PayPal MCP server

# Fix list: PayPal MCP server

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/paypal-mcp-server, the October 2026 research run, assessed 8 October 2026. Grade C, 56.3 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on PayPal MCP server: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 52 out of 100, up to 9.6 more on the total

Why it scored 52: Graded on the hosted lines, because the remote server at mcp.paypal.com is the surface the quickstart leads to. PayPal's status page at paypal-status.com covers its production products (20). The page is drawn by script, so we read its Atom feed, which on 8 October went back only to 15 August 2026. It shows two resolved degraded-performance events on PayPal payments, on 15 August (refunds, payments, authorisations, orders) and 22 September (Checkout), with no durations given, and no entry naming the MCP server (15 of 30). PayPal states that it doesn't publish rate limits (0). The rate-limiting guide tells callers to read `Retry-After` on 429 and lengthen delays, and the REST APIs take `PayPal-Request-Id`, but the MCP tools have no per-call idempotency argument (10 of 15). No SLA found, and the Developer Agreement gives no uptime guarantee for the sandbox (0). No beta label on the server, but the remote commerce tools are limited to gift cards behind a feature flag and the documented `/http` path returned 404 (7 of 10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Security & auth, 48 out of 100, up to 9.1 more on the total

Why it scored 48: The remote server publishes OAuth metadata with authorisation code, PKCE S256, refresh tokens, revocation and dynamic client registration, and no scope list. The local server takes a REST access token that carries the whole app's permissions for about nine hours (22 of 30). `--tools=` narrows the local tool set and the sandbox is the default, but no read-only mode and no confirmation step for refunds, captures or accepting a dispute claim was found (8 of 20). Tools return invoice notes, dispute text and product descriptions written by other people, and the only guidance found is to write clear system prompts and keep human oversight (4 of 15). No audit log or per-call view for MCP use was found in the reviewed pages, and we did not look inside the dashboard (0). security.txt names a HackerOne bug bounty and has no Expires field. We did not read PayPal's certification pages, and neither repository has a security policy file (14 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 3. Payments & pricing, 40 out of 100, up to 7.5 more on the total

Why it scored 40: Payment platforms take the highest step that applies on the 40-point protocol line. No x402, MPP or L402 was found in the MCP docs, the developer llms.txt indexes or either repository, and PayPal's Agentic Commerce Protocol and Universal Commerce Protocol pages concern checkout, not paying for calls (0). US fees are public without a login, 3.49% plus $0.49 for PayPal Checkout and 2.99% plus $0.49 for standard cards (20). The sandbox is free with a developer account, with no card and no monthly fee on most products (20). A person has to create the PayPal account and app, and log in for the remote server (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 4. Agent ergonomics, 66 out of 100, up to 5.5 more on the total

Why it scored 66: `--tools=all` loads 28 tools (15), and `--tools=` takes a comma-separated subset, so a client can load only what a task needs (8 back, 23 of 25). List tools take `page`, `page_size`, `total_required` and status or date filters (16 of 20). Errors come back as `ok`, `status`, `code` and a message cut to 200 characters, which is usable but undocumented (10 of 20). Tools are registered with a name, description and schema only, so there is no `readOnlyHint` or `destructiveHint`, and `PayPal-Request-Id` is set only from toolkit context (4 of 20). The sandbox is the default, most tools need few fields, and the toolkit ships for TypeScript and Python (13 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 5. Schema & documentation, 70 out of 100, up to 4.9 more on the total

Why it scored 70: Each tool is registered with a zod shape that the MCP SDK turns into JSON Schema (25). developer.paypal.com/llms.txt indexes the docs by section, including the MCP and toolkit pages (10). Tool descriptions state the purpose and some say when to choose another tool, such as cancelling instead of deleting an active recurring series, while many are one line (13 of 20). Inputs carry 32 enums plus length and range limits, and list tools bound `page_size` at 100 (12 of 15). The quickstart gives one example prompt and no error reference for the tools was found (6 of 15). The toolkit changelog stops at 1.3.5 from April 2025 although 1.11.0 is current, the MCP repository has none, and in August 2026 the version went to 1.12.0 and back to 1.9.0 after failed builds (4 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 6. Maintenance & community, 64 out of 100, up to 3.2 more on the total

