AI Code With Blog

Tutorials Articles

Model usage, API engineering and product practice around Tutorials.

Featured
Connecting Codex to AICodeWith: Dedicated Endpoint, Auth Setup, and Call Verification
Tutorials

Connecting Codex to AICodeWith: Dedicated Endpoint, Auth Setup, and Call Verification

Configure Codex with the dedicated AICodeWith Responses endpoint from the current docs, including config backup, auth.json, model selection, call verification, and recovery steps.

AI Code With•7 min•

Latest 76

What is the UI SDK, and how do you use it? Vercel and shadcn's behavior library for agents, installed and tested in Claude Code / Codex / Cursor (0.1.0-beta.1)
Tutorials

What is the UI SDK, and how do you use it? Vercel and shadcn's behavior library for agents, installed and tested in Claude Code / Codex / Cursor (0.1.0-beta.1)

The UI SDK is Vercel's behavior library for agents, released on 12 November 2026 as 0.1.0-beta.1. Behaviors live in the framework-free @ui-sdk/core package and are rendered by adapters for React, Vue, Svelte, Solid, and Web Components. This guide follows the official docs on what it is, how to install it, and how to add the skill for Claude Code, Codex, and Cursor, and it records the commands, timings, and output from our test machine.

AI Code With · 11/12/202611 min
What Is AWS Emulator and How Do You Install It? Test AWS Apps Locally with Claude Code and Codex (Tested on 1.0.322)
Tutorials

What Is AWS Emulator and How Do You Install It? Test AWS Apps Locally with Claude Code and Codex (Tested on 1.0.322)

AWS Emulator is AWS's official local emulator of its service APIs, generally available since November 11, 2026 (US time). It has no additional service fee, its source is open, it needs no AWS account, and it covers more than 200 services. We tested version 1.0.322 on Debian 13. This guide covers the install commands, pointing the AWS CLI at the local endpoint to create an S3 bucket and a DynamoDB table, the errors we hit, and how to let Claude Code, Codex and Cursor write and test code against the emulator.

AI Code With · 11/12/202612 min
What Is T3 Code and How Do You Install It? Connect Claude Code and Codex, Use a Third-Party API Key, Fix Common Errors (Tested on 0.1.9)
Tutorials

What Is T3 Code and How Do You Install It? Connect Claude Code and Codex, Use a Third-Party API Key, Fix Common Errors (Tested on 0.1.9)

T3 Code is pingdotgg's open-source control surface for coding agents. It ships no model of its own; it drives the Claude Code, Codex, Cursor and other CLIs already on your machine. We tested T3 Code 0.1.9 on Linux. This guide covers the install commands, how to point Claude Code and Codex at an AICodeWith API Key so T3 Code can drive them, a full run of both agents with the t3 thread command, and the errors we hit along the way with their fixes.

AI Code With · 11/12/20266 min
How to Set Up Claude Code / Codex / Cursor Agent Hooks: PreToolUse Comparison, Minimal Script, Fail-Open Test
Tutorials

How to Set Up Claude Code / Codex / Cursor Agent Hooks: PreToolUse Comparison, Minimal Script, Fail-Open Test

Instructions, AGENTS.md, and skills are suggestions the model can ignore. Hooks are deterministic scripts the harness runs at fixed points. This post compares PreToolUse (or equivalent) across Claude Code, Codex, and Cursor—config paths, event names, exit 2 / deny, fail-open vs fail-closed—and includes a minimal blocking script tested on Claude Code.

AI Code With · 11/07/20265 min
Claude Duet + Claude Code: Check Progress and Approve Changes from Your Phone (Plus a Tested Option for API Key Users)
Tutorials

Claude Duet + Claude Code: Check Progress and Approve Changes from Your Phone (Plus a Tested Option for API Key Users)

According to Claude Duet lead @davidsuttonca, Duet now connects to Claude Code sessions: see what is running, approve changes and send updates from your phone, for sessions on your computer or in the cloud. This guide covers the official requirements (subscription sign-in, Remote Control, cloud sessions), explains why a third-party base URL or API key cannot use it, and tests an alternative: steering Claude Code and Codex on a server through tmux and SSH.

