Rating certificate · AW/26-09/69126 Sept 2026
Not enough evidence
The Register / Coding agents / Kimi Code CLI
Register entry · Coding agents · updated 26 Sept 2026
Kimi Code CLI
Kimi Code CLI is an MIT-licensed terminal agent that reads and edits code, runs shell commands, searches files and fetches web pages, defaulting to Moonshot's Kimi models (K3, K2.x) via a Kimi Code subscription or an Open Platform API key, and can use other compatible providers. It replaced the Python kimi-cli in 2026 and is distributed as a single binary. Features include subagents, hooks, MCP configuration, a plugin marketplace and video input.
What reviewers and users praise
- Distinct harness design (bidirectional shell integration, preserved thinking, tool injection) noted by reviewer Rohit Ghumare
- Low-cost access to Kimi K3 compared with US frontier subscriptions (HN users)
- Default approval for every non-read action (docs)
What they criticise
- Subscription usage limits exhausted quickly with K3 (SwellJoe review; multiple HN reports)
- Capacity problems: K3 subscriptions paused after demand (press), 'can't keep up with the existing demand' (HN)
- Rough edges: web UI 'slightly broken', users unable to finish projects (HN, Reddit)
Experts vs the crowdneeds both scores
Experts—
Crowd—
| Price | Free, paid from $19 |
|---|---|
| Use it via | CLI, IDE extension, Web app |
| Runs on | macos, linux, windows |
| Licence | MIT |
| Updated | Evidence collected 26 Sept 2026 |
How the rating was worked out
Weights for Coding agents: adoption 35%, experts 40%, crowd 25%. Fewer than two pillars have enough evidence. The method.
| Adoption & momentumweight 35% | 64 | GitHub stars 7.7k → 57 · npm weekly downloads 32k → 41 · VS Code / Open VSX installs 592k → 48 · 52 releases in 90 days → 100reach × 0.7 + momentum × 0.3 |
|---|---|---|
| Expertsindependent reviews · weight 40% | — | Only 2 independent reviews (3 needed) |
| Crowdcommunity sample · weight 25% | — | Community sample from a single platform |
| Rating | — | Not rated: fewer than two pillars have enough evidence. |
Trust & safety
Seven dimensions graded A–D from documented facts. Overall: A.
B
Permission model
Default 'Always Ask' mode: read-only operations run automatically, every edit or command asks for approval; 'Ask When Needed' (formerly YOLO) and Plan modes; permanent allow/deny rules. No built-in sandbox found in docs. source
B
Data access scope
Operates on the working directory with broad file and shell access gated by approvals; sensitive files such as .env and SSH keys still prompt in Ask When Needed mode. Broad access, documented. source
A
Data storage
Runs locally; sessions stored under the local Kimi Code home directory; model calls go to Moonshot or the configured provider. source
n/e
Data retention & training
Moonshot's training/retention terms for Kimi Code subscription traffic not located.
A
Incident history (24 months)
No GitHub Security Advisories for MoonshotAI/kimi-code or kimi-cli and none in the GitHub Advisory Database for @moonshot-ai/kimi-code or PyPI kimi-cli; not named in the September 2026 GitSpawn disclosure. NVD not searched separately. source
n/e
Compliance
No attestation found; not researched further.
A
Transparency
MIT source with public releases and a dated 'What's New' changelog. source
The experts
2 named independent reviews found; dated reviews from the last 12 months are scored.
- positive
No other mainstream harness treats the shell boundary as bidirectional.
Rohit Ghumare · 31 Jul 2026 · rohitghumare.com - mixed
this tiny project (at least, the project is tiny for other models), consumed 94% of one five hour window, and 19% of my weekly usage!
Joe Cooper (SwellJoe blog) · 18 Jul 2026 · swelljoe.com
The crowd
11 coded posts from hackernews.
3 positive3 mixed5 negative
- negative
there's Kimi code and similar which can't keep up with the existing demand
news.ycombinator.com · 20 Sept 2026 · news.ycombinator.com - positive
I'm using Kimi Code with few skills, MCPs and that's basically everything I need.
news.ycombinator.com · 31 Aug 2026 · news.ycombinator.com - mixed
it looks like Z Code is just a Codex fork. Just like the Kimi Code one is. What irks me about this is that the harnesses seem to be just…
news.ycombinator.com · 26 Aug 2026 · news.ycombinator.com - positive
I clicked join on the $20/m plan and got in the same day. Apparently 5x usage when using Kimi Code too.
news.ycombinator.com · 13 Aug 2026 · news.ycombinator.com - positive
am also using Kimi K3 via kimi-code and also GLM 5.2 via ZCode happy with all three
news.ycombinator.com · 12 Aug 2026 · news.ycombinator.com - negative
has stupidly low rate limits for and cost, compared to OpenAI which has massive rate limits
news.ycombinator.com · 11 Aug 2026 · news.ycombinator.com - negative
At least with the Kimi Code plan, its limits are pretty abysmal compared to ChatGPT/Claude plans.
news.ycombinator.com · 27 Jul 2026 · news.ycombinator.com - negative
1.5 days later ive went through more then 50% of my weekly usage and decided to renew the regular anthropic sub too.
news.ycombinator.com · 22 Jul 2026 · news.ycombinator.com
Pricing
As published, checked 26 Sept 2026.
| Plan | Price | Notes |
|---|---|---|
| Kimi Code CLI software | Free | MIT; model access via Kimi Code subscription or Moonshot Open Platform API key. |
| Moderato (legacy) | $19 / monthly | 15 USD/mo billed annually. Reported by HN user 2026-07-19 and corroborated by an independent review snippet; vendor pricing page is JS-rendered and could not be read. Moonshot said Kimi and Kimi Code benefits would be split into new plans (docs now reference 'Plus' and 'Pro'). K3 256k requires Moderato+ (legacy). |
| Allegretto (legacy) | $39 / monthly | 31 USD/mo annually; required for K3 1M context on legacy plans. Same caveat as above. |
| Allegro (legacy) | $99 / monthly | 79 USD/mo annually. Same caveat. |
| Vivace (legacy) | $199 / monthly | 159 USD/mo annually. Same caveat. |
| Plus / Pro (new plans) | Custom | Named in model-access docs (K3 1M on 'Pro and above', K3-256k on 'Plus and above'); prices not visible. |
Recent changes
Ranked alongside
Other ranked entries in Coding agents.
OpenCodeAnomaly75
Claude CodeAnthropic73
Zed (Agent Panel)Zed Industries73
GitHub CopilotGitHub (Microsoft)72
Codex (CLI and app)OpenAI71
Sources
Every figure above traces to one of these, observed up to 26 Sept 2026.