Head to head · Commerce catalogues · October 2026 research run

Commerce Layer API + MCP vs Swell

Commerce Layer API + MCP scores 67.6 (B) on agent readiness against Swell's 57 (C), and leads in 4 of 7 scored categories. Both do commerce catalogues.

Best commerce platforms and checkout APIs for AI agents · All 189 commerce comparisons

Which one, for what

Commerce Layer API + MCP B

Best for Teams that want a hosted headless backend and an agent that can work across every commerce object with scoped OAuth roles.

Ahead on

  • Reliability, 88 against 67
  • Schema & documentation, 83 against 71
  • Security & auth, 64 against 42

Also in its favour

  • Rated higher by the reviewer agents, 3.5 against 2.5 out of 5

Watch for

No price between the free plan and a sales-quoted Enterprise contract

Swell C

Best for Developers who want a hosted headless store with a flexible data model and an agent that writes server-side.

No category where it leads by five points or more, and no fact that sets it apart.

Watch for

One full-access secret key per environment, with no scoped or read-only key

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.

Security & auth Commerce Layer API + MCP +22

Commerce Layer API + MCP 64

OAuth 2.0 tokens throughout. Integration credentials follow a custom role set per resource and per operation, sales channel tokens are scoped to a market, and the Core MCP accepts the same tokens or runs the OAuth flow (28). The docs tell you to give the agent a dedicated role with minimal permissions, but delete_resource has no documented confirmation step (12). Tool results include merchant- and shopper-entered data with no prompt-injection guidance found (5). The versions endpoint, which gave change history per resource, was removed on 8 May 2026. Event stores remain, with a retention policy added on 18 June, but since 2 September 2026 they and the Dashboard resource inspector are Enterprise only (6). SOC 2 Type 2, ISO 27001 and PCI DSS Level 1 claimed on the security page, with annual penetration tests and fixes within 90 days. The SDK's SECURITY.md gives two email addresses. No security.txt and no bug bounty found (13).

Swell 42

One secret key per environment (test or live) with full access to every collection, revocable, plus public keys for the Frontend API (15). No scoped or read-only secret key found. Role-based permissions are listed only on the Unlimited plan (5). The API returns merchant- and shopper-entered content with no prompt-injection guidance found (5). An events audit log in the developer console shipped on 30 September 2026 (10). The Security and Privacy page at /legal/security, not found on 1 October, has a Vulnerability Reporting section asking reporters to email Swell, with no acknowledgement time, and states annual third-party penetration tests, encryption at rest and in transit, and breach notice within 72 hours. No security.txt (the path still redirects to itself), bug bounty or SOC 2 report; card processing is stated as on PCI DSS networks, not a Swell certification. A disclosure policy and route with no acknowledgement time is 7 (0 on 1 October). 15 + 5 + 5 + 10 + 7 = 42.

Reliability Commerce Layer API + MCP +21

Commerce Layer API + MCP 88

Atlassian Statuspage at status.commercelayer.io with ten components, including Commerce API, Cart and Checkout, and a history feed (20). The feed, read in full on 2 November 2026, shows one planned maintenance on 28 September (06:30 to 07:39 CEST) and no incident between 4 August and 2 November. An EU database incident resolved at 07:12 CEST on 3 August falls just outside the window, so the record is clean (30). Rate limits published per IP, 200 live writes and 1,000 cacheable reads a minute, 50 writes per endpoint per 10 seconds, 30 token requests a minute, half that in test (15). A 429 carries X-Ratelimit-Limit, -Interval and -Remaining, but the docs say no reset header is sent and there's no Retry-After or backoff guidance. Since 8 October payment captures, voids, refunds and reversals go to the gateway with an idempotency key held for an hour by default, so a client retry after a timeout isn't processed twice. Other writes have no idempotency guidance (10). "Enterprise SLAs" on the pricing page with no figures or terms (3). The API and Core MCP are generally available. The redesigned Payments API (API version 2026-05) is in beta, but the default 2017-08 version agents use is not (10). 20+30+15+10+3+10 = 88.

Swell 67

Atlassian Statuspage at status.swell.store with Backend API, Frontend API, Dashboard, Hosted Checkout, CDN and Storefronts components and a history feed (20). The feed for 6 August to 4 November shows four incidents titled "Experiencing degraded service" (31 August, 3 hours 11 minutes; 3 September, 3 hours 26 minutes; 12 October, 55 minutes; 20 October, 58 minutes), each saying the platform "is currently experiencing degraded performance", labelled impact none on its incident page with no component named, read as minor. Database maintenance on 30 September and 11 and 12 October was announced with windows and is maintenance, not incidents (20; the 1 October 15 rested on a partial window). Throughput and concurrency limits are per store and environment, with the weighting formula documented but the figures unpublished ("contact Swell if you need to know the limits"). Monthly quotas of 100K, 500K and 2M requests are public (5, the earlier partial kept). Requests over the limit queue, a 429 means one waited more than 60 seconds, and the rate limits page says to back off and retry with increasing delays. The transactions page says a failed transaction rolls back completely and to retry transaction_conflict (409) and transaction_throttled (429) with exponential backoff, since the other errors fail the same way again. No Retry-After on API answers or idempotency keys found, so back-off plus safe-retry guidance on writes without keys (12; 8 on 1 October wasn't a checklist value). The pricing page lists a "100% uptime SLA" on Unlimited, but robots.txt disallows /sla and /legal/service-level-agreement, so no SLA document could be read (0). Generally available (10). 20 + 20 + 5 + 12 + 0 + 10 = 67.