AI Code With · 11/07/20268 min
Claude Code Put Your Account Email in User-Agent? How to Upgrade the 2.1.317+ Privacy Mod, Self-Check, and Read Logs
Tutorials

Claude Code Put Your Account Email in User-Agent? How to Upgrade the 2.1.317+ Privacy Mod, Self-Check, and Read Logs

Per GitHub #107757 and HN reports, Claude Code sometimes wrote the signed-in account email into User-Agent on curl-style calls. 2.1.317+ ships cc-plugin-privacy; this guide covers upgrade, plugin confirm, log self-check, coverage limits, and a local version-check timing.

AI Code With · 11/06/20265 min
How to Use Codex 0.167: Agent Command Center, Multi-Backend Switching, Local Image Preview, and Vertex Wizard (Upgrade + Hands-on)
Tutorials

How to Use Codex 0.167: Agent Command Center, Multi-Backend Switching, Local Image Preview, and Vertex Wizard (Upgrade + Hands-on)

Codex CLI 0.167.0 (rust-v0.167.0) adds Agent Command Center, keeps drafts when switching backends, previews local images, and groups API Key / Vertex options under “Sign in another way.” This guide follows the official release notes and times a short coding task with our API Key + custom provider.

AI Code With · 11/06/20264 min
How to Use Agent Autoloop: Let Claude Code / Codex Pick Up Work in the Background and Ask You Only When Needed (Install, rules.json, Hands-on)
Tutorials

How to Use Agent Autoloop: Let Claude Code / Codex Pick Up Work in the Background and Ask You Only When Needed (Install, rules.json, Hands-on)

Agent Autoloop turns Claude Code, Codex, and similar CLI agents into a background loop—events or cron dispatch work, the agent exits after each round, and you get a card only when a decision is yours. This guide follows the official README and SECURITY docs, walks through install plus our measured try → answer → status loop, a morning-brief rules.json example, aicodewith API Key setup, and unattended-security boundaries.

AI Code With · 11/06/20266 min
How to Use Cursor Live Apps: Build, Provision a Database, Pass Security Checks, and Share a Link with Cloud Agent
Tutorials

How to Use Cursor Live Apps: Build, Provision a Database, Pass Security Checks, and Share a Link with Cloud Agent

According to @cursor_ai and Cursor's official docs, Cursor Live Apps let a Cloud Agent build a web app (dashboard, portal, or microsite) from a description, with hosting, Postgres, auth, and Secrets included. You preview first, pick a URL, and only get a shareable live link after deterministic and agentic security checks pass. This guide walks through the official steps and records what we saw when we opened the official demo URL. It is for readers who want Cursor Apps deployment, a Cloud Agent share link, or a one-click publish path.

AI Code With · 11/05/202612 min
Claude Code Background Subagents Use Less of Your Limit: Haiku 5.5 Now Writes the Status Line in 2.1.316 (Upgrade, Check, Configure)
Tutorials

Claude Code Background Subagents Use Less of Your Limit: Haiku 5.5 Now Writes the Status Line in 2.1.316 (Upgrade, Check, Configure)

According to Anthropic's @ClaudeDevs, the status line a Claude Code background subagent refreshes every 30 seconds used to re-read the subagent's whole conversation each time. Haiku 5.5 now writes it from short notes, so each refresh takes about 99% less, sessions with background subagents use about 5% less of the limit and get about 6% more prompts. It needs v2.1.316+. This guide covers who gets it, how to upgrade and check your version, a hands-on run of a background subagent on Claude Code 2.1.318 with an API Key, and how to set the subagent model with CLAUDE_CODE_SUBAGENT_MODEL, agent files and /tasks.

