Numeral

# Fix list: Numeral

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/numeral, the October 2026 research run, assessed 8 October 2026. Grade D, 49.3 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 Numeral: 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, 25 out of 100, up to 15 more on the total

Why it scored 25: Graded on the REST API with the hosted lines, with the MCP server noted where it differs. No status page was found on numeral.com, the docs or the help centre, and `status.numeral.com` and `status.numeralhq.com` did not answer (0). With no page there is no readable incident history (5). The docs name 429 responses for the API and the MCP server but give no number. The product page says rate limits can be removed on request, and the MCP docs describe a sliding window per account (0). The best practices guide advises exponential backoff on 429 and one retry on 500 or 503. No Retry-After header is documented, and there are no idempotency keys, though the docs give a duplicate-recovery pattern on `reference_order_id` (10). No uptime SLA found. The filing guarantee covers late returns, not availability (0). The API has four dated versions and no beta label, and the MCP docs carry none (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. Payments & pricing, 10 out of 100, up to 11.3 more on the total

Why it scored 10: No x402, MPP or L402 (0). The pricing page gives $75 per filing and $150 per registration on Professional. The tax engine and API sit under Enterprise at a custom price, so the surface an agent calls has no public price (5). The Monitoring plan is free and test keys cost nothing, but the get-started page is a qualification form that ends in a demo booking, and whether a free account can create API keys isn't stated (5). Keys are created by a person in the dashboard (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.

## 3. Security & auth, 47 out of 100, up to 9.3 more on the total

Why it scored 47: Secret API keys as Bearer tokens, `sk_test_` or `sk_prod_`, created and revoked in the dashboard, with revocation immediate per the MCP docs. No scopes, expiry or OAuth found (20). Test keys reach sandbox data only. The MCP server is read-only by default and its four write tools stay unregistered until Numeral enables them for an account. The REST API has no read-only key and no confirmation step (10). Responses carry merchant, customer and product names written by others, with no injection guidance found. The docs do state that no tool accepts an account identifier (3). The dashboard's Logs tab shows each request and response (8). A trust centre at security.numeral.com runs on Oneleet and is drawn by script, so we couldn't read it. No security.txt, bug bounty or disclosure policy found. The DPA lists security measures (6).

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.

## 4. Transparency & trust, 45 out of 100, up to 4.8 more on the total

Made of editorial 36, provenance 53.

Why it scored 45: Closed service under terms last updated 10 June 2025 that name Numeral Two, Inc. and California law. The SDKs are Apache-2.0 (15). The terms let Numeral use customer content to improve the services and other Numeral offerings and say it has no obligation to store it. The DPA of 2 January 2024 limits use to the customer's purposes and promises deletion or return at the end of service. The privacy policy of 11 May 2026 excludes platform data. No retention period or statement on AI training was found (12). No deprecation policy found. Four API versions stay documented with no end dates (4). The DPA says the subprocessor list is available by email and promises 30 days' notice of changes. Hosting location for platform data isn't stated (5).

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):

- Endpoint on the vendor's domain: api.numeralhq.com is not on numeral.com (0 of 15)
- Terms of service: read, states 5 of the 7 things a reader expects (8.3 of 10)
- Privacy policy: not found (0 of 10)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 5. Agent ergonomics, 71 out of 100, up to 4.7 more on the total

Why it scored 71: List endpoints take `cursor`, and merchants and certificates take `limit` up to 100. No field selection. The MCP server documents 14 tools, list tools return summaries, and `get_sales_summary` aggregates in one call (16). Cursor paging throughout. Certificates filter by status, customer, jurisdiction and dates, while product and customer lists take only a cursor (16). Errors are flat JSON with `code`, `type` and `message` and a documented list of types. MCP tool errors return a stable `error_code` with a corrective message (19). No idempotency keys, and the docs say an `Idempotency-Key` header is silently ignored. Commits are made safe by handling `REFERENCE_ORDER_ID_ALREADY_EXISTS`, and cancelling a certificate request is idempotent. MCP annotations unread (11). A calculation needs an address or IP, line items, `origin_address` and `automatic_tax`. SDKs exist for Go, Python and Node, all below 1.0 and two marked alpha (9).

