What did Hlido score Claude Code?

Claude Code scored 91/100 (VITAL) in Hlido's independent, hands-on review.

Is Claude Code reliable?

Claude Code is rated VITAL (91/100) by Hlido. Claude Code 2.1.241 — Anthropic’s terminal coding agent. CLI surface live-tested from a cold npm install: non-interactive print mode, first-class MCP management, legible permission flags, fast clean failure without credentials. Agentic loop requires paid auth (untested); medium c

How much does Claude Code cost?

Hlido recorded the pricing model as: Free tier · Subscription. Derived from evidence Hlido holds, not vendor-supplied; verify current prices on the vendor's pricing page.

Does any vendor pay Hlido for placement?

No. Hlido takes no money from the agents it rates — scoring weights stay private and the evidence behind every verdict is public.

Claude Code

Name: Claude Code review
Item: Claude Code
Rating: 91
Author: Hlido Editor

Coding · tested 2026-08-24 · re-test due 2026-11-22 · by the Hlido desk, not the vendor

In short: The reference terminal coding agent — a disciplined, safety-legible CLI whose public surface passed every check we ran.

5 PASS · 0 FAIL of 5 public-surface claims

Visit Claude Code → scorecard.json

Quick answer

Claude Code scores 91/100 (VITAL) on Hlido’s independent, hands-on test (reviewed 2026-08-24). VITAL (91) because every mechanically verifiable claim on the public CLI surface passed, failure modes are fast and legible (the strongest craft signal a CLI can show without credentials), the agent-integration story (MC Pricing: Free tier · Subscription (free entry point documented).

Hlido Editor · 2026-08-24

We installed Claude Code cold from npm (v2.1.241, ~11s, zero vulnerabilities) and probed every public surface a buyer can verify without an account. All of it held: non-interactive -p/--print mode for scripting and CI, first-class MCP server management with inline examples, permission scoping via --allowedTools and --add-dir, a working doctor, and — the detail we weight heavily — when run without credentials it fails in 1.3 seconds with a clean, legible authentication error instead of hanging or crashing. Even its most dangerous flag is named --allow-dangerously-skip-permissions with help text steering it to sandboxes: the safety posture is legible in the surface itself. What this review does NOT cover: the agentic editing loop requires a paid subscription or API key, so we did not live-test multi-file edits the way we did for Aider — the score carries medium confidence accordingly. Disclosure: Hlido’s own review pipeline runs on Claude models from Anthropic, the vendor of this product. The tested checks above are mechanical and reproducible by anyone; the editorial layer is where a reader should apply that knowledge.

Why VITAL

VITAL (91) because every mechanically verifiable claim on the public CLI surface passed, failure modes are fast and legible (the strongest craft signal a CLI can show without credentials), the agent-integration story (MCP management, non-interactive print mode, SDK) is the most complete in the category, and the product is the de-facto reference against which other terminal agents document themselves — our own corpus’s Aider review already compares against it by name. Scored one point under Aider (92), whose full agentic edit loop we did live-test; Claude Code’s we could not, and an untested loop never outranks a tested one on this register.

Public-surface checklist

PASS Version-output (required) — 2.1.241 (Claude Code), exit 0
PASS Help-core-usage (required) — -p/--print, --allowedTools, --add-dir, MCP, background agents documented
PASS Mcp-management (required) — claude mcp add/list with HTTP-transport examples
PASS Doctor-health (required) — doctor exit 0; install/platform/deps reported
PASS Unauthenticated-behavior (required) — clean auth error, exit 1, 1.3s — no hang

What it does well

Fails fast and legibly: unauthenticated non-interactive call returned a clean auth error, exit 1, in 1.3s (tested)
MCP server management is first-class and documented with working examples (tested)
Non-interactive -p/--print mode makes it scriptable by other agents and CI (documented on the tested help surface)
Permission model is visible at the flag level: --allowedTools, --add-dir, and a deliberately scary name for the bypass flag (tested help surface)
Clean cold install from npm, 0 vulnerabilities flagged (tested)

What it fails at

No free tier to verify the agentic loop: everything behind authentication is invisible to an unauthenticated evaluator (tested: auth wall reached in 1.3s)
The product page (claude.com) is a login-walled marketing surface — our browser runner could not evaluate it, which pushes all public verification onto the CLI itself (prior run evidence)

Red flags

Reviewer-vendor overlap, disclosed: Hlido’s pipeline runs on Claude models. The mechanical checks are reproducible; weigh the editorial layer accordingly.

Best for

Teams already on Anthropic models who want the deepest native integration
Agent builders who need a scriptable, non-interactive coding agent (-p mode) with MCP wiring
Developers who weight legible permission scoping and sandbox-aware defaults

Not recommended for

Anyone needing a credential-free trial of the actual editing loop before buying
Model-agnostic teams who want to swap LLM vendors freely (see Aider)

Pricing & access

ModelFree tier · Subscription
Free entry pointYes — a free tier or open-source edition is documented

Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-24.

Compared to

Aider
Aider is open-source, model-agnostic, and its full edit loop is live-tested on this register (92). Claude Code is Anthropic-native with a stronger agent-integration surface (MCP management, print mode) but its edit loop sits behind paid auth we did not exercise. Choose Aider for vendor independence; Claude Code for depth in the Anthropic stack.
Codex CLI
Both are vendor CLIs with MCP support. Claude Code failed fast without credentials (1.3s, clean error); Codex CLI hung for 40+ seconds in the same test. Codex counters with a built-in sandbox subcommand and the ability to run itself as an MCP server.

Agent relevance

API CLI MCP SDK Behavioral-testable

Agentic-Commerce Readiness 73/100 · INTEGRABLE

Independent readiness for agent delegation & transaction. How it’s scored · check live

Install via npm (@anthropic-ai/claude-code); invoke non-interactively with `claude -p "<prompt>"`; scope tools with --allowedTools; attach MCP servers via `claude mcp add`. Requires ANTHROPIC_API_KEY or subscription auth.

Evidence

— source

scorecard.json · registry · methodology

More: compare agents · best of · developer tools · incident registry

Verdict by Hlido Editor, our automated editorial system · Method: cli-tier-1+editorial-narrative-v2+claude-native · Methodology version 2026.05 · Next review due 2026-11-22

How this page was produced. The scores, claim verdicts and evidence come from automated hands-on testing of the product’s public surface. The written analysis is drafted by an AI system, and pages publish without a person reviewing each one. Hlido publishes this record and answers for it — tell us if anything here is wrong and we will correct it.

Embed this trust badge

Live, always-current independent score — free to embed on your site or README. No vendor pays for placement.

Markdown

[![Hlido trust score](https://hlido.eu/badge/claude-code.svg)](https://hlido.eu/check/?agent=claude-code)

HTML

<a href="https://hlido.eu/check/?agent=claude-code"><img src="https://hlido.eu/badge/claude-code.svg" alt="Hlido trust score"></a>