ANNOUNCEMENT
Claude Opus 5.5 is live in Command Code
Claude Opus 5.5 is live in Command Code Max and API.
Command Code raises $5MRead the launch announcementRead more
//Command Log
// Command Log
Agentic engineering, coding taste, and research insights from building Command Code.
ANNOUNCEMENT
GPT 6 Sol and Luna are live in Command Code
Luna is available on all plans and API. E.g. in $10 GOAT plan: 30K reqs, 1.5B tokens. Sol is available on Pro, Max and API
ANNOUNCEMENT
Grok 4.7 is live in Command Code
Grok 4.7 is now live in Command Code. 40% for the next 7 days. 35 credits (3.5x usage) in GOAT
ANNOUNCEMENT
Jev is free in Command Code
Jev is free in Command Code
DEVELOPMENT
Parallel sessions with Git worktrees
Command Code supports Git worktrees. Use /worktree to create or switch and run parallel sessions, each on its own branch.
ANNOUNCEMENT
Step 5 Preview is live in Command Code
Step 5 Preview is now live in Command Code. Available across all plans and API.
ANNOUNCEMENT
MiMo V2.6 Pro and Flash are live in Command Code
MiMo V2.6 Pro & Flash are live in Command Code. Available on all plans. Command Code has the best usage in market for both models.
RESEARCH
Qwen 3.8 Omni Flash vs DeepSeek V4.1 Flash vs Gemini 3.8 Flash on FlappyBench
Ran FlappyBench on Qwen3.8-Omni-Flash, DeepSeek-V4.1-Flash, and Gemini 3.8 Flash with the same design prompt.
ENGINEERING
taste-1 vs Jev: two decision models that are not LLMs
Jev is a general-purpose decision model. taste-1 is a meta neuro-symbolic decision model for one domain, coding taste, running at 100T-token scale this year. Neither is an LLM. Architecture, pipeline, and the differences, in diagrams.
ANNOUNCEMENT
GLM-5.3 FlashX is live in Command Code
GLM-5.3 FlashX is now live in Command Code. High speed variant of GLM-5.3 Flash. 200 TPS with 1M context window
ANNOUNCEMENT
Qwen 3.8 Omni Flash is live in Command Code
Qwen 3.8 Omni Flash is now live in Command Code. Available across all plans and API.
DEVELOPMENT
Fix errors with IDE diagnostics
Stop copy-pasting errors. IDE diagnostics are quite strong. Open Command Code. Select your code. Type fix this
ANNOUNCEMENT
Introducing the Command Code Desktop App
Introducing Command Code Desktop App! Available today in beta for Mac, Linux, & Windows. Building the most powerful agentic app, starting with code....
DEVELOPMENT
Mods are more powerful than plugins
Command Code can build mods and modify any behavior from programmatic hooks to a completely different inference provider.
ANNOUNCEMENT
New last week: DeepSeek V4.1 Flash, Muse Spark 1.3 and Ling 3.0 Flash Sante
Last week these models went live in Command Code. DeepSeek V4.1 Flash. Muse Spark 1.3 (with max reasoning)
ANNOUNCEMENT
Command Code has built-in herdr support
Command Code has built-in herdr support now!
RESEARCH
DeepSeek V4.1 Flash vs GLM 5.3 vs Kimi K3 on FlappyBench
Ran FlappyBench on DeepSeek V4.1-Flash, GLM 5.3, and Kimi K3 with the same /design prompt.
ANNOUNCEMENT
DeepSeek V4.1 Flash is live with 6x usage
DeepSeek V4.1 Flash is live in Command Code. 6x usage on $10/mo GOAT for a week. 60 usage. 154K requests. 7.8B tokens.
ANNOUNCEMENT
Ling 3.0 Flash Sante is free in Command Code
Ling 3.0 Flash Sante is now free in Command Code. 5.1B active params. all subs. all plans. npm i -g command-code@latest
ANNOUNCEMENT
More GOAT customers than Go
GOAT > Go → more GOAT customers than Go now.
DEVELOPMENT
Fork a session with /fork
The /fork command creates a new session from the current one with the same context.
ANNOUNCEMENT
Muse Spark 1.3 Max is live in Command Code
New reasoning effort param for harder problems and frontier performance. Run /model, select Muse Spark 1.3
DEVELOPER STORIES
527 Million Tokens at 20 Cents: How Sivaram Turned DeepSeek V4 Flash Into His Workhorse
Full-stack engineer Sivaram Pandariganthan burned through 527M tokens on DeepSeek V4 Flash for 20 cents using Command Code's tool call repairs, cache-friendly sessions, and vision tool.
DEVELOPMENT
Connect to MCP servers
Connect to MCP servers to query data, use tools and automate workflows.
ANNOUNCEMENT
GPT-6 Astra is live in Command Code
GPT-6 Astra is live in Command Code
ANNOUNCEMENT
Meta Muse Spark 1.3 Contributor is live in Command Code
Meta Muse Spark 1.3 Contributor now live in Command Code. Most capable lowest cost model yet.