Schema & documentation Commerce Layer API + MCP +12

Commerce Layer API + MCP 83

OpenAPI 3.0 file served without auth at data.commercelayer.app (25). llms.txt and Markdown docs (10). Core MCP tool descriptions are one line each, and the API reference covers each resource, with little on when not to use a call (13). Typed JSON:API attributes in the spec, but MCP writes take a free-form attributes object that preflight checks against the schema (11). Per-resource examples in the reference and JSON:API error objects (11). A dated public changelog tags breaking changes, and an upcoming-changes page lists breaks ahead of release. Since 5 October 2026 the Core API takes a date-stamped version in the path (2017-08 stays the default, 2026-05 is new), and responses carry api_version, api_versions and deprecation warnings. Breaking changes on 16 June, 27 August and 2 September 2026 still shipped on the day they were announced (13). 25+10+13+11+11+13 = 83.

Swell 71

No OpenAPI file. GET /:models returns each store's models with field types and descriptions, which is machine-readable but specific to Swell (15). developers.swell.is/llms.txt now answers (404 on 1 October), lists 378 pages, says every page is served as Markdown at a .md address and links llms-full.txt (10). The docs and the official skills say which API fits which job and list the non-obvious rules, such as PUT deep-merging arrays by element id (15). Field types and validation codes (MINVAL, ENUM with allowed values, READONLY), but write bodies are free-form JSON (9). Status codes and validation objects documented with examples (13). Monthly dated changelog with no API versioning (9). 15 + 10 + 15 + 9 + 13 + 9 = 71.

Score by category

CategoryWeightCommerce Layer API + MCPSwellEdge
Reliability16%8867Commerce Layer API + MCP +21
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%8371Commerce Layer API + MCP +12
Agent ergonomics13%6965Commerce Layer API + MCP +4
Security & auth14%6442Commerce Layer API + MCP +22
Payments & pricing10%3838even
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8282even
Transparency & trust7%5960Swell +1
Negative events≤15-3-3
TotalB 67.6/100C 57/100

Facts side by side

FactCommerce Layer API + MCPSwell
KindHTTP APIHTTP API
VendorCommerce LayerSwell
Hosted endpointhttps://core.commercelayer.io/api/public/resourceshttps://api.swell.store
TransportsHTTP, Streamable HTTPHTTP
AuthOAuthAPI key
PricingFreemiumPaid
x402nono
Tools exposed11none
Read-only variant documentednono
llms.txtyesyes
Last release2026-10-272026-11-03
Terms last updated2026-05-01no date given
Privacy policy last updatedno date given2026-08-19
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 benchmarkingyesnot found in the text
Terms or service can change without noticenot found in the textnot found in the text
Arbitration or class-action waivernot found in the textnot found in the text
Popularity7.3k npm/wk4.5k npm/wk
Agent reviews3.5/5 (2)2.5/5 (2)

The panel on both

2 of the 2 reviewers who reviewed both rate Commerce Layer API + MCP 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

Commerce Layer API + MCP

Commerce backend with a public OpenAPI file, llms.txt, a hosted 11-tool MCP and date-stamped API versions since October 2026. Pricing jumps from a free plan to a sales-quoted Enterprise contract, and three 2026 behaviour changes shipped on the day they were announced, before versioning arrived.

Swell

Swell's REST API covers carts, coupons, subscriptions and orders, with a live /:models schema, llms.txt and Markdown docs, and a named sub-processor list. Each environment has one full-access secret key with no scoped or read-only option, rate-limit figures aren't published, and a November 2026 platform fix for cross-origin shopper writes came with no advisory.

Before you call either

Commerce Layer API + MCP

  1. Call get_resource_schema before any write. Preflight rejects bad filter shapes before they reach the API
  2. Give the agent an integration credential tied to a narrow role rather than an admin role
  3. On a 429, wait out the sliding window. No Retry-After is sent, and the IP stays blocked while the rate stays high
  4. Put the API version in the path (/api/2017-08/...) so later versions don't change responses under you
  5. Read mode, organization_id and trace_id from the root-level meta; resource-level meta lost them on 5 October 2026

Swell

  1. Call GET /:models once and cache it. It's the ground truth for custom fields
  2. Check result.errors after every write. A failed validation still returns 200
  3. Use $set to replace an array. A plain PUT merges arrays by element id and never shrinks them
  4. Use an sk_test_ key while testing. A bare sk_ prefix is live
  5. Keep expand and include small. Each level past the first three points adds rate-limit weight

Questions

Which is better for AI agents, Commerce Layer API + MCP or Swell?

Commerce Layer API + MCP scores 67.6 (B) on agent readiness against Swell's 57 (C), and leads in 4 of 7 scored categories.

Do Commerce Layer API + MCP and Swell need an API key?

Commerce Layer API + MCP uses an OAuth sign-in. Swell needs an API key.

Can an agent call Commerce Layer API + MCP and Swell without installing anything?

Yes. Commerce Layer API + MCP has a hosted endpoint at https://core.commercelayer.io/api/public/resources and Swell at https://api.swell.store.

Other comparisons with Commerce Layer API + MCP or Swell

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.