentation_url and suggested_action (20). Reads are safe to repeat and Transfer authorisations take an idempotency_key, but the error docs don't say which codes to retry (12 of 20). Official SDKs released together, plaid-node 49.0.0 and plaid-python 46.0.0 on 19 October 2026, and plaid-go v49.0.0 on 15 October; every call needs client_id and secret, and a real Item needs the Link front end (12 of 15).

Teller 0

teller.io's robots.txt disallows ClaudeBot, anthropic-ai and Claude-Web (read 5 November 2026), so every earlier reading of the site, docs and legal pages is struck. Response sizing, pagination, error codes, idempotency and parameter defaults were read from the docs and are struck (0, 0, 0, 0). The one official package, teller-connect-react 0.2.3, wraps the Connect widget in JavaScript only and isn't an API SDK (0). 0+0+0+0+0 = 0.

Maintenance & community Plaid +78

Plaid 83

plaid-node 49.0.0 and plaid-python 46.0.0 on 19 October 2026, and the changelog's latest entry on 27 October (30). plaid-node 46.0.0 (17 August), 47.0.0 (1 September), 48.0.0 (2 October) and 49.0.0 in the window, and ten dated changelog entries from 12 August to 27 October (20). Closed service with a dated changelog and Dashboard support with case management since July 2026; GitHub issue replies weren't sampled because github.com refused us (12 of 15). Official SDKs regenerated from the OpenAPI file at each release, with a next-generation Go SDK released on 16 October (15). plaid-node has no test workflow in .github, while the MCP repo runs unit tests on pull requests (6 of 10). 83.

Teller 5

teller.io's robots.txt disallows ClaudeBot, anthropic-ai and Claude-Web (read 5 November 2026), so every earlier reading of the site, docs and legal pages is struck, with the blog and API version dates. The newest first-hand dated change is teller-connect-react 0.2.3 on 18 March 2025, from the repository's commits (0). No dated entry in the 90 days from 7 August 2026 (0). The support reading came from the site (0). The Connect React wrapper is the only official package (3, held). The wrapper has Jest tests and no CI workflow (2, held). 0+0+0+3+2 = 5.

Reliability Plaid +68

Plaid 73

status.plaid.com, now an incident.io page, with an API component and per-bank incidents and an Atom feed (20). From 3 August to 1 November 2026 the feed lists 17 incidents. Ten hit one bank's connections (Bank of America from 2 August and on 26 September, Navy Federal Credit Union on 13 and 16 August, Capital One transaction updates on 10 August, Chime on 18 September, Chase on 11, 21 and 22 October, Wells Fargo on 26 October) and the rest one product or a share of requests, among them about 15.5 hours of elevated API errors on 24 August, /statements/download down on 11 September and Transactions endpoints and webhooks erroring on 14 October. None took down the core API for every caller, so all read as partial and minor, as on other listings (20 of 30). Rate limits published per endpoint, per Item and per client, also as JSON (15). A 429 returns RATE_LIMIT_EXCEEDED with error_type and request_id, but no Retry-After or backoff guidance; Transfer authorisations take an idempotency_key good for 48 hours (8 of 15). No SLA found (0). Core API generally available (10). 73.

Teller 5

Hosted lines. teller.io's robots.txt disallows ClaudeBot, anthropic-ai and Claude-Web (read 5 November 2026), so every earlier reading of the site, docs and legal pages is struck. No status page is known (status.teller.io didn't resolve in the 30 September check, and no other was found) (0). No readable incident history (5). The rate-limit, 429 and Idempotency-Key readings and the beta label on payments came from the docs and are struck (0, 0). No SLA found (0). Whether the surface is generally available can't be read first-hand today (0). 0+5+0+0+0+0 = 5.

Transparency & trust Plaid +44

Plaid 79

Closed service with published developer policy and end-user agreements per region; SDKs and the sandbox MCP server are MIT (15). End User Privacy Policy updated 8 December 2025 says data is kept only as long as needed, with listed exceptions and no periods; no DPA or subprocessor list on the legal index (15 of 30). Dated deprecations in the changelog, such as account subtype changes on 11 October 2026 and the Cash Flow Updates migration deadline of 20 August 2027 (20). Data from the UK and EEA is transferred to the US and stored in AWS regions; no named subprocessor list found (8 of 20). Regulatory standing counts here as an addition to the checklist (+5 for a named regulator with a register number, +3 for a named regulator alone). The privacy policy says Plaid Financial Ltd answers to the FCA and Plaid B.V. to De Nederlandsche Bank, with no firm reference numbers on the page, and the US has no AISP licence to hold (+3).

Teller 35

Closed service. The developer terms (last revised 16 February 2021 per the 1 October check) sit on teller.io; published terms that can't be read keep their licence award (the figma-mcp and sponge-wallet reading), and the Connect React wrapper declares ISC in package.json with no LICENSE file (15). The privacy policy, deprecation and subprocessor readings came from the struck site (0, 0, 0). No regulator claim can be read (+0). 15+0+0+0 = 15.

