CLI agent
Claude Code
Anthropic's agentic coding tool that reads a codebase, edits files, and runs commands from terminal, IDE, desktop app, and web.
Compare only what the shape makes comparable.
The client runs locally, but inference requires a hosted provider. Model serving can be routed to an organization's own cloud account via Amazon Bedrock, Claude Platform on AWS, Google Cloud's Agent Platform (formerly Vertex AI), or Microsoft Foundry, or through a self-hosted LLM gateway or corporate proxy. Docs also reference self-hosted environments for running Claude Code on the web sessions on organization-operated compute. No fully offline/self-hosted model option is documented.
Named: Claude Fable 5, Claude Opus 4.6, Claude Sonnet 5, Haiku
- Free$0 per user / month
- Pro$20 per user / month
- Max$100 per user / month
- Team (Standard seat)$25 per seat / month
- Team (Premium seat)$125 per seat / month
- Enterprisecontact
- Anthropic Console (API)contact
- MCP support
- Plugins / custom tools
- Subagents / parallel agents
- Hooks / lifecycle events
- Persistent memory
- Background execution
- Sandboxing / permissions
- Git / PR integration
The Bash sandbox "runs on macOS, Linux, and WSL2. Native Windows is not supported. On Windows, run Claude Code inside a WSL2 distribution." src↗
Claude Code is not included in the Free plan; the Pro card is the lowest tier stating "Includes Claude Code". src↗
Teleport (claude --teleport) and the /desktop handoff each require a claude.ai subscription; the /desktop handoff is available on macOS and x64 Windows only. src↗
Homebrew and WinGet installations do not auto-update; only native installations "automatically update in the background". src↗
- pricing (Team Premium seat)
Two official Anthropic sources disagree. claude.com/pricing shows Team premium seats at "$100" annual / "$125" monthly. code.claude.com/docs/en/third-party-integrations states billing of "Teams: $150/seat (Premium) with PAYG available". Both recorded; the pricing-page figure was used in the tiers array.
- pricing (Max 20x)
claude.com/pricing publishes one Max card at "From $100 Per month" covering both 5x and 20x, and does not display a separate dollar figure for Max 20x anywhere on the page.
- interfaces
Claude Code ships a standalone desktop app (macOS, Windows x64, Windows ARM64) that has no matching value in the allowed interfaces enum, so that surface is not represented in the array.
- model_support
Recorded single-vendor: all named models are Anthropic Claude models. The overview page says the Terminal CLI and VS Code "also support third-party providers", but the third-party integrations page shows those providers (Amazon Bedrock, Claude Platform on AWS, Google Cloud's Agent Platform, Microsoft Foundry) serve Claude models rather than non-Anthropic models.
What people say about Claude Code
69 opinions · 2025-11-19 to 2026-08-19 · medium confidence
Counts are conservative estimates of distinct opinion-holders across the threads read, each tagged by overall polarity toward Claude Code; not a logged per-comment tally. Score is round(100*(22-30)/69) = -12; the negative tilt is channel-driven, see skew_notes.
The largest and most polarized sample of the three, but skewed toward complaint channels by construction: the GitHub slice comes from degradation bug reports, where satisfied users never post, and the biggest HN thread is Anthropic's own April 2026 quality postmortem, which selects for people who were already angry. Against that, one HN thread ('I'm 60 years old, Claude Code has re-ignited a passion') is a pure enthusiasm thread and supplies most of the positive weight, so both poles are over-represented and the middle is thin. Claude Code users are also the most likely to be running the tool daily and professionally, so complaints about limits reflect heavy use, not casual trial.
- hackernews+16
- reddit0
- github-93
GitHub reads 109 points lower than the highest platform here. Issue trackers collect defects only, so that number measures activity more than opinion. Why we publish it anyway.
Sentiment moved in a visible arc rather than a straight line. January-February 2026 brought a wave of independent degradation reports (GitHub #21431 and its duplicates, HN 'Claude Code is being dumbed down?'), which Anthropic largely validated in its 2026-04-23 postmortem confirming three separate harness regressions; a default effort change shown incorrectly in the UI for a month, thinking tokens stripped on session resume, and a verbosity system prompt that hurt code quality. That post converted some anger into grudging trust and some into departure; the highest-signal comments in that thread are about honesty of communication rather than model quality. Positive sentiment about the harness itself never collapsed and is the reason users cite for staying. A second degradation wave appears in August 2026 around the gen-5 models, so the pattern has recurred rather than resolved. The platforms disagree sharply on this product: +16 on Hacker News against -93 on GitHub, a 109-point split. GitHub issues are a breakage queue rather than a mood reading, so treat the repo score as a defect signal and the HN score as the closer read on how users feel.
Praised for
- Best harness and developer experience10×
- Reignites the joy of building8×
- Strong at large multi-step agentic work7×
- Best at UI and design work4×
- Debugging power beyond human patience3×
Complained about
- Perceived quality degradation over time18×
- Rate limits too tight for the price12×
- Poor communication about silent changes7×
- Ecosystem lock-in and third-party bans5×
- Hands-off coding feels unfulfilling4×
- Support quality is poor3×
This requires strong agentic skills. Of which Claude Opus 4.5 is massively better in, in Claude Code, than ChatGPT 5.2 in codex.
Enter Claude Code 2.0. The UX had evolved. The harness is more flexible and robust. Bugs are fixed.
the agent’s performance has noticeably degraded, to the point where it’s nearly unusable.
we noticed same behavior, 3 devs in out team complained independently of each other how bad claude code performed since 2 or 3 days.
I used to be delighted how often claude made the right decision, even though I didn't give it specific guidance. Now it's flipped to making the wrong decision much of the time.
Yes, it's been terrible the last few days, it's making decisions without gathering the full context. I used to be able to add ultrathink when it wasn't quite grasping what I needed, but that's gone now too.
Fast forward decades and Claude Code is giving me that same energy and drive. I love it. It feels like it did back then. I’m chasing the midnight hour and not getting any sleep.
I spent the last 2 days primarily using Claude instead of coding things myself at work. I felt the exact opposite way. It was so unfulfilling. I’d equate it to the feeling of getting an A on a test, knowing I cheated.
I gave Claude the task, a log parser (which it also wrote), and told it to find what each issue was. It did the job in a few minutes.
it's like programming with a couple of buddies. Occasionally they goof off and wreck everything, but we put it back together and end up with a finished project.
Claude allows me to grind through my backlog of ideas in minutes between other tasks, and see which ones take flight.
Now I can just feed them into claude and guide it to completion. It feels great.
anthropic really did a good job with that harness, its feature set is unmatched from the skills integration, background agents/tasks, rules organisation, hooks, it just has the best DX imo
I can easily hit the weekly limit on Claude even on the $200 plan. I have yet to ever hit a rate limit on Codex $100. And the results are almost as good. And don't get me started on Anthropic's extra usage scam.
the experience of suspecting a model is getting worse while Anthropic publicly gaslights their user-base: "we never degrade model performance" is frustrating.
Even just one of these things (messing with adaptive thinking) is enough for me to not trust them anymore. And then their A/B testing this week on pricing.
After all, Claude Code is delivering and I use it to build stuff. Still, I experienced a degradation in quality.