Ref

# Fix list: Ref

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/ref, the October 2026 research run, assessed 8 October 2026. Grade C, 59.4 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 Ref: 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, 58 out of 100, up to 8.4 more on the total

Why it scored 58: Scored as a hosted MCP server. ref.tools/status is Ref's own page with eight components and 30-day bars, but the components and incidents are written into the site's script by hand, with no monitoring feed behind them (15 of 20). It lists four incidents, the latest from 28 to 29 November 2025, and none in the last 90 days. A hand-kept list can't be verified, so 25 of 30. No rate limit is documented. Credit allowances are published, and a response on 8 October 2026 carried `ratelimit-policy: "500-in-1min"` (5). No 429 or retry guidance was found in the reviewed documentation, though responses carry `ratelimit` headers (3). No SLA found (0). The hosted endpoint is generally available (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, 35 out of 100, up to 8.1 more on the total

Why it scored 35: No x402, MPP or L402 (0). One credit per tool call and $10 per 1,000 extra credits are published without a login (20). The free plan is 200 one-time credits that never expire, not a recurring allowance, and the docs don't say whether sign-up needs a card (15). A person has to sign up in a browser to get a key (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, 64 out of 100, up to 6.3 more on the total

Why it scored 64: OAuth through Scalekit with two scopes, `public_docs:read` and `private_docs:read`, or an API key. Key rotation and revocation aren't described in the docs (26). Less 10 because the install docs give `?apiKey=YOUR_API_KEY` in the URL as a documented option (16). Both tools are read-only, the scopes separate public from private search, and team roles limit who can add to the index (16). `ref_read_url` returns third-party web content. Ref's comparison page says Centure.ai detects and blocks prompt injection in scraped and uploaded content, which we can't test (11). ref.tools/activity logs user identity, tool calls and arguments (13). A valid security.txt and a security@ref.tools contact. SOC 2 is in progress, not complete. No bug bounty or published advisories found, and the repository has no SECURITY.md (8).

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. Agent ergonomics, 67 out of 100, up to 5.4 more on the total

Why it scored 67: Two tools whose definitions and server instructions total under 1,000 characters (25). There are no size or paging parameters. The README says a session never returns the same search result twice and a read returns the most relevant sections, about 5,000 tokens at most (12). In the open-source server, failures come back as ordinary text without `isError`, and any 401 from the API becomes a message to verify an email address. No error list is documented, and the hosted server's code is closed (6). Both tools carry `readOnlyHint: true`, with no `destructiveHint` or `idempotentHint`. Both only read, so retries are safe (16). One required parameter per tool. No SDK was found (8).

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, 68 out of 100, up to 5.2 more on the total

Why it scored 68: Both tools have typed JSON Schema inputs in the open-source server (25). docs.ref.tools/llms.txt, llms-full.txt and a Markdown copy of every docs page exist (10). The descriptions say what each tool does and that a read takes the exact URL from a search, with nothing on when not to use them (12). Each tool takes one required string with no enums or bounds, and private search is switched on by writing `ref_src=private` inside the query text (7). The README shows two worked search-and-read sequences with token counts. No error responses are documented (6). The npm package follows semver, but the repository has no changelog file and no GitHub releases, and product changes appear only in monthly newsletters (8).

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, 50 out of 100, up to 4.4 more on the total

Why it scored 50: The hosted service's dated updates on 31 July and 30 September 2026 cover Ref Plans. For Ref Context, the `ref-tools-mcp` package was last published on 4 November 2025 (3.0.3), and the repository's last commits, on 28 September 2026, are dependency updates (15 of 30). Two dated updates in the last 90 days, short of three (0). Closed issues show two to five comments each, September's open pull requests have one comment each, and the docs promise a reply within one business day (15). The official MCP registry lists `tools.ref/ref-tools-mcp` under Ref's own domain, at 3.0.1 from 28 October 2025 (15). Dependabot alerts were cleared on 28 September 2026. The only workflow builds a bundle on release, with no tests (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. Transparency & trust, 68 out of 100, up to 2.8 more on the total

Made of editorial 53, provenance 82.

Why it scored 68: The stdio server is MIT, and it is a thin client. The search service is closed under terms of service that forbid automated access, which the product itself needs (18). The security page describes encryption, tenant isolation, zero-data-retention terms at model providers and a seven-day limit on agent transcripts. The privacy policy of 27 May 2025 gives no retention period for queries or activity logs, no DPA was found, and Centure.ai is named on the comparison page but not in the subprocessor table (18). No deprecation policy found. The pricing page says legacy $9 and $19 subscriptions continue unchanged (3). 19 subprocessors are named with what each can see. Data locations aren't stated (14).

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):

- Domain age: ref.tools, registered 2025-01-20 (1 year) (3 of 15)
- Terms of service: published, but our reader couldn't read it (7 of 10)
- Privacy policy: published, but our reader couldn't read it (7 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: trust.ref.tools returned only its page title to our reader, so the canonical subprocessor list and any policies there were unread
- unchecked: the hosted server's live tool definitions and error behaviour, which need an account. The tool definitions and errors described are from the open-source stdio server
- unchecked: whether sign-up for the 200 free credits needs a card, and whether keys can be rotated or revoked at ref.tools/keys
- Whether the `ratelimit-policy: "500-in-1min"` header seen on an unauthenticated response is the limit for authenticated calls
- Whether Centure.ai processes customer content as a subprocessor, since the security page's table doesn't list it
- Whether the terms' ban on automated or non-human access is meant to apply to MCP use
- How long search queries and activity logs are kept
- The comparison page still quotes a $9 a month plan for 1,000 queries, which the pricing page lists as a legacy subscription

## Weaknesses

- The install docs give `?apiKey=YOUR_API_KEY` in the URL as a documented way to authenticate
- No rate limit, 429 handling or SLA was found in the reviewed documentation
- The free plan is 200 one-time credits, and an account needs a browser sign-up
- The `ref-tools-mcp` npm package was last published on 4 November 2025, and the changelog index stops at July 2026
- The terms of service forbid access "through automated or non-human means", which the MCP product itself requires

## 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 the key in the `x-ref-api-key` header, not the `?apiKey=` query parameter, so it stays out of URLs and logs
- Add `ref_src=private` to the `query` text to search private repositories and PDFs. Searches cover public documentation only by default
- Pass `ref_read_url` the exact URL from a search result. Reads return the sections most relevant to the session's searches, about 5,000 tokens at most per the README
- Keep one MCP session per task. Repeated similar searches in a session never return the same result twice, so refine the query to see more
- Treat tool errors as text. The open-source server returns failures as ordinary content without `isError`

## 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.