Pennylane

# Fix list: Pennylane

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/pennylane, the October 2026 research run, assessed 9 October 2026. Grade B, 66.3 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 Pennylane: 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. Payments & pricing, 20 out of 100, up to 10 more on the total

Why it scored 20: No x402, MPP or L402 (0). Plan prices are public, 7 to 79 euros a month before VAT on the tab for sole traders and single-shareholder companies, with the API from Essentiel at 24 euros. No per-call price, so plan-only credit (10). A 15-day trial with no commitment and a sandbox inside an account. Whether the trial needs a card, and whether it includes the API, was not established, so half credit (10). A person signs up in a browser, OAuth apps need the Partnerships team, and the MCP flow needs a browser sign-in (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.

## 2. Security & auth, 68 out of 100, up to 5.6 more on the total

Why it scored 68: OAuth 2.0 authorisation code with 24-hour access tokens and rotating refresh tokens for partners, company and firm tokens with chosen scopes, expiry and deletion, and OAuth 2.1 with PKCE on the MCP server. No documented option to send a token in a URL (30). Every resource has a `:readonly` scope, tokens can be read only and the MCP server cannot write. No confirmation step for irreversible calls such as finalising an invoice (16). The API returns invoice and supplier text written by other parties. The MCP docs warn that data goes to the AI vendor and that answers can be wrong, with no prompt-injection guidance (4). The token list shows last use, and changelog endpoints and the `get_changelog` tool list record changes over four weeks. No per-call log for the operator was found (6). ISO 27001 with a 2026 certificate published, an annual third-party penetration test in the customer terms, a security white paper and an incident response plan. No security.txt, bug bounty or disclosure policy was found (12).

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

Why it scored 66: The REST API is the main surface. Responses are sized with `limit` from 1 to 100, default 20, and filters, with no field selection found. The MCP server lists 31 tools, each data type switched on by its own scope (15). Cursor pagination, a `filter` parameter with operators and `sort` on list endpoints (20). Errors return `error`, `message` and `details` naming the field and rule, with a table of which codes to retry (17). No idempotency keys. The docs say creation endpoints do not enforce idempotency except supplier invoice imports, and a duplicate file returns 409. The MCP server is read only, and its annotations were not read (7). Sensible defaults, but no official SDK was found. The reference shows generated samples in six languages (7).

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.

## 4. Reliability, 75 out of 100, up to 5 more on the total

Why it scored 75: Graded on the hosted lines. Atlassian Statuspage at status.pennylane.com with seven components, one of them the API on app.pennylane.com (20). Four incidents between 11 July and 9 October 2026. On 20 July several services were disrupted intermittently from 10.55 until the 19.03 CEST resolution, with the application unavailable from 16.53 to 17.00. The others were mobile uploads for about three hours on 27 August, slowdowns for 50 minutes on 14 September and instability for 37 minutes on 25 September. We scored this between minor incidents and one major outage (15). 25 requests per 5 seconds per token (15). 429 carries `retry-after`, every response carries rate limit headers and the error guide asks for exponential backoff on 5xx, but creation endpoints do not enforce idempotency except supplier invoice imports (10). The customer terms commit to 99.9 per cent for app.pennylane.com with credits, while the API agreement supplies the API as is with no warranty, so half credit (5). v2 is described as the official stable version (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.

## 5. Maintenance & community, 65 out of 100, up to 3.1 more on the total

Why it scored 65: Newest changelog entry 30 September 2026 (30). Five dated entries between 15 July and 30 September 2026 (20). Closed service with a dated changelog, an RSS feed, a public roadmap, support by chat or account manager for customers and Zendesk for partners, and a community forum we did not read (10). No official SDK was found, and a search of the official MCP registry for pennylane returned no server (0). No package to assess. The docs are current, with pages updated on 23 September and 5 October 2026 (5).

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.

