X MCP

# Fix list: X MCP

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/x-mcp, the October 2026 research run, assessed 8 October 2026. Grade C, 55.4 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 X MCP: 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). Per-resource and per-request prices are published without a login, such as $0.005 a post read, $0.010 a user read and $0.005 a bookmark write (20). No free tier. Credits are bought in advance, and the one-time $20 credit needs a saved payment card (0). A person signs in to the developer portal, creates an app and completes a browser login (0). 20.

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, 48 out of 100, up to 9.1 more on the total

Why it scored 48: OAuth 2.0 authorisation code with PKCE (S256), scopes and refresh tokens, and the resource metadata accepts bearer tokens in the header only. Less 6 because xurl 1.3.4 requests a fixed set of 24 scopes including tweet.write and dm.write with no flag to narrow them, and the bridge config holds the client secret in plain text (24 of 30). The app-only Bearer route is read-only per the docs. No confirmation step is documented for `article_publish` or for removing bookmarks, so approval rests with the MCP client (10 of 20). Posts, profiles and news text come from anyone on X, and the security section has no guidance on prompt injection (0 of 15). The Developer Console shows usage and spend live and /2/usage/tweets returns daily counts. No per-call log was found (6 of 15). x.com's security.txt points to HackerOne but expired on 1 January 2024, and we found no SOC 2 or ISO 27001 statement in what we read (8 of 20). 48.

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, 45 out of 100, up to 8.9 more on the total

Why it scored 45: The number of tools isn't published. One docs page lists six groups and another says XMCP exposes 200+ endpoints as tools, so we scored context cost as not established (5 of 25). The API has `max_results`, `pagination_token`, field selection and expansions. Whether each MCP tool passes them through is unread (12 of 20). Errors are problem+json with a type URI, title and detail, and the bridge turns transport and token failures into JSON-RPC errors keyed to the request (15 of 20). No idempotency keys, and no readOnlyHint or destructiveHint documented. The bridge refreshes the token and retries after a 401 (4 of 20). One npx line starts the bridge and the URL has a default, but each user first creates an app, registers a redirect URI and sets a 300-second startup timeout. Official XDKs exist for Python and TypeScript (9 of 15). 45.

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. Schema & documentation, 65 out of 100, up to 5.7 more on the total

Why it scored 65: A public OpenAPI spec covers the API behind it (version 2.169, 175 paths, 218 operations). The MCP tool list and input schemas aren't published, and tools/list answers 401 without a token, so we didn't read them (18 of 25). llms.txt, llms-full.txt, skill.md and Markdown for every page (10). The MCP page describes six groups of tools and names one tool, `article_publish`, with no per-tool guidance on when to use which (6 of 20). The API parameters are typed in the spec. Whether the MCP tools carry the same enums and constraints is unread (8 of 15). Client configs for five clients, a troubleshooting table of eight symptoms and a documented problem+json error format with eleven error types (12 of 15). The API is versioned in the path and has a dated changelog with an RSS feed, latest entry 2 October 2026, but the changelog has no entry for the hosted MCP server and the server publishes no version history (11 of 15). 65.

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.

## 5. Reliability, 72 out of 100, up to 5.6 more on the total

Why it scored 72: Read with the hosted lines, since the server runs at api.x.com and the bridge only relays. A status page exists at developer.x.com/status but needs JavaScript and showed our reader nothing. The dated incident history at docs.x.com/incidents.md lists incidents by title, with no component view and no MCP entry (15 of 20). That history has no incident dated after 1 July 2026, so the 90 days to 8 October are clean, apart from a 'Missing entities.annotations' entry marked ongoing since 24 March 2026 (25 of 30). API rate limits are published per endpoint, such as 300 per 15 minutes and 1 a second for full-archive search and 50 per 15 minutes for bookmark writes. The MCP page gives no numbers of its own and says only that writes are stricter than reads (12 of 15). 429s come with `x-rate-limit-reset` and the docs advise waiting for it and backing off. No Retry-After and no idempotency keys for writes were found (10 of 15). No SLA found for Pay-per-use (0). The MCP page carries no beta or preview label (10). 72.