RESEARCH
Muse Spark 1.3 vs Gemini 3.8 Flash vs Grok 4.6 on FlappyBench
Ran FlappyBench on Muse Spark 1.3, Gemini 3.8 Flash, and Grok-4.6 with the same /design prompt.
ANNOUNCEMENT
Qwen 3.8 Max 0902 is live in Command Code
Qwen 3.8 Max 0902 is now live in Command Code. It's an upgraded Qwen 3.8 Max: 2.4T params. 1M context
DEVELOPER STORIES
How Command Code Taste Learned to Tell Engineering Process Apart From Production Data Pipelines
Evan uses Command Code for agents, evals, and data work. The taste feature learned to distinguish between his engineering process and production data lake workstreams, even where they overlap.
ANNOUNCEMENT
Gemini 3.8 Flash with 4x usage on the GOAT plan
Gemini 3.8 Flash is now available in Command Code GOAT with 4x usage.
ANNOUNCEMENT
LongCat-2.0 is free in Command Code
LongCat-2.0 is now free in Command Code. 1.6T params. 1M context. 48B activate. Available on all plans to all subscribers.
ANNOUNCEMENT
Fable 5.1 is live in Command Code
Fable 5.1 is now live in Command Code. Claude's most advanced coding model. 75% cheaper cache reads.
ANNOUNCEMENT
DeepSeek V4 Flash Fast is available in Command Code
DeepSeek V4 Flash Fast is now available in Command Code. 2x usage in GOAT. Available all plans and API
DEVELOPMENT
Write multi-line prompts in your editor with Ctrl+G
Use Ctrl+G to open an external editor for your multi-line prompts.
RESEARCH
Hy4 Preview vs Kimi K3 vs GLM 5.3 on FlappyBench
Tested FlappyBench with /design on Hy4 Preview, Kimi K3, and GLM 5.3. Hy4 Preview ($0.0480): Most distinct UI and gameplay
DEVELOPMENT
Review plans with /plan-review
Use /plan-review mode to comment on the plan, revise, and approve it when it’s ready.
ANNOUNCEMENT
Tencent Hy4 Preview is live in Command Code
770B mode, 49B active, 1M context. cmd update available in GOAT and above plans. Our early internal benchmarks show strong perf yet cheaper per task...
ANNOUNCEMENT
Qwen 3.8 Flash with 2x usage on the GOAT plan
Qwen 3.8 Flash is now available in Command Code with 2x usage limits on the GOAT plan.
DEVELOPMENT
Resume previous conversations with /resume
Type /resume and get back to your previous conversation. $ cmd --resume opens a picker for earlier sessions
ANNOUNCEMENT
GLM 5.3 Flash, aka Ox Alpha, is live with 4x usage
GLM 5.3 Flash aka Ox Alpha is now live in Command Code. 4x usage in GOAT plan. 40 credits. 24K reqs. 1.2B tokens
ANNOUNCEMENT
MiniMax M3 and M2.7 are free for 10 days
MiniMax M3 and M2.7 are now free in Command Code. next 10 days. all subs. all plans
DEVELOPMENT
Paste screenshots as chat input
Paste images as your chat input. Take a screenshot. Paste it into the chat. Describe what’s wrong
ANNOUNCEMENT
Alipay is now supported in Command Code
More ways to pay and be a GOAT.
ANNOUNCEMENT
DeepSeek V4 Flash Vision Exp is available on all plans
DeepSeek V4 Flash Vision Exp is live in Command Code. Available on all plans, Go, GOAT, Pro, Max, API.
ANNOUNCEMENT
DeepSeek V4 Flash Vision Exp is live in Command Code
DeepSeek V4 Flash Vision Exp is now live in Command Code. GOAT/$10 per month · 43K requests · 2.2B tokens
ANNOUNCEMENT
Ox Alpha, a stealth frontier model, is free in Command Code
Ox Alpha (stealth frontier model) is free in Command Code. 100T tokens per day free capacity, no 5hr usage limits.
RESEARCH
Qwen 3.8 27B vs DeepSeek V4 Pro 0813 vs Gemini 3.7 Flash on FlappyBench
Tested FlappyBench on Qwen 3.8 27B, DeepSeek V4 Pro 0813, and Gemini 3.7 Flash with /design command.
ANNOUNCEMENT
Qwen 3.8 27B gets 7x usage in the GOAT plan
Qwen 3.8 27B now has 7x usage in GOAT plan. the best coding plan on the market, we've increased usage of Qwen 3.8 27B across all plans.
DEVELOPMENT
Text-only models now have vision
Text-only models now have vision in Command Code. A vision model reads the attached image. Your main model gets the context in the same thread
ANNOUNCEMENT
Qwen 3.8 27B is live in Command Code
Qwen 3.8 27B is now live in Command Code. Beats Opus 4.6 Max, GPT 5.6 Luna Max on the Artificial Analysis Intelligence Index.
DEVELOPMENT
Accessibility checks with /design a11y
Launching a11y mode in /design. Run /design a11y to catch: Tap targets that are too small. Focus management problems
ANNOUNCEMENT
GLM-5.3: same price as GLM-5.2, about 50% better
GLM-5.3 is already live in Command Code. Same price as GLM-5.2, 50% better on Zai's internal Code Bench.
