Cloudflare CLI (cf)

# Fix list: Cloudflare CLI (cf)

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cloudflare-cli, the October 2026 research run, assessed 17 October 2026. Grade B, 63.7 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 Cloudflare CLI (cf): 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, 48 out of 100, up to 10.4 more on the total

Why it scored 48: Read with the local-software lines, since the owner runs the CLI on their own machine. Official npm package `cf`, with Node.js 22.18 or later stated and Bun named as unsupported (20). A public GitHub Actions workflow runs lint and type checks, the CLI's vitest suite on Linux, macOS and Windows (137 test files) and the Wrangler compatibility corpus on every pull request and push to main. We couldn't read the run history, because GitHub's web pages and API refused our requests, so whether main is passing is unchecked (15 of 25). The open issue list was unreadable for the same reason, so crash and regression reports are scored absent (0, unchecked). Changesets and a package changelog that labels breaking changes, such as the kebab-case renames in 1.0.0-beta.16 and the exit-status change in 1.0.0-beta.15. Breaking changes ship between prereleases, which semver allows for a beta (13 of 15). 1.0.0-beta.18, declared beta in the docs and SECURITY.md (0).

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, 50 out of 100, up to 6.3 more on the total

Why it scored 50: No x402, MPP or L402 (0). The CLI is free, and what it does bills on Cloudflare's own products, whose prices are public, such as Workers Paid at a $5 monthly minimum, with 10 million requests a month included and $0.30 per additional million (20). New accounts start on the Workers Free plan, per the pricing page. We didn't sign up to confirm that no card is asked for (20). `cf cli search`, `cf docs search`, `cf schema` and `--dry-run` work with no account. Anything that touches an account needs a person to sign up and approve OAuth in a browser, or to create the token an agent uses (10 of 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.

## 3. Agent ergonomics, 65 out of 100, up to 5.7 more on the total

Why it scored 65: A CLI adds no tool definitions to context, and `cf cli search` returns at most five matches. JSON is always printed with two-space indentation, and there is no field selection beyond piping to jq (18 of 25). List commands pass `--page` and `--per-page` through, but the output strips the API's `result_info`, so an agent can't tell whether more pages exist (10 of 20). Errors print the API's numeric codes and messages to stderr, an unknown command lists the closest matches, and the multi-account error lists the accounts to pick from. Errors are formatted text boxes, not JSON, and exit codes are only 0, 1 and 130 (13 of 20). `--dry-run` on generated commands prints the request without sending it or needing credentials. Destructive commands prompt, need `--force` outside a terminal and exit 1 without it since 1.0.0-beta.15. The API client sets `maxRetries: 0` and the docs give no retry or idempotency guidance (14 of 20). Account and zone are resolved automatically, a zone can be named by domain, and few options are required. One CLI in TypeScript. Cloudflare's API SDKs are separate listings' concern (10 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: OAuth sign-in that a person approves, with `--scopes` to request a subset, or a scoped and revocable API token in `CLOUDFLARE_API_TOKEN`. The Global API Key is not supported and no flag takes a token. OAuth credentials are stored as plain JSONC in the config directory, with keyring support described as future work in the source (27 of 30). All DELETE operations and those the schema marks destructive need confirmation, and outside a terminal only `--force` bypasses it. Named profiles can be bound to one directory, so an agent can hold separate credentials. No read-only mode was found, and the docs warn that `--force` is also an API parameter on some commands, as with `cf workers delete --force` (15 of 20). Output is the account's own API data and Cloudflare's docs search results, which can include text other people wrote, such as logs or form submissions. The docs give privacy guidance for search queries but none on untrusted content (6 of 15). Cloudflare account audit logs record API changes, per the Cloudflare MCP listing's check, and cf's requests carry its version, execution mode and the detected agent harness in headers. `--dry-run` shows each request before it is sent (10 of 15). security.txt names HackerOne and a disclosure policy, read on 17 October 2026. The repository's SECURITY.md is only a beta warning with no route of its own. Actions are pinned by SHA, zizmor checks the workflows, and npm releases carry provenance attestations since 1.0.0-beta.4. Certifications weren't checked this run (14 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, 81 out of 100, up to 3.1 more on the total

Why it scored 81: Graded on the CLI as an agent calls it. Commands are generated from the public Cloudflare API schemas. `cf schema <command>` returns the operation ID, method, path and typed parameters with required flags as JSON, and a hidden `cf tools` prints MCP tool definitions with JSON Schema inputs for every command. No output schema per command (22 of 25). llms.txt for the cf docs and a Markdown copy of every page (10). The agents page says when to use cf and when to keep Wrangler, which search to use for what, and never to run `cf dev`, `cf build` or `cf deploy` in an unmigrated Wrangler project. Generated command descriptions are short API summaries such as "Create D1 Database" (14 of 20). Flags are typed with enum choices and required fields from the schema. The docs say body field lists are incomplete or empty for some commands, which then need a whole `--body` JSON (11 of 15). Worked examples on every docs page. Errors are documented only as a non-zero exit with a message on stderr, and the agents page still gives the old exit status for an aborted delete (9 of 15). Dated package changelog, git tags per release and posts on the developer changelog (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, 68 out of 100, up to 2.8 more on the total

Why it scored 68: 1.0.0-beta.18 on 16 October 2026 (30). 19 prereleases between 25 September and 16 October 2026, and 18 versions since 13 April (20). Issues and pull requests weren't readable, so responsiveness is scored absent. During the beta, a workflow closes outside pull requests that weren't coordinated with the team in advance (0, unchecked). Cloudflare's official CLI, current and published from GitHub Actions with provenance, but in no registry (10 of 15). Renovate and Dependabot update dependencies, and CI runs on three operating systems. Run results unchecked (8 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, 87 out of 100, up to 1.1 more on the total

Made of editorial 80, provenance 94.