AI Code With · 11/04/20269 min
Conductor is shutting down: run several Claude Code / Codex agents in parallel without it (git worktree guide)
Tutorials

Conductor is shutting down: run several Claude Code / Codex agents in parallel without it (git worktree guide)

Meta has acquired Conductor. According to the official announcement, Conductor Cloud goes read-only on November 13 and shuts down on November 20, when cloud data starts being deleted, while the local Mac app keeps getting updates until early 2027. This guide covers the timeline and what to export, then shows how to do the same thing without Conductor - one git worktree per task, one Claude Code or Codex session per worktree, an API Key, then merge, verify and clean up - with tests of claude --worktree, codex --worktree, a merge conflict, and alternatives such as Orca, Emdash and Claude Squad.

AI Code With · 11/04/202611 min
Did Postman install plugins into Claude Code and Codex? 3-step check and removal (plugins, hooks, MCP, config.toml)
Tutorials

Did Postman install plugins into Claude Code and Codex? 3-step check and removal (plugins, hooks, MCP, config.toml)

According to GitHub issue #14370 and The New Stack, the Postman desktop app (macOS 12.33.4 and 12.35.1) installed its plugin into Claude Code and Codex during an existing-user upgrade and wrote a trusted_hash into ~/.codex/config.toml so its own SessionStart hook runs. This guide covers what happened and Postman's response, a 3-step check for Claude Code, Codex and Cursor, the official removal command and the leftovers it does not remove, plus our reproduction with Postman's official installer in an isolated home directory.

AI Code With · 11/04/202612 min
How to Use DeepSeek Harness Creator Mode: Install, Connect a Third-Party API and Build Your First Plugin
Tutorials

How to Use DeepSeek Harness Creator Mode: Install, Connect a Third-Party API and Build Your First Plugin

On November 2 the official DeepSeek Harness account showed off Creator Mode, where you describe a tool in one sentence and the agent writes the plugin, installs it into Harness and checks that it loaded. This guide explains what Creator Mode is, how to start Harness with npx, how to connect an OpenAI-compatible third-party API (aicodewith base URL + key) to use Claude, GPT and DeepSeek models, and walks through building a todo_scan plugin with Creator Mode, with real outputs, common errors and the Claude Code / Codex equivalents.

AI Code With · 11/03/202610 min
Claude Code auto mode failing every turn behind a gateway? 2.1.317 fixes gateway 403/422 (upgrade + tests)
Tutorials

Claude Code auto mode failing every turn behind a gateway? 2.1.317 fixes gateway 403/422 (upgrade + tests)

Claude Code 2.1.317 (released November 3) fixes auto mode failing every turn when ANTHROPIC_BASE_URL points at an LLM gateway that answers 403 or 422 to the server-side safety check. This guide covers what changed, which third-party gateway and relay users are affected, how to upgrade and check the version, how to set the Base URL and key and turn on auto mode, plus side-by-side tests of 2.1.316 and 2.1.317, an error table and FAQ.

AI Code With · 11/03/20268 min
How to Use Claude Message Threads: Messages API Threads Beta, Send Only New Messages, ~35% Faster First Token (Claude Code / API Tested)
Tutorials

How to Use Claude Message Threads: Messages API Threads Beta, Send Only New Messages, ~35% Faster First Token (Claude Code / API Tested)

On November 2, Anthropic put Message Threads into public beta on the Messages API. The first request sends the full conversation and creates a thread; after that each turn sends only the new messages and the id of the previous response. Long conversations get a faster first token, while token usage and billing stay the same. This guide explains the thread parameter, the beta header, the 5-minute expiry and the 404 replay from the official docs, gives curl, Python and TypeScript examples, and reports our test with an aicodewith key: through aicodewith the thread field is currently ignored and continue turns lose earlier context, Claude Code is not affected, plus the workaround and common errors.