ANNOUNCEMENT
GPT 5.6 Sol is now in the GOAT plan
GPT 5.6 Sol is now in Command Code GOAT. Best coding plan with best GPT model. $70 credits.
ANNOUNCEMENT
DeepSeek raised prices, and your credits are not changing
First things first: your usage credits are not changing. DeepSeek officially raised their prices today, so we've updated our limits and requests accordingly.
RESEARCH
GLM 5.3 vs Fable 5 vs GPT-5.6 Sol on FlappyBench
Tested FlappyBench with GLM 5.3, Fable 5 and GPT-5.6 Sol. 3 models. Same prompt with /design command.
DEVELOPMENT
Let Command Code write and fix your tests
Writing tests is now the easiest part of your workflow. Pick a file, and Command Code will:
ANNOUNCEMENT
GLM 5.3 is live with 2x usage
GLM 5.3 is live in Command Code (2x usage). The best deal: 2x usage, permanently. 10 GOAT plan has $20 worth of GLM 5.3 usage.
ANNOUNCEMENT
GLM 5.3 is the most impressive GLM model yet
GLM 5.3 is easily the most impressive GLM model yet. Excited to partner up and ship GLM 5.3 in Command Code AI soon.
ANNOUNCEMENT
DeepSeek V4 Pro GA: how much cheaper it is
DeepSeek V4 Pro already live in Command Code. Plans. 1 Go ($10 credits). 10 GOAT ($70 credits usage)
ANNOUNCEMENT
Gemini 3.7 Flash is live at 50% off
Gemini 3.7 Flash is now live in Command Code at 50% OFF. It's fast with frontier level intelligence
RESEARCH
Grok 4.6 vs GPT-5.6 Sol vs Opus 5 on FlappyBench
Tested FlappyBench with Grok 4.6, GPT-5.6 Sol, and Opus 5. 3 frontier models. Same prompt with /design command.
DEVELOPMENT
Open well-documented pull requests by asking
Pull requests are easier now. Just ask to open a PR, It will: Read your changes. Draft the description and test plan
ANNOUNCEMENT
DeepSeek V4 Pro 0813 is live in Command Code
DeepSeek V4 Pro 0813 (latest) is live in Command Code. Same price and same model ID. DeepSeek-V4-Pro-0813 is the GA release
RESEARCH
DeepSeek V4 Pro 0813 vs Kimi K3 vs GLM 5.2 on FlappyBench
Tested DeepSeek V4 Pro 0813, Kimi K3, and GLM 5.2 through FlappyBench. 3 models, same prompt with the /design command.
ANNOUNCEMENT
A GOAT dad gets his 8-year-old started with Command Code
GOAT dad gets his 8yrs old started with cmd cli. what's your excuse? > "I paid for my 8-year-old son's @CommandCodeAI
ANNOUNCEMENT
Hermes and Command Code: 25 billion tokens saved
Hermes @CommandCodeAI. 25 Billion tokens saved by Command Code now x 10.
DEVELOPMENT
Code reviews in the terminal with /review
Code reviews should help you ship faster, not block you. Run /review [PR number] and get your code reviews directly in the terminal
ANNOUNCEMENT
The GOAT plan wins over developers switching from Claude and OpenAI
GOAT plan won developers' 💜 switching from claude/openai.
DEVELOPMENT
Align your workflows with /plans
Use /plans to align your workflows. Browse, review, and annotate saved plans. Select plans from current session vs all.
DEVELOPMENT
Browse saved plans with /plans
Browse through plans with /plans. Browse, review, and annotate saved plans. Select plans from current session vs all.
ANNOUNCEMENT
Muse Spark 1.2 and Muse Spark 1.2 Contributor are live
Muse Spark 1.2 by Meta now live in Command Code. Muse Spark 1.2. Available in plans: GOAT, Pro, Max
RESEARCH
Muse Spark 1.2 vs Kimi K3 vs GLM 5.2 on FlappyBench
Tested Muse Spark 1.2, Kimi K3, and GLM 5.2 with our FlappyBench. 3 models, same prompt with the /design command.
ANNOUNCEMENT
Introducing the GOAT plan
Command Code GOAT plan is the best low-cost coding plan on the market today. $10/mo to get $70 credits. 7x what you pay
DEVELOPMENT
Plan mode removes the guesswork
When the change is too big for a quick fix: Your code stays untouched. Plan it first, then review it
ANNOUNCEMENT
Qwen 3.8 Max is live, and going open weight
Qwen 3.8 Max is now live in Command Code Go. ... and it's going open weight!! most importantly, next week open-weights of Qwen3.8-Max and Qwen3.8-27B will...
RESEARCH
Qwen3.8-Max vs Opus 5 vs GPT-5.6 Sol on FlappyBench
We ran a test between the new Qwen3.8-Max, Opus 5 and GPT-5.6 Sol. 3 models. same prompt. one-shot with the /design command.
