Moonshot AI Launches Kimi Code CLI, an Open-Source Terminal Coding Agent

·2 views

Moonshot AI has released Kimi Code CLI, an open-source terminal-based coding agent that reads and edits files, runs shell commands, and searches codebases using plain-language instructions. The tool is built with TypeScript, distributed as a single binary under an MIT license, and works natively with Moonshot AI's Kimi models while also supporting providers like OpenAI, Anthropic, and Google. Notable features include video input for converting screen recordings into code, conversational MCP server setup, built-in subagents, lifecycle hooks, and editor integration via the Agent Client Protocol. Kimi Code CLI can be installed on macOS, Linux, and Windows through an official install script, Homebrew, or npm, requiring no Node.js setup for the binary version. Users can launch the interactive terminal UI by navigating to a project directory and running the 'kimi' command, then authenticating via an in-CLI login prompt.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How Category Theory Concepts Map to Functional Programming in JavaScript

A technical explainer outlines how category theory — the mathematics of composition — underpins functional programming primitives in JavaScript. Although JavaScript is not a purely functional language, its support for first-class functions makes it possible to implement core constructs such as Functors, Monoids, and Monads. A Functor wraps a value in a container and exposes a map method, allowing chained transformations without extracting the value. Monoids provide an identity value and an associative combining operation, forming the basis of list reduction via Array.prototype.reduce. Monads extend Functors with a flatMap or chain method that prevents nested containers when functions themselves return wrapped values.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Akhouri Systems Reaches 417 Downloads for ATLOCK Security Tool With No Ad Spend

Akhouri Systems, a small software outfit with no prior audience or mailing list, reached 417 downloads of its Windows security tool ATLOCK as of July 19, 2026, without spending on advertising or promotion. The product, which offers NTFS ACL-level file locking and AES with PBKDF2 vault encryption, gained traction through platform-specific organic posting across developer communities. The team credits transparent communication — including acknowledging product weaknesses — as a key trust-building factor over generic marketing language. An independent third-party 'clean' certification also contributed to credibility with cautious users. Consistent maintenance and visible product updates were cited as further reasons users chose to trust and download the software.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Split Terraform S3 Lifecycle Ownership Caused Silent Data Cleanup Failures

A developer building a media upload system discovered that uncommitted S3 objects uploaded via presigned URLs had no cleanup path because committed and uncommitted files shared the same key prefix. The fix introduced a tmp/ prefix namespace, where unconfirmed uploads expire after 7 days via a lifecycle rule, while validated files are promoted to their final keys using CopyObject. During deployment, the Terraform apply failed because two separate Terraform resources each managed lifecycle rules on the same S3 bucket, unaware of each other. Since S3's PutBucketLifecycleConfiguration replaces the entire lifecycle document rather than appending rules, each resource would overwrite the other's configuration. The incident also surfaced three security and logic defects in the promotion flow, including an idempotency bug and an arbitrary copy primitive, all of which were patched before rollout.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Moonshot AI Launches Kimi K3, a 2.8T Open-Weight Model Topping Code Benchmarks

Moonshot AI released Kimi K3 on July 16, 2026, just ahead of the World Artificial Intelligence Conference in Shanghai. The Beijing-based lab, backed by Alibaba, built K3 as a Mixture-of-Experts model with 2.8 trillion total parameters and approximately 50 billion active parameters per token. The model debuted at first place on Arena.ai's Frontend Code Arena with a score of 1,679, a jump of 17 positions from its predecessor K2.6. K3 supports a 1,048,576-token context window, native image and video understanding, and introduces two new architectural mechanisms — Kimi Delta Attention and Attention Residuals — aimed at faster decoding and improved training efficiency. The open-weight model is compatible with the OpenAI SDK and integrates with developer tools such as Claude Code, Cursor, Cline, and RooCode, though it is priced at frontier-model rates rather than the budget tier of earlier Kimi releases.

0 comments Read more at DEV Community