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.

## 6. Maintenance & community, 60 out of 100, up to 3.5 more on the total

Why it scored 60: Closed service. The 2026-03-01 spec was last modified on 15 September 2026 and the docs were rebuilt on 8 October 2026 per their Last-Modified headers (30). The changelog has no dated entry since 1 March 2026. Dated activity in the last 90 days is the spec change, the Go SDK v0.1.0 on 27 August and the npm package on 18 August 2026 (10). A public changelog by version, a help centre and support by email. No public issue tracker in use (7). Go, Python and Node SDKs generated by Stainless, none at 1.0, and the Python package dates from 18 December 2025. Numeral's MCP server isn't in the official registry, where the only entry for the name is a third party's (8). The Go SDK has CI and a changelog (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.

## 7. Schema & documentation, 87 out of 100, up to 2.1 more on the total

Why it scored 87: OpenAPI 3.0.1 at `docs.numeral.com/api-reference/v2026-03-01/openapi.json` with 32 operations and 74 schemas. An older unversioned spec with 15 operations is the one llms.txt links (25). `docs.numeral.com/llms.txt` indexes every page and each page is served as Markdown (10). Endpoint pages state purpose, the MCP overview says when to use MCP and when REST, and the AI help page publishes skills that list integration mistakes. We couldn't read the live MCP tool definitions without a key (16). 57 enums and 35 required lists, with 6 numeric bounds and 2 length limits (11). 235 examples in the spec and an error reference listing each `type`. The 2026-03-01 spec declares no 429 response (13). Versions are dates sent in `X-API-Version`, and the changelog covers each version with migration notes. Additions made after 1 March 2026 carry no date (12).

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.

## 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 trust centre at security.numeral.com, a Oneleet page drawn by script. Certifications, penetration tests and any disclosure policy listed there were not read. Numeral's llms.txt says it holds SOC 2 information
- unchecked: the live MCP tool definitions, input schemas and annotations, which need an API key. The count of 14 comes from the tool reference page
- unchecked: the dashboard, including whether a free Monitoring account can create test API keys without a demo
- Whether a status page exists. None is linked from the site, docs or help centre, and two guessed hostnames did not answer
- The numeric rate limits for the API and the MCP server. Neither is published
- The product page names Java, Python and Ruby SDKs. We found Go, Python and Node packages and no Java or Ruby one
- The price of API access. The pricing page lists the tax engine and API under Enterprise as Custom
- The privacy policy excludes platform data, so `provenance.privacy` points at the DPA, the published document that governs personal data in the product
- lastRelease is the Last-Modified date of the 2026-03-01 OpenAPI spec, since the changelog carries no dated entry after 1 March 2026

## Weaknesses

- No status page, uptime SLA or numeric rate limit was found. The product page says rate limits can be removed on request
- The API ignores an `Idempotency-Key` header. A retried commit is safe only if the client handles `REFERENCE_ORDER_ID_ALREADY_EXISTS`
- Secret keys carry no scopes, and the MCP server has no OAuth flow. One key reads all of the account's tax data
- API access is listed under the Enterprise plan at a custom price, and the product page says to book a demo first
- SDKs are early. Go is v0.1.0, `numeral-tax` on PyPI is 0.1.0a1 from December 2025 and on npm is 0.1.0-alpha.2

## 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.

- Send `X-API-Version: 2026-03-01` on every call to `https://api.numeralhq.com`. Merchant, platform and certificate endpoints reject requests without it
- Commit with a stable `reference_order_id`. On `REFERENCE_ORDER_ID_ALREADY_EXISTS` look up the existing transaction, because `Idempotency-Key` is ignored
- Commit before the calculation's `expires_at`. If it has expired, calculate again and commit the new `calculation_id` with the original `reference_order_id`
- Send `amount` as the line total in minor units to `/tax/calculations` and as the per-unit price to `/tax/platform/calculations`
- Start with an `sk_test_` key. Certificate endpoints and MCP certificate tools are live only and answer `TESTMODE_NOT_SUPPORTED` or `live_mode_required`

## 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.