DEVELOPMENT
Custom slash commands
Create custom slash commands and use them frequently. Save a prompt as a markdown file. Define its scope per project or user level
RESEARCH
DeepSeek V4 Flash vs Kimi K3 vs GLM 5.2 on FlappyBench
We ran a test between the new DeepSeek V4 Flash, Kimi K3 and GLM 5.2. 3 open models. 1 prompt. /design command
ANNOUNCEMENT
GPT 5.6 Luna on the $1 Go plan
GPT 5.6 Luna is now available in Command Code Go ($1 plan) 10x free credits. It’s 95% off right now.
ANNOUNCEMENT
GPT-5.6 Terra and Luna are 50% off
GPT-5.6 Terra and Luna are 50% off in Command Code. Get 2X requests for every dollar of credit you spend.
ANNOUNCEMENT
Command Code co-signs the Open Weights and American AI Leadership letter
Command Code is proud to co-sign the Open Weights and American AI Leadership letter, alongside NVIDIA and 200+ other industry leaders.
ANNOUNCEMENT
DeepSeek V4 Flash 0731 is live in Command Code
DeepSeek V4 Flash 0731 now live in Command Code. DeepSeek V4 Flash now outperforms GLM 5.2
DEVELOPMENT
Agent Skills: stop repeating the same prompt
Stop repeating the same prompt in every session. Use Agent Skills: an open standard for teaching agents specialized knowledge and workflows.
ANNOUNCEMENT
Inkling-Small by Thinking Machines Lab is live
Inkling-Small by Thinking Machines Lab is now live in Command Code. $0.50/M input, 1.20/M output tokens
DEVELOPMENT
Stash prompts with ctrl+s
New in v1: Stashing Prompts. ctrl+s sets the prompt you are typing aside. press it again and it reappears
ANNOUNCEMENT
Qwen 3.7 Flash: 1M context and tiered pricing
Qwen 3.7 Flash is now live in Command Code. 1M context window. Available on all plans (Go, Pro, Max, Team).
ANNOUNCEMENT
Qwen 3.7 Flash is live in Command Code
Qwen3.7 Flash is now live in Command Code. 1M context at $0.03/$0.13 per 1M tokens. Available on all plans (Go, Pro, Max, Team).
DEVELOPMENT
Restore conversation or code with /rewind
Bring back any message or code instantly. Use /rewind to pick a checkpoint, then choose what to restore: conversation, code, or both.
RESEARCH
Opus 5 vs GPT-5.6 Sol vs Kimi K3: Flappy Bird
We ran a test between Opus 5, GPT-5.6 Sol, and Kimi K3. One prompt. Across all three models. With our /design command.
ANNOUNCEMENT
Claude Opus 5 arrives in Command Code
Claude Opus 5 is now live in Command Code. Anthropic's latest thoughtful & proactive model
ANNOUNCEMENT
Claude Opus 5 is now live in Command Code
Anthropic's latest thoughtful and proactive model. On frontier intelligence, closet to Fable 5 at half price
DEVELOPMENT
dont-ask permission mode for unattended runs
New in Command Code: dont-ask permission mode. Runs exactly what your allowlist permits. Everything else gets denied
RESEARCH
Fable 5 vs GPT-5.6 Sol vs Kimi K3: an endless runner
We ran a test between Fable 5, GPT-5.6 Sol, and Kimi K3. One prompt. Across all three models. With our /design command.
ANNOUNCEMENT
Ling-3.0-flash is free in Command Code
Ling-3.0-flash is now free in Command Code. a hybrid-reasoning MoE model super good in our internal benchmarks, beats AntLing's 1T flagship model.
ANNOUNCEMENT
Why one developer picked Command Code over Claude Code, Codex and OpenCode
This developer picked Command Code over Claude Code, Codex, OpenCode here’s why: > I keep coming back to Command Code for personal projects, and it's not...
RESEARCH
GPT-5.6 Sol vs Kimi K3 vs Gemini 3.6 Flash: a brick breaker game
We ran a test between GPT-5.6 Sol, Kimi K3, and Gemini 3.6 Flash. One prompt. Across all three models. Using our /design command.
ANNOUNCEMENT
Laguna S 2.1 is free in Command Code
Laguna S 2.1 is now free in Command Code. US open source model that's most capable of agentic coding in its weight class by a wide margin.
DEVELOPMENT
Push and pull your taste packages
You can pull and push your taste packages. Push what it learned in a project: npx taste push
ANNOUNCEMENT
Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are live
Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are live in Command Code. Available on Pro, Max and Team plans
ANNOUNCEMENT
Command Code is the most used coding agent on OpenRouter
Command Code is the most used coding agent on open router overtaking Claude Code, Kilo, OpenClaw, etc.
RESEARCH
Kimi K3 builds a 3D portfolio in three prompts
Tested Kimi K3 to build a 3D portfolio website. Used the /design command. Built in 3 prompts: build, continue, then recolor
RESEARCH
Kimi K3 builds liquid metal text in Three.js
We tested Kimi K3 to build liquid metal text that wraps with your cursor in Three[.]js. Used the /design command
DEVELOPMENT
Create custom agents with /agents
You can create custom agents for yourself or for your team. Run /agents, describe what it should do, which tools it can access, and Command Code handles the...