AI Code With · 11/02/202611 min
How to Build Karpathy's HackerTube with Claude Code or Codex: Mine Hacker News Comments for a Video Site Without an Algorithm (Tested)
Tutorials

How to Build Karpathy's HackerTube with Claude Code or Codex: Mine Hacker News Comments for a Video Site Without an Algorithm (Tested)

Karpathy built HackerTube in about two hours with Claude Code and Opus 5.5, picking 6,454 YouTube videos that Hacker News commenters kept recommending over 20 years. We had Claude Code and Codex each build a mini version with our API key: fetch two months of comments from the official HN Algolia API, extract YouTube links, count and dedupe them, and generate a static page. Includes the full prompt, run times, a side-by-side comparison and the steps between the mini version and HackerTube.

AI Code With · 11/02/202611 min
How to Edit Video with Claude Code: levelsio-Style Vlog Rough Cut with Whisper and ffmpeg (Tested)
Tutorials

How to Edit Video with Claude Code: levelsio-Style Vlog Rough Cut with Whisper and ffmpeg (Tested)

levelsio dropped his camera clips into a folder and had Claude Code (Opus 5.5) cut a 10-minute vlog. We rebuilt the workflow: Whisper transcription, picking the best take, a cuts.json edit list, ffmpeg render and subtitles. You get the exact prompt, how to run it in Claude Code and Codex, timings from our test with an aicodewith key, and fixes for misheard names and the PyAV error.

AI Code With · 11/02/20269 min
How to Use Vercel agent-desktop: Let Claude Code, Codex and Cursor Operate Desktop Apps (Setup, Tests, Errors)
Tutorials

How to Use Vercel agent-desktop: Let Claude Code, Codex and Cursor Operate Desktop Apps (Setup, Tests, Errors)

agent-desktop, open-sourced by Vercel Labs on October 31, lets coding agents read and click native apps on macOS, Windows and Linux through the OS accessibility tree instead of guessing from screenshots. This guide explains what it is, how to install it, how to wire it into Claude Code, Codex and Cursor, what happened when we ran it with our API key, and how to fix six real errors.

AI Code With · 11/01/202612 min
Claude Opus 5.5 Effort Levels: How to Set /effort in Claude Code and the API (low to max, Tested)
Tutorials

Claude Opus 5.5 Effort Levels: How to Set /effort in Claude Code and the API (low to max, Tested)

Opus 5.5 has five effort levels (low, medium, high, xhigh, max) and defaults to medium. Many users have recently moved down from xhigh and max to high or medium, reporting similar quality with far fewer tokens. This guide checks those claims against Anthropic's docs and Artificial Analysis data, shows how to set effort in Claude Code with /effort, --effort and settings.json, how to write output_config.effort in API requests, common errors, and timing tests run with our API Key.

AI Code With · 11/01/202611 min
Claude Watermark Starts November 3: Will Claude Code and API Output Be Marked? Impact for Opus 4.7 and Sonnet 4.6
Tutorials

Claude Watermark Starts November 3: Will Claude Code and API Output Be Marked? Impact for Opus 4.7 and Sonnet 4.6

Anthropic will reportedly watermark the output of Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6 from November 3, 2026; newer models such as Opus 5.5 already carry it. The mark is applied at the model level, so it covers the API, Claude Code and cloud partners alike, cannot be switched off and needs no config change. This guide covers which models are affected, whether code carries the mark, impact on quality and speed, whether API Key calls are marked, who can detect it, plus real Claude Code and API test output.

AI Code With · 11/01/20269 min
OpenCode + Claude: Claude Agent Subscription Login vs API Key Setup (2.2.3+)
Tutorials

OpenCode + Claude: Claude Agent Subscription Login vs API Key Setup (2.2.3+)

OpenCode 2.2.3 adds a built-in Claude Agent provider. Install Claude Code 2.1.248+, run claude auth login, then pick Claude Agent in /connect to use your Claude Code login and models. The other route is the Anthropic provider with a baseURL and an API Key. This guide compares both routes, walks through each setup, shows how to move CLAUDE.md rules to AGENTS.md, lists common errors, and includes real output from the API Key route.