Why it scored 87: MIT OR Apache-2.0, with both licence files in the repository and the package (30). telemetry.md lists what command telemetry contains and what is never collected, and says the IP address is handled under Cloudflare's privacy policy. No retention period for telemetry is given. Worker deploys upload npm dependency metadata unless `CLOUDFLARE_DEPENDENCIES_INSTRUMENTATION=false`, which the changelog records but the environment variables page doesn't list. The privacy policy wasn't re-read this run (20 of 30). The docs warn that commands, configuration and Build Output can change before the stable release, the changelog names each breaking change, and the launch post commits to 18 months of Wrangler maintenance after the beta ends. No dated deprecation for cf commands (12 of 20). Telemetry is on by default and fully disclosed, with `cf cli telemetry disable`, `CF_SEND_TELEMETRY=false`, `DO_NOT_TRACK=1` and a `DEBUG=1` view of each payload. Search query text and a hashed agent session ID are part of it (18 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 7 of the 7 things a reader expects, and has 2 clauses that cost points (6 of 10)
- Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10)

## Deductions

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

- Misleading claim, -2. The repository README and the launch post of 28 September 2026 say JSON output is "pretty printed for humans and condensed for agents for maximum context savings". The source has printed every result with `JSON.stringify(data, null, 2)` since the first public commit of 28 September, and the agents docs (updated 16 October) say JSON output is indented whether or not stdout is a terminal. Checked first-hand on 17 October 2026 (https://github.com/cloudflare/cf/blob/main/README.md, https://github.com/cloudflare/cf/blob/main/packages/cli/src/lib/output.ts, https://blog.cloudflare.com/cloudflare-cf-cli-launch/, https://developers.cloudflare.com/cf/agents/).

## 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.

- The scout was wrong on the date. It took 1.0.0-beta.0 of 25 September 2026 as the first release, but Cloudflare published `cf` 0.0.4, described as "The Cloudflare CLI", on 13 April 2026, the day of its technical preview post. `firstReleased` is 2026-04-13. The open beta and the public repository date from 28 September 2026, and the window's change is `cf docs search`.
- Category infrastructure, as the scout suggested, because the CLI's main job is acting on Cloudflare resources across the whole API. `code.deploy` and `code.docs` are kept for `cf deploy` and `cf docs search`.
- Neither of our products. It does not run on the person's device as an assistant (LocalGhost) or route a task to the best tool (letme).
- Judgement call on the deduction. The condensed-output claim is a small one, and the docs state the true behaviour, so -2, the bottom of the range.
- unchecked: GitHub issues, pull requests, stars and CI run results. github.com web pages and the API answered 403 to our requests, and WebFetch's permission request was not answered. The repository was read from a clone.
- unchecked: the default OAuth scope set, which lives in the `@cloudflare/workers-auth` package, not in this repository.
- unchecked: whether the npm dependency metadata upload on deploy is on by default. The changelog implies it, and the default is set outside this repository.
- We did not install or run the CLI. Every claim about behaviour comes from the docs, the source and the changelog.

## Weaknesses

- Beta at 1.0.0-beta.18, with 42 commands renamed in 1.0.0-beta.16 and other breaking changes between prereleases
- List output strips the API's `result_info`, so an agent can't see whether more pages exist
- Telemetry is on by default and records the text of `cf cli search` and `cf docs search` queries
- No automatic retry on 429 or 5xx (the API client sets `maxRetries: 0`), and no backoff guidance in the docs
- The agents and CI docs still describe the old exit status 0 for an aborted delete

## 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.

- Run `cf cli search "<task>"` with the task in quotes, then `cf schema <command>`, before guessing a command name
- Add `--dry-run` to any write first. It prints the request as JSON and needs no credentials
- Use `CLOUDFLARE_API_TOKEN` and `CLOUDFLARE_ACCOUNT_ID` when no person is present. `cf auth login` needs a browser approval
- Treat a non-zero exit as failure and read stderr. Pass `--page` and `--per-page` yourself, since list output carries no paging metadata
- Keep search queries free of account IDs, domains and tokens, or set `CF_SEND_TELEMETRY=false`, because queries are sent in telemetry

## What the review panel asked for

- a programmatic account route
- a read-only token preset
- a date for stable 1.0
- agents docs updated for beta.15
- Return `result_info` in list output
- Add a JSON error format
- keyring credential storage
- read-only profile

## 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.