ANNOUNCEMENT
Last days of free Hy3 and discounted MiniMax M3
Time to use Hy3 & M3. Developers have used over 1T free tokens on Hy3 using Command Code. We're excited to see what y'all will build in the last few days...
DEVELOPER STORIES
AICodeKing: "The harness matters more than the model"
YouTuber AICodeKing tested Command Code's harness against open models like GLM 5.2 and concluded the gap between open and closed models is smaller than the gap between good and bad harnesses.
RESEARCH
Kimi K3 vs Fable 5: a creative portfolio for $0.035
We compare Kimi K3 with Fable 5. We use our /design command for this test. The result is insanely good for only $0.035. The section and scroll animations,...
RESEARCH
Kimi K3 vs GPT-5.6 Sol vs Fable 5 with /design
Kimi K3 vs GPT-5.6 Sol vs Fable 5. Three top-tier models. Same /design prompt across all. Parameters reviewed:
ANNOUNCEMENT
Command Code is the #1 trending agent on OpenRouter
Command Code is #1 top trending agent on openrouter! #1 in Programming App. #1 in IDE Extensions
ANNOUNCEMENT
Kimi K3 is live in Command Code
First 3T scale open model. Most capable open model ever. It's Fable/Sol class open model beating Opus 4.8
DEVELOPMENT
Project memory with /init and AGENTS.md
Your agents forget everything between sessions. agents[.]md fixes that. Run /init to generate project memory in Command Code. Scans your project and files...
RESEARCH
Four models, one Chrome Dino game
GPT-5.6 Sol vs Fable 5 vs Muse Spark 1.1 vs Grok 4.5. We tested all four using the Chrome Dino game via the /design command.
ANNOUNCEMENT
Inkling by Thinking Machines Lab is live in Command Code
Inkling by Thinking Machines Lab is now live in Command Code. Available on all plans (Go, Pro, Max, Team)
RESEARCH
GPT-5.6 Sol vs Fable 5 vs Grok 4.5: a car racing game
GPT-5.6 Sol vs Fable 5 vs Grok 4.5. Built a car racing game with a UX-over-UI focus, using /design command.
DEVELOPMENT
Permission modes in Command Code
Command Code has permission modes for your coding workflows. Three modes: Default: Approve each action as you go (diff + reasoning)
ANNOUNCEMENT
The first coding agent to support Muse Spark 1.1
Command Code is the first & atm only coding agent that supports Muse Spark 1.1 by Meta. Shipped on the launch day. Excellent score for the cost on Command...
RESEARCH
Fable 5 vs Grok 4.5 vs GPT 5.5: same game, same prompt
Fable 5 vs Grok 4.5 vs GPT 5.5. We put three top-tier models to build a same game challenge.
ANNOUNCEMENT
Muse Spark 1.1 by Meta now available
Muse Spark 1.1 by Meta now available in Command Code.
RESEARCH
Fable 5 vs GPT 5.5 vs GLM 5.2 vs DeepSeek V4 Pro: a retro space shooter
Fable 5 vs GPT 5.5 vs GLM 5.2 vs DeepSeek V4 Pro. We tested 4 models on a retro pixel-art space shooter game.
ANNOUNCEMENT
GPT-5.6 Sol, Terra and Luna are in Command Code
Shipped: GPT-5.6 Sol, Terra, and Luna. Sol, the flagship for long-horizon work. Terra balances intelligence and cost.
ANNOUNCEMENT
Muse Spark 1.1 by Meta is now in Command Code
Muse Spark 1.1 by Meta is now available in Command Code. Built for coding, agentic tasks, and knowledge work.
DEVELOPMENT
Share your sessions with /share
Share your sessions with others. Run /share and get an instant, easy-to-copy link. Best for reviews and debugging
ANNOUNCEMENT
Grok 4.5 is now available in Command Code
Built for coding, agentic tasks, and knowledge work. Available on all plans (Go, Pro, Max)
ANNOUNCEMENT
Tencent Hy3 is free in Command Code
Tencent Hy3 is now completely FREE in Command Code. LIVE NOW. 100% off on tencent-hy3 till JUL 21
DEVELOPMENT
Theme support with /theme
Shipped: Theme Support. Use /theme to switch to dark/light mode. Persists across sessions.
ANNOUNCEMENT
MiniMax M3 is now 62.5% off, permanently
Big news!! MiniMax M3 is now 62.5% off in Command Code. Permanent price drop on one of the best open models.
RESEARCH
DeepSeek V4 Pro vs GLM 5.2 vs Fable 5 on a game build
We tested 3 models on a game build with /design. DeepSeek V4 Pro, GLM 5.2, and Fable 5. Same prompt. Open models shipped UX at 500x less cost than Fable 5.
DEVELOPMENT
Windows DX improvements shipped
Agent stops mis-invoking cmd. MCP servers no longer hang on /exit. Fixed WSL and SSH extension installs
ANNOUNCEMENT
Describe a /goal and get it done
Runs the loop until the objective is met. Completion verified by independent review. /goal status to check progress.