Why it scored 64: `@paypal/mcp` 1.8.1 was published on 28 October 2025, but it loads `@paypal/agent-toolkit` at `latest`, whose 1.11.0 came out on 1 September 2026, 37 days before this check (20 of 30). The toolkit had three npm releases in the last 90 days, 1.9.0 on 10 August, 1.10.0 on 20 August and 1.11.0 on 1 September (20). Pull requests were merged in August 2026. GitHub's API refused us, so the issue queue went unread (8 of 25). The server is in the official registry as `io.github.paypal/paypal-mcp-server`, a GitHub-verified namespace, with a 1.8.1 entry that lists the stdio package and not the remote server (13 of 15). Neither repository has test files, the toolkit's test script only prints an error, CI publishes without testing, and the server depends on the toolkit at `latest` (3 of 10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 7. Transparency & trust, 78 out of 100, up to 1.9 more on the total

Made of editorial 58, provenance 98.

Why it scored 78: The local server and toolkit are Apache-2.0 on GitHub, though the toolkit's npm metadata says ISC, and the hosted server is closed under the PayPal Developer Agreement (20 of 30). The Privacy Statement of 28 September 2026 gives retention as the relationship plus 10 years, says personal information may be used to train PayPal's AI models, and describes a third-party list updated quarterly with 30 days to object. No statement specific to MCP traffic was found (20 of 30). The Developer Agreement lets PayPal change or discontinue APIs on notice at any time, with 30 days' notice for substantial changes to the agreement, and no deprecation notice for the MCP server or toolkit was found (8 of 20). The toolkit changelog discloses User-Agent telemetry with no opt-out found, and the Privacy Statement's third-party list was not read for data locations (10 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Privacy policy: read, states 8 of the 8 things a reader expects, and has 1 clause that costs points (8 of 10)

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 8 October 2026. The MCP quickstart (last updated 30 June 2026) gives `https://mcp.paypal.com/http` and `https://mcp.sandbox.paypal.com/http` as the streamable HTTP endpoints. Both returned 404 to an initialise request on 8 October 2026, while `/mcp` on each host returned the OAuth challenge. A documented endpoint that doesn't answer, minus 2 (https://developer.paypal.com/ai-tools/mcp-server).

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- unchecked: the tool list, annotations and scopes the hosted server at mcp.paypal.com returns after login, because we did not authenticate
- unchecked: GitHub stars, open issues and security advisories for paypal/agent-toolkit and paypal/paypal-mcp-server, because the GitHub API refused our requests
- unchecked: status history before 15 August 2026 and incident durations, because the status page is drawn by script and its feed holds 13 entries
- unchecked: PayPal's certification pages (PCI DSS, SOC reports) and the third-party list linked from the Privacy Statement
- unchecked: API call logs in the PayPal Developer Dashboard, which need a login
- The lead gave a second remote transport without naming its path. The docs name `/http`, which returned 404, and `/mcp` is what answers
- The registry entry describes `PAYPAL_ENVIRONMENT` as 'SANDBOX' or 'LIVE', but the code treats every value other than `production` as sandbox
- Whether the remote commerce tools (`search_product`, `create_cart`, `checkout_cart`) are open to any account or only by feature flag

## Weaknesses

- The quickstart gives `/http` for streamable HTTP. On 8 October 2026 it returned 404 on both hosts, and `/mcp` returned the OAuth challenge
- Tools are registered without `readOnlyHint` or `destructiveHint`, and no approval step for `create_refund`, `pay_order` or `accept_dispute_claim` was found
- `@paypal/mcp` 1.8.1 dates from 28 October 2025 and selects 28 tool keys, while the toolkit it loads defines 48 tools
- PayPal states that it doesn't publish rate limits, and no SLA was found. The Developer Agreement gives no uptime guarantee for the sandbox
- Neither repository has tests in CI, and the toolkit changelog stops at 1.3.5 from April 2025 although 1.11.0 is current

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Connect a remote client to `https://mcp.paypal.com/mcp` or `/sse`. The documented `/http` path returned 404 on 8 October 2026
- Set `PAYPAL_ENVIRONMENT=PRODUCTION` for live calls. Any other value, including the registry's `LIVE`, runs against the sandbox
- Refresh `PAYPAL_ACCESS_TOKEN` before it expires. The README gives `expires_in` 32400 seconds, and the local server takes no client ID or secret
- Pass `--tools=` with only the keys the task needs, such as `invoices.list,orders.get`. `--tools=all` loads 28 keys including refunds and captures
- Ask the user before `create_refund`, `pay_order`, `cancel_subscription` or `accept_dispute_claim`. The server applies them without a confirmation step

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.