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.

## 6. Maintenance & community, 70 out of 100, up to 2.6 more on the total

Why it scored 70: The xurl bridge reached 1.3.4 on 29 September 2026 and the platform changelog's latest entry is 2 October 2026 (30). Six xurl releases since 16 July 2026, and ten platform changelog entries since 21 July (20). The xurl changelog credits fixes to numbered issues, and the repository showed 7 open issues and 8 open pull requests on 8 October. We couldn't read the threads or the developer forum, which answered 403 (12 of 25). No X entry in the official MCP registry under searches for xmcp, xdevplatform and com.x (0). Go 1.24 with build, test, gofmt and release-build jobs in CI (8 of 10). 70.

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, 70 out of 100, up to 2.6 more on the total

Made of editorial 50, provenance 89.

Why it scored 70: The hosted server is closed, under a Developer Agreement last updated 27 April 2026, and the bridge is MIT (18 of 30). The privacy policy, effective 15 January 2026, describes retention by data type and names the controllers. The terms of service let X use content to train its AI models. No DPA or subprocessor list for developers was found in what we read (14 of 30). The versioning page limits breaking changes to major versions, at most yearly, with migration guides. Dated notices exist but can be short, such as five days for the Account Activity replay endpoint in March 2026 and four days for the removal of follows, likes and quote posts from self-serve tiers in April 2026 (14 of 20). Controllers for the US and Europe are named. No subprocessor or data location list found (4 of 20). 50.

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 1 clause that costs points (7.1 of 10)
- Privacy policy: published, but our reader couldn't read it (7 of 10)
- security.txt: published but past its Expires date (5 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: the MCP tool list, input schemas, descriptions and annotations. tools/list needs a token from a Pay-per-use app, which we don't have
- unchecked: the live status page at developer.x.com/status, which needs JavaScript. Incident history was read from docs.x.com/incidents.md
- unchecked: issue and pull request threads on xdevplatform/xurl (the GitHub API was rate-limited) and devcommunity.x.com, which answered 403
- unchecked: the HackerOne programme page, which is a JavaScript app. The only pointer read is the expired security.txt
- How many tools the server exposes. The MCP page lists six groups and the Agent Resources page says 200+ endpoints
- Whether MCP calls are billed at the published API rates and whether failed calls are charged. The MCP page requires a Pay-per-use app but doesn't say
- When the hosted server launched. The docs give no date, and the first dated evidence is `xurl mcp` on 25 June 2026
- The docs say the endpoint has no MCP OAuth discovery, yet it serves protected resource metadata. We didn't test whether a client can sign in directly with a pre-registered client ID
- Whether the developer.write and developer.billing.write scopes listed for the MCP resource back tools that create apps or buy credits
- The category has no capability for searching or reading a network. We used social.analytics as the nearest and left out social.post, since the only publishing tool documented is for Articles
- No deduction was made for the short-notice API changes of February and April 2026, which predate the hosted server and concern posting endpoints it doesn't document

## Weaknesses

- No dynamic client registration, so each user creates an X developer app on the Pay-per-use package in Production
- The tool list and tool schemas aren't published, and the endpoint answers 401 without a token
- The documented tools don't create posts, replies, likes or follows. Writes cover bookmarks and Articles
- xurl 1.3.4 asks for a fixed set of 24 scopes, including tweet.write and dm.write, with no flag to narrow them
- No guidance on prompt injection, though every post returned is text written by strangers
- Not in the official MCP registry, and the platform changelog has no entry for the hosted server

## 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.

- Use the app-only Bearer header for read-only work. Bookmarks, Articles and any tool working in the user's name need the xurl bridge
- Set the client's startup timeout to 300 seconds or more, because the bridge holds the handshake until the first browser login ends
- On a headless host run `xurl auth oauth2 --headless` first, with CLIENT_ID and CLIENT_SECRET exported in that shell
- On `client-not-enrolled`, move the app to Pay-per-use and Production in the developer portal
- On 429, wait until the Unix time in `x-rate-limit-reset`. No Retry-After header is documented

## 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.