RESEARCH
GLM 5.2 vs Sonnet 5 with /design
We ran a test comparing GLM 5.2 vs Sonnet 5. We used the same prompt, with /design in the command code.
ANNOUNCEMENT
Introducing /goal: The first step toward autonomous loops
Set one objective and walk away. /goal lets Command Code work toward a goal across many turns, and only calls it done after an independent check verifies it from real evidence.
ANNOUNCEMENT
Claude Sonnet 5 is now available in Command Code
Near-Opus performance at Sonnet pricing. Agentic tool use, browser, and terminal access. 2/M input, $10/M output through August 31.
DEVELOPER STORIES
6.9B tokens, 83,022 agent runs, $156: 3 weeks of Command Code
JeffersonNunn.eth ran 6.9B tokens across 83,022 agent runs on Command Code for about $156 in three weeks, keeping 16 windows going every day, and still had spend left.
ANNOUNCEMENT
Usage limits doubled, and models run 5x to 8x faster
We are doubling the usage limits to: 5-hour limit — 30% of your monthly credits. Weekly limit — 60% of your monthly credits
DEVELOPER STORIES
The $1 line in a $281 stack: Command Code is "Insane ROI"
vel (@Velcrafting) ran 4.8B tokens in May across a $281 AI stack. Command Code was the $1 line item, tagged "Insane ROI" next to $200 of OpenAI and $40 of grok.
DEVELOPER STORIES
An entire Android app for under a dollar: 500M+ tokens on Command Code
Yzn (@yazanbruv) ran 500M+ tokens on DeepSeek V4 through Command Code's $1 plan and shipped a fully functional Android app, now in review, for less than a dollar.
ANNOUNCEMENT
GLM-5.2 Fast is live in Command Code
GLM-5.2 Fast is now live on Command Code! The high-throughput build of GLM-5.2, same frontier coding, now at 120–250 tokens/sec for snappier long-horizon runs.
ANNOUNCEMENT
Fugu Ultra by Sakana AI is live in Command Code
Fugu Ultra is live in Command Code. A multi-agent model from Sakana AI built to orchestrate top models for hard, multi-step tasks.
DEVELOPMENT
/fork: branch a session from the current context
New in v0.40.0 ⋅ /fork. Fork new sessions from current context. The original stays intact.
ANNOUNCEMENT
Share your usage stats as a card
"Share your usage" is now live in Command Code! Turn your usage stats into a shareable card for X, Discord, and Reddit.
ANNOUNCEMENT
GLM-5.2 is live in Command Code
GLM-5.2 is now live on Command Code! A frontier open-source reasoning model built for long-horizon tasks.
ANNOUNCEMENT
Kimi K2.7 Code HighSpeed is in Command Code
Kimi K2.7 Code HighSpeed is now in Command Code! A high-speed mode of open-source multimodal coding model, Kimi K2.7 Code.
ANNOUNCEMENT
MiniMax M3 is free in Command Code
MiniMax M3 is now FREE in Command Code! We're partnering up with MiniMax. One of the best open models.
ANNOUNCEMENT
Kimi K2.7 Code is now in Command Code
Kimi K2.7 Code is now in available in Command Code. 10x free credits in Go. Our new #1 open mode in internal benchmarks.
DEVELOPMENT
Everything we shipped in one day
THIS. IS. HOW. WE. DO. IT. all this in one day. feat: add Claude Fable 5. fix: resolve [Text#N] paste placeholder into text for slash commands on submit
ANNOUNCEMENT
Claude Fable 5 is now available in Command Code
Claude Fable 5 now available on Command Code. The Mythos class, state-of-the-art model shows exceptional performance in software engineering tasks, agentic...
DEVELOPMENT
Web search and web fetch in your terminal
New in v0.33.2. Browse the web now from your terminal. Web Search → ranked results. Web Fetch → reads full pages, not a snippet
DEVELOPMENT
/configure-models: pick a model for background tasks
New in v0.32.3 ⋅ /configure-models. Routes background requests to a model of your choice. Run compaction on MiniMax M2.5
ANNOUNCEMENT
NVIDIA Nemotron 3 Ultra is now in Command Code
NVIDIA Nemotron 3 Ultra now available in Command Code! Strongest US open model yet! 1M context
DEVELOPER STORIES
A $200/mo Claude Code user is moving Opus 4.7 work to DeepSeek Pro
Austin Ginder, a heavy $200/mo Claude Code user, ran 21.5M tokens through DeepSeek Pro on Command Code for 75 cents — and started moving his Opus 4.7 work to it.
ANNOUNCEMENT
Command Code crossed 10K paying customers in 30 days
Thank you everyone for all your support. We're excited to ship a lot of cool new stuff we're cooking this month. GUI, ACP, SDK, and more.
ANNOUNCEMENT
Qwen 3.7 Plus is free for an hour
NEW MODEL. Qwen 3.7 Plus is free in Command Code for next hour. Cheaper than max and prev 3.6 plus.
ANNOUNCEMENT
MiniMax M3 is 50% off for a week
MiniMax M3 is now 50% off for the next week! If you haven’t tried it yet, our Go plan starts at just $1 and includes $10 in usage credits.
