Brex

# Fix list: Brex

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/brex, the October 2026 research run, assessed 8 October 2026. Grade C, 60.7 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 Brex: 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, 25 out of 100, up to 9.4 more on the total

Why it scored 25: Read with the hosted rubric. No x402, MPP or L402 (0). Plan prices are public, $0 per user per month for Essentials, which lists API access, and $12 for Premium, with Enterprise on request. The API has no per-call price (10). The $0 plan needs no payment card, but it needs an approved Brex business account, and there is no customer sandbox, so partial credit (15 of 20). A person creates each token in the dashboard or completes an OAuth 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. Reliability, 60 out of 100, up to 8 more on the total

Why it scored 60: Read with the hosted lines and scored on the REST Developer API, the surface an outside agent calls with a token. status.brex.com on Atlassian Statuspage has 12 components, among them Partner API and Authentication, with incident history (20). Thirteen entries between 12 July and 8 October 2026, seven marked critical. Errors affecting API requests ran from 16:01 to 20:34 UTC on 4 August, with a fix in place at 17:50. An entry on 21 September reports developer API tokens invalidated unexpectedly, with affected customers told to create new ones. Maintenance took the dashboard and external APIs down for about nine minutes on 12 July. One API outage of over an hour plus the token invalidation, read as between one major and several (5). Limits with numbers, 1,000 requests in 60 seconds per client and account, plus daily caps on transfers, wires and cards (15). The docs recommend exponential backoff with jitter on 429, and every POST and PUT accepts an `Idempotency-Key`, required on transfers and cards. A Retry-After header isn't documented (13 of 15). No SLA found in the Platform Agreement or the Access Agreement (0). Team, Budgets, Payments, Transactions, Travel and Onboarding are at version 1.0, while Fields is beta, Accounting alpha, Expenses and Webhooks are numbered 0.1 and the MCP server is beta (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.

## 3. Agent ergonomics, 70 out of 100, up to 4.9 more on the total

Why it scored 70: `limit` sizes a page, related objects arrive only through `expand[]`, and custom fields only with `load_custom_fields`. No field selection (18 of 25). Cursor pagination with `next_cursor` on every list endpoint and 15 filters on List expenses. The FAQ says card transactions can't be filtered to a time range beyond a start date, and pending transactions aren't returned (17 of 20). Errors carry `type`, `message` and an optional `code`, with an `X-Brex-Trace-Id` header and a short table of fixes. No full list of codes (12 of 20). `Idempotency-Key` on all POST and PUT requests, required where a duplicate would move money or issue a card. MCP tool annotations couldn't be read without an account (17 of 20). List calls need no parameters. No official SDK, three community libraries the docs say Brex doesn't support, and a Postman guide (6 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.

## 4. Security & auth, 72 out of 100, up to 4.9 more on the total

Why it scored 72: User tokens with scopes chosen at creation, shown once, revocable in the dashboard and expired after 90 days unused. Partners use the OAuth 2.0 authorisation code grant with scopes, optional PKCE and one-hour access tokens. All credentials travel in the `Authorization` header (30). Most resources have a read-only scope, card numbers need the separate `cards.pan` scope, and MCP users act with their own Brex permissions, with approvals and card management left out of the MCP. No approval step was found for transfers or card creation made with a token that holds the write scope (13 of 20). Expense memos, merchant and vendor names and receipts are untrusted text. The MCP page advises requiring human approval before actions, with no guidance on injected content (5 of 15). The MCP page points to API logs in the dashboard and a list of connected integrations that can be revoked, and responses carry a trace ID. Log contents and retention aren't documented publicly (9 of 15). SOC 1 Type II, SOC 2 Type II and PCI DSS per the trust page, and a disclosure policy with a form run with Bugcrowd. No bounty is stated and there is no security.txt (15 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.