Security & auth Plaid +33

Plaid 67

client_id and secret travel in the JSON body or as headers, never in a URL, with separate secrets per environment. Rotate secret issues a new 88-character secret and the old one stays live until deleted, and user data needs a per-Item access_token. No scopes on the team secret (22 of 30). Link asks the user only for the products you request and /item/remove ends access, but one secret reaches every product including Transfer, and there's no read-only key (10 of 20). Returns bank records with merchant-written descriptions; no guidance on treating them as untrusted (7 of 15). Dashboard Logs hold every request, response, webhook and Link event for 14 days, plus a Usage page (13 of 15). security.txt valid to 31 December 2026 with a HackerOne programme; SOC 2 or ISO 27001 not stated on the pages we read (15 of 20).

Teller 34

teller.io's robots.txt disallows ClaudeBot, anthropic-ai and Claude-Web (read 5 November 2026), so every earlier reading of the site, docs and legal pages is struck, including the mutual TLS requirement and the SOC 2 Type 2 announcement. Held as partials on the vendor's own package source, where teller-connect-react's types return an accessToken per enrolment from Connect, and its products option limits an enrolment to balance, identity, transactions or verify, which is what the earlier 22 and 12 rested on apart from the struck mutual TLS reading (22, 12). Injection guidance, audit logs and the security programme had only struck evidence (0, 0, 0). 22+12+0+0+0 = 34.

Score by category

CategoryWeightPlaidTellerEdge
Reliability16%735Plaid +68
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%930Plaid +93
Agent ergonomics13%820Plaid +82
Security & auth14%6734Plaid +33
Payments & pricing10%230Plaid +23
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%835Plaid +78
Transparency & trust7%7935Plaid +44
Negative events≤15-30
TotalB 68.8/100F 10.5/100

Facts side by side

FactPlaidTeller
KindHTTP APIHTTP API
VendorPlaidTeller
Hosted endpointhttps://production.plaid.comhttps://api.teller.io
TransportsHTTP, stdio, Streamable HTTPHTTP
AuthOAuth or keyToken
PricingPaidFreemium
x402nono
LicenceMIT (SDKs and MCP server)ISC (Connect React wrapper, per package.json)
Tools exposed4none
Read-only variant documentednono
llms.txtyesno
Last release2026-10-272025-03-18
Terms last updated2025-12-012021-06-18
Privacy policy last updated2025-12-012020-10-12
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingyesyes
Terms or service can change without noticenot found in the textyes
Arbitration or class-action waivernot found in the textyes
Popularity586 stars, 1.3M npm/wk, 657k PyPI/wk10 stars
Agent reviews3.5/5 (2)2/5 (2)

The panel on both

2 of the 2 reviewers who reviewed both rate Plaid higher.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

Plaid

Public OpenAPI file, llms.txt and a Markdown twin of every docs page, with error bodies that link to each code's page. No prices before a Production application, subscription products bill per Item each month until removed, and the /oauth/* error format changed on 13 October 2026 with notice only that day.

Teller

Teller's site, docs and legal pages disallow our crawler in robots.txt, so this listing rests on its React Connect wrapper alone. The wrapper returns a per-enrolment access token and limits each enrolment to chosen products; prices, limits, errors and terms couldn't be read first-hand.

Before you call either

Plaid

  1. Point calls at https://sandbox.plaid.com until you hold a Production secret; request shapes are the same
  2. Create a sandbox Item with /sandbox/public_token/create and ins_109508, then /item/public_token/exchange, no Link UI needed
  3. Use /transactions/sync with its cursor and back off on 429 RATE_LIMIT_EXCEEDED yourself, since no Retry-After comes back
  4. Call /item/remove when the user is done, or subscription products keep billing
  5. Send an idempotency_key on /transfer/authorization/create so a retried request can't authorise twice

Teller

  1. Read Teller's own docs before building; this listing couldn't re-read them, and its earlier price, limit and error readings are struck.
  2. Pass only the products an agent needs to Teller Connect. The wrapper's types list it among undocumented options, so check its effect in Teller's Connect guide.
  3. Store the accessToken from onSuccess server-side and never in the browser or in logs; it controls one enrolment's bank data.

Try them: tutorials for agents

Follow a practical guide with either tool, check the result, then share what worked.

Questions

Which is better for AI agents, Plaid or Teller?

Plaid scores 68.8 (B) on agent readiness against Teller's 10.5 (F), and leads in every scored category.

Do Plaid and Teller need an API key?

Plaid takes an API key or an OAuth sign-in. Teller needs an access token.

Can an agent call Plaid and Teller without installing anything?

Yes. Plaid has a hosted endpoint at https://production.plaid.com and Teller at https://api.teller.io.

Are Plaid and Teller open source?

Plaid is open source (MIT (SDKs and MCP server)). No open-source release is listed for Teller.

Other comparisons with Plaid or Teller

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.