ANNOUNCEMENT
MiniMax M3 is live in Command Code
MiniMax M3 is live now on CommandCode! A frontier-class open-weight model with 1M context, frontier coding, agentic performance, and native multimodality
ANNOUNCEMENT
Qwen 3.7 Max is free in Command Code
Qwen 3.7 Max is now FREE in Command Code. One of the most premium open models! LIVE NOW. for all subscribers $1 Go, Pro, Max, Ultra!!
ANNOUNCEMENT
$1M run rate, 1 trillion tokens and 9K customers in 24 days
BIG day for us!! Command Code is now one of the fastest-growing coding agent harness for open models. 3rd largest by usage.
ANNOUNCEMENT
Step 3.7 Flash is live in Command Code
Step 3.7 Flash is one of the fastest low-cost open models, now live in Command Code. 400 tokens/sec
DEVELOPMENT
36K+ tool repairs shipped in Command Code 0.28.0
Big!! We just shipped 36K+ tool repairs in command-code@0.28.0. This would improve 29 different models and 36K+ tool errors won't happen.
ANNOUNCEMENT
MiMo-V2.5-Pro and MiMo-V2.5 are ~99% off
MiMo-V2.5-Pro & MiMo-V2.5 are now 99% off on Command Code. This is like 100x more usage! Input, output, and cache pricing are all lower.
ANNOUNCEMENT
Command Code now accepts UPI
Go plan billed in INR at ₹140/month. Card still works everywhere else. Pay in INR, code with 20+ models.
ANNOUNCEMENT
Qwen 3.7 Max is 50% off in Command Code
Qwen 3.7 Max is now 50% off on Command Code. Same budget, 2x the requests. Input, output, cache read/write all discounted
DEVELOPMENT
--model now sets the model for any run
New in v0.26.21. Works in the TUI. Works headless with -p. Fuzzy matching on every model ID.
ANNOUNCEMENT
Introducing the Command Code Provider API
Use your Command Code plan with OpenAI and Anthropic compatible clients using your Command Code API key.
ANNOUNCEMENT
Introducing /design: your terminal-native design partner
Introducing "/design": the skill that turns Command Code into your terminal-native design partner.
ANNOUNCEMENT
Introducing the Command Code Provider API
Introducing Command Code Provider API. Use your Command Code plan through an OpenAI and Anthropic compatible API.
ANNOUNCEMENT
Qwen 3.7 Max is now in Command Code
New in v0.26.20. Command Code supports Qwen 3.7 Max. Frontier coding model with long-horizon agent execution.
DESIGN
/design: the skill that turns Command Code into your design partner
The skill that turns Command Code into your design agent. Eighteen modes for color, type, layout, motion, states, and accessibility. Built against glassmorphism by default
ANNOUNCEMENT
A dollar for $40 of DeepSeek V4 Pro usage
Hard to say no to that. how did we make deepseek outperform opus 4.7? btw we've raised $5M from the best investors in silicon valley, we got you!!
DEVELOPMENT
/plan now takes a task inline
Reads your code before it writes any. Outputs the plan as a file. Edit it. Then run it. Execution is the easy part.
ANNOUNCEMENT
Gemini 3.5 Flash lands in Command Code
New in v0.26.11. Command Code supports Gemini 3.5 Flash. Pro-level coding, parallel agentic execution, 1M token context.
DEVELOPMENT
Configurable reasoning effort in /model
New in v0.26.5. /model now supports configurable reasoning effort. Set it from low through max.
ANNOUNCEMENT
Milestone: Command Code broke 100 billion tokens
Milestone: we broke 100 billion tokens! Command Code was 3rd & 6th most used coding agent1 this week and probably the most used coding harness in the world...
ANNOUNCEMENT
DeepSeek V4 Pro is 75% off: your credits just got 4x stronger
Your credits just got 4x stronger. DeepSeek V4 Pro is 75% off on Command Code until May 31:
ENGINEERING
Rules Rot. Skills Decay. Taste Is What Scales.
Rules files and skills decay over time. The real solution is coding taste, patterns learned from how you actually write code.
ANNOUNCEMENT
OpenAI GPT models, now available in Command Code
Ship faster with OpenAI GPT models in Command Code, with a purpose-built coding harness, taste conditioned on your coding patterns, and cross-model PR review.
ANNOUNCEMENT
OpenAI GPT models are now available in Command Code
Big update. Build, debug, test, refactor, and review code with GPT models while Command Code learns your coding taste.
ENGINEERING
I Gave Two AI Coding Agents the Same Prompt. Only One Wrote My Code.
Ran the same 3 prompts through Command Code and Claude Code. Claude wrote correct code, but Command Code wrote *my* code every single time. The difference is worth seeing.
DESIGN
From Designing Generic UI to Yours: 10 Taste Packages for AI-Assisted Design Systems
Run npx taste pull and Command Code starts building UI in the style of Apple, Vercel, Shopify, or seven others. Learn how design taste files work.