## 5. Schema & documentation, 80 out of 100, up to 3.3 more on the total

Why it scored 80: Ten public OpenAPI 3 specs, 115 operations, each downloadable from the reference page (25). llms.txt and a Markdown twin of every docs and reference page (10). Operation descriptions state the purpose and which role may call them, and some carry constraints, such as the ten physical cards a user may hold. Many filter parameters have no description, and none says when not to use an endpoint (12 of 20). Enums on status and type fields, date-time formats and required bodies. `expand[]` takes free strings (12 of 15). An examples page per API and an error guide with the body shape and a table of common errors. The Team API spec documents only 200 responses for its 35 operations, and no spec lists 429 (9 of 15). Path versions v1 and v2, launch-stage suffixes on spec versions and a public changelog, dated to the month only (12 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, 66 out of 100, up to 3 more on the total

Why it scored 66: The changelog's latest entry is October 2026 (30). Entries for July, August, September and October 2026, with nine items in September alone (20). Closed service with a dated changelog, a developer support address and a Slack community. Response times couldn't be observed (10 of 15). No official SDK, and no entry for the Brex MCP server in the official MCP registry (3 of 15). No packages to assess. The OpenAPI bundles match the changelog's October additions (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, 67 out of 100, up to 2.9 more on the total

Made of editorial 47, provenance 86.

Why it scored 67: Closed service with terms published at stable URLs and as Markdown, the Platform Agreement and the Brex Access Agreement for API use (15 of 30). A privacy policy and a DPA with breach notice within 72 hours and deletion within 30 days of a written request after termination. Retention periods aren't stated, and the Access Agreement lets Brex monitor API use for any business purpose (17 of 30). A launch-stages page promises 15 days' notice of breaking changes in alpha and 60 in beta, and new versions with deprecation timelines once generally available. Deprecated endpoints are marked in the specs. Removal dates aren't published, and the versioning page promises only sufficient lead time (11 of 20). The DPA points to a sub-processor list on trust-portal.brex.com, which returned 403 to our reader, and says data may be transferred globally. The trust page names AWS storage services (4 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):

- Terms of service: read, states 5 of the 7 things a reader expects, and has 1 clause that costs points (6.3 of 10)
- Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10)
- security.txt: not found (0 of 10)

## Deductions

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

- 2026-02 and 2026-09. The changelog records the List expenses maximum `limit` cut from 1,000 to 100 in February 2026, and `GET /v2/users/{id}/limit` removed from the Team API in September 2026 without a deprecated label in the entry. The Team API is at version 1.0, and the launch-stages page says breaking changes to generally available APIs come as new versions with deprecation timelines. Both changes are documented in the month they shipped, and notice by email couldn't be checked, so the smallest deduction applies (https://developer.brex.com/changelog).

## 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 sub-processor list and security documents on trust-portal.brex.com, which returned 403 to our reader
- unchecked: MCP tool input schemas and the readOnlyHint and destructiveHint annotations, which need a signed-in Brex account
- unchecked: whether customers were told by email before the February 2026 limit cut and the September 2026 endpoint removal
- unchecked: whether 429 responses carry a Retry-After header
- unchecked: what the dashboard's API logs record and how long they are kept
- The changelog is dated to the month, so `lastRelease` is recorded as 2026-10-01 for an entry labelled October 2026
- The 4 August 2026 API incident and the 21 September token incident have no published cause or count of affected customers
- Bills appear only as read-only MCP tools (`list_bills`, `get_bill_by_id`). The REST API has vendors and transfers but no bill endpoints in the public specs
- No purchase request or procurement endpoint was found in the public specs
- The last-updated dates of the privacy policy and the Access Agreement weren't visible in the pages we read

## Weaknesses

- The Expenses API update endpoint accepts only `memo`, so an outside agent can't set a category or custom field on an expense through it
- No customer sandbox. The docs say staging isn't a sandbox and won't accept customer tokens
- No official SDK. The docs list three community libraries that Brex doesn't support
- status.brex.com logs API request errors from 16:01 to 20:34 UTC on 4 August 2026 and invalidated developer tokens on 21 September 2026
- The MCP server is beta with 43 tools, and approvals and card management aren't available through it

## 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.

- Ask an account admin or card admin for a user token with only the scopes the task needs, and prefer the `.readonly` variants. A token unused for 90 days expires.
- Send a stored `Idempotency-Key` on every POST and PUT. Create transfer and Create card reject requests without one.
- Only settled transactions are returned. Poll card and cash transactions with `posted_at_start` and a lookback of at least one day.
- Keep under 1,000 requests in 60 seconds per client and account, and back off exponentially with jitter on 429.
- Send only ASCII in free-text fields, and quote the `X-Brex-Trace-Id` response header when reporting an error.

## 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.