## 6. Schema & documentation, 85 out of 100, up to 2.4 more on the total

Why it scored 85: Each reference page carries an OpenAPI 3.0.1 definition for its operation, readable through the Markdown twin. One downloadable file for the whole API was not located, and the MCP tool schemas need a signed-in session, so we did not read them (23). llms.txt indexes the guides and 179 reference pages, and every page has a Markdown twin (10). Descriptions state purpose, the scopes required and side effects, such as a finalised invoice no longer being editable, and guides compare importing with creating an invoice (14). Typed parameters with minimum and maximum values and enums on the page read, though `filter` is a JSON array passed as a string (11). Examples in the definitions and an error guide covering nine status codes with fixes (12). Versioned as v2 with a dated changelog and RSS feed (15). Based on one reference page read in full.

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.

## 7. Transparency & trust, 76 out of 100, up to 2.1 more on the total

Made of editorial 71, provenance 80.

Why it scored 76: Closed service. The API agreement of January 2024 and the customer terms of 16 December 2025 are public and name the entity and French law. The API agreement lets Pennylane change or revoke access without notice and takes an exclusive licence over integrations, and its registered office differs from the one in the customer terms (13). Privacy policy of 30 January 2026, a data processing agreement annexed to the terms, deletion 30 days after termination, EU hosting and a commitment not to train AI models on customer data. The privacy policy allows providers outside the EU while the sub-processor list shows none (24). The 2026 breaking changes had dated phases over 24 weeks and v1 had an end date, but no standing deprecation policy was found and the agreement reserves changes without notice (14). Sub-processor list with locations, dated 16 December 2025, with update notices 30 days ahead (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):

- Terms of service: read, states 6 of the 7 things a reader expects, and has 2 clauses that cost points (5.1 of 10)
- Privacy policy: read, states 1 of the 8 things a reader expects (4.8 of 10)
- security.txt: not found (0 of 10)

## 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: no request was sent to the API or the MCP server. The robots.txt on app.pennylane.com disallows every path except /auth/login, so the 401 body, live headers and MCP tool schemas and annotations are unread
- unchecked: whether the 15-day trial needs a card and whether it includes API access. The signup flow is inside the app
- unchecked: prices for company types other than the default pricing tab, which the page draws by script
- unchecked: a single downloadable OpenAPI file, the Firm API reference at firm-pennylane.readme.io and the webhook guides
- unchecked: the security white paper, the incident response plan, the ISO 27001 certificate PDF and the terms for accounting firms, none of which we opened
- unchecked: security incidents in the last 12 months. Web search was not available and nothing was found on the vendor's pages
- unchecked: the community forum and the help centre
- The MCP product page says a connection cannot be cut from the Pennylane account, while the MCP docs say it can be revoked under connected applications
- The introduction guide still shows a first request to `https://api.pennylane.com/v1/me`, although v1 was due for retirement at the end of 2025
- Whether the 99.9 per cent commitment for app.pennylane.com applies to API calls, given that the API agreement disclaims any warranty
- No official SDK and no public repository were found in the docs. GitHub was not searched

## Weaknesses

- The API agreement allows changes, suspension or revocation with or without notice and supplies the API as is
- Creation endpoints do not enforce idempotency, except supplier invoice imports
- API and MCP access for companies starts at the Essentiel plan, and OAuth apps need approval from the Partnerships team
- No official SDK was found, and the MCP server is not in the official MCP registry
- No security.txt, bug bounty or published disclosure policy was found
- The API agreement grants Pennylane an exclusive licence over API integrations and their source code on request

## 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.

- Call `GET https://app.pennylane.com/api/external/v2/me` first to confirm the token, user and company context
- Stay under 25 requests per 5 seconds per token and wait for the `retry-after` value on a 429
- Do not retry a create after a timeout without checking first. Only supplier invoice imports are deduplicated
- Send `filter` and `sort` again with every `cursor`. The cursor stores position only
- Send amounts as strings such as "120.00", and ask for `:readonly` scopes when the task only reads
- Use the MCP server for reads only. It cannot create, edit or delete, and firm connections need `company_id`

## 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.