ENGINEERING
Taste-Driven Development: How Command Code Learns Your Coding Style
Command Code learns from your feedback to generate code that matches your coding taste. How we built a system that adapts to how you actually write.
ANNOUNCEMENT
Ten free design system taste packages
Today we're shipping ten free design system taste packages. Inspired from industry best examples like:
ENGINEERING
Deterministic hooks for non-deterministic agents
We shipped hooks today, a programmable API to the Command Code lifecycle. Run your own code, deterministically, at the boundary where the agent crosses from deciding to acting.
ANNOUNCEMENT
Introducing Hooks in Command Code
Gives you deterministic control over agent behavior without relying on the model to follow instructions.
TUTORIAL
AI Terminology #28: Beam Search
A decoding algorithm that explores multiple possible output sequences in parallel and keeps the most likely ones at each step to improve overall results.
TUTORIAL
AI Terminology #27: Top-k Sampling
A decoding method that limits the model’s choices to the k most likely next tokens, balancing coherence and diversity in generated text.
ENGINEERING
Taste vs Skills vs Rules
Why static configuration decays, why tutorials only go so far, and why continuous learning from developer behavior is the missing layer.
ANNOUNCEMENT
Claude Opus 4.7 is live in Command Code
Claude Opus 4.7 is now available in Command Code. Try now! npm install -g command-code
ANNOUNCEMENT
Claude Opus 4.7 is now available in Command Code
Try now!!! npm install -g command-code
TUTORIAL
AI Terminology #26: Temperature
A decoding parameter that controls randomness in model outputs. Lower values make responses more deterministic, while higher values increase creativity and...
ENGINEERING
Agentic PR review workflow with Command Code
Shift from scattered PR feedback to confident shipping decisions with agentic review that understands your codebase, your workflow, and how you think about code.
TUTORIAL
AI Terminology #25: Softmax
A function that converts raw model outputs (logits) into probabilities that sum to 1, helping the model choose the most likely next token.
TUTORIAL
AI Terminology #22: Logits
The raw, unnormalized output values from a model before probabilities are applied, typically passed through a softmax function.
TUTORIAL
AI Terminology #21: Perplexity
A metric used to evaluate language models, measuring how well a model predicts a sequence. Lower perplexity indicates better performance.
ENGINEERING
The Full Stack of Terminals Explained: Terminal, Shell, TTY, Console, POSIX, ANSI Escapes & PTYs
A deep dive into the full stack of terminals — what makes a terminal, how shells work, what TTY and PTY mean, POSIX standards, ANSI escape codes, and more.
TUTORIAL
Better prompting cheat sheet
Be specific (who, what, why, format). Give context (background + goal). Define the role (“Act as...”)
ENGINEERING
Making Engineering Conventions Portable with npx taste push/pull
Taste files solve a personal problem. But engineering is a team sport. Here's how we built the infrastructure to make accumulated engineering judgement portable across teams and open source projects.
ENGINEERING
Why compaction fails in coding agents when it matters most
When long coding sessions hit their limits, most agents break by losing context. Here’s why compaction fails and how a better approach fixes it.
TUTORIAL
AI Terminology #20: Distillation
A training technique where a smaller “student” model learns to replicate the behavior of a larger “teacher” model, preserving performance while reducing...
TUTORIAL
AI Terminology #19: Quantization
A technique that reduces the precision of a model’s numbers (e.g., from 32-bit to 8-bit) to make it smaller, faster, and more efficient to run.
TUTORIAL
AI Terminology #18: Backpropagation
A training algorithm that computes gradients of the loss function and propagates them backward through the network to update model weights.
ENGINEERING
Why Teams Should Be Building Coding Taste
AI coding agents write correct code, not your code. Teams that build and share coding taste (learned preferences from accept/reject/edit signals) compound a 5-10x productivity advantage. Here's the information-theoretic argument for why taste is the next frontier of AI-assisted development.
ANNOUNCEMENT
Introducing Command Code: the coding agent that learns your taste
We've raised $5M to launch the first coding agent that can continuously learn your coding taste.
MARKETING
Replacing a $299/mo Influencer Tool With 200 Lines of Command Code Skill in 3 cents
Learn how our PMM team built a creator-discovery pipeline as a Command Code skill that finds dev creators better than the SaaS tools like Modash and Upfluence. Runs in 10 minutes, and costs about 3 cents per run.
TUTORIAL
AI Terminology #17: Gradient Descent
An optimization algorithm used to minimize a model’s error by iteratively adjusting its parameters in the direction that reduces loss.
TUTORIAL
AI Terminology #16: Multi-Head Attention
A mechanism within transformer models that allows the system to attend to different parts of the input simultaneously, capturing multiple relationships at once.
AI AGENTS
What serious coding agents do on a production codebase
AI that can operate directly on a production codebase. Serious coding agents don’t autocomplete. They build a symbol graph, resolve cross-file references,...
TUTORIAL
AI Terminology #15: Autoregressive Model
A type of model that generates outputs sequentially, where each token is predicted based on previously generated tokens.
TUTORIAL
AI Terminology #14: Tool Calling
The ability of an AI model or agent to invoke external functions, APIs, or services to perform actions beyond text generation.