AI Code With · 10/31/20268 min
Claude Code VS Code Extension 2.1.313 Shipped an SSH Backdoor: a 5-Minute Check and Cleanup for Cursor / VSCodium Users
Tutorials

Claude Code VS Code Extension 2.1.313 Shipped an SSH Backdoor: a 5-Minute Check and Cleanup for Cursor / VSCodium Users

According to OpenSourceMalware and Phorion, Claude Code extension 2.1.313 on Open VSX had a 66-line loader prepended to extension.js. As soon as the extension loads it adds an attacker key to ~/.ssh/authorized_keys, tries to enable SSH, opens a reverse tunnel to 134.209.82.225 and installs persistence. This guide gives 5 check steps for macOS, Linux and Windows, plus evidence capture, cleanup and credential rotation.

AI Code With · 10/30/20268 min
Claude API 413 request_too_large: the limit is now 128 MB — Bedrock and third-party limits, and how to send large PDFs and screenshots
Tutorials

Claude API 413 request_too_large: the limit is now 128 MB — Bedrock and third-party limits, and how to send large PDFs and screenshots

The Claude API Messages request limit went from 32 MB to 128 MB, but Bedrock is still 20 MB, Google-managed Vertex is 30 MB, and third-party gateways have their own caps. How to recognize 413 request_too_large, a per-platform limit table, compressing and downsampling images, sending big PDFs by Files API file_id, trimming old screenshots in long computer use and Claude Code sessions, plus our own test from 10 MB to 130 MB.

AI Code With · 10/30/20266 min
Google Jules Is Shutting Down (Dec 31): Export Your Data and Move to Claude Code / Codex Background Agents
Tutorials

Google Jules Is Shutting Down (Dec 31): Export Your Data and Move to Claude Code / Codex Background Agents

Google will shut down Jules on December 31, 2026, and it becomes read-only after November 11. This guide walks through the official export (settings, env vars, scheduled tasks, repos, memories, setup scripts, every session and patch) and maps each piece to Claude Code and Codex, with timings from running the same export in both clients (Claude Code about 15 s, Codex about 44 s).

AI Code With · 10/29/20265 min
Why Does Claude Ban Accounts? The Community "Points System" Claim, and Using Claude Code Without Switching Accounts
Tutorials

Why Does Claude Ban Accounts? The Community "Points System" Claim, and Using Claude Code Without Switching Accounts

Since October 28, two posts claiming "Claude bans accounts with a points system" have drawn about 770,000 views on X. We separate the community claims from what Anthropic actually publishes about bans and appeals, then test Claude Code 2.1.310 with an API Key from our site: simple requests came back in about 1.4–1.7 seconds, and a small fix-a-bug-and-write-tests task finished in 8.5 seconds. Codex and Cursor setup plus 401/429 troubleshooting included.

AI Code With · 10/29/20269 min
How to install and use tstack: Theo's 9 agent skills tested in Claude Code and Codex, with sol-loop running GPT-6.1 Sol on the same Key
Tutorials

How to install and use tstack: Theo's 9 agent skills tested in Claude Code and Codex, with sol-loop running GPT-6.1 Sol on the same Key

Theo open-sourced tstack, the 9 agent skills he uses every day. We tested it in Claude Code 2.1.310 and Codex 0.164.0 with a single API Key from this site: install took about 3 s; sol-loop in Codex ran 2 rounds in 62 s, and Claude Code ran 2 rounds in 70 s once we added a Sol subagent. On both sides gpt-6.1-sol caught an integer-division bug. We also cover why full-send lets the Agent merge PRs on its own, and the config that blocks merging in both clients.

AI Code With · 10/28/20268 min
Can an AGENTS.md in Downloads hijack Codex and Claude Code? How to audit config.toml and settings.json, prevent it, and rotate your Key (with check commands)
Tutorials

Can an AGENTS.md in Downloads hijack Codex and Claude Code? How to audit config.toml and settings.json, prevent it, and rotate your Key (with check commands)

According to a Forever Security report, an AGENTS.md dropped into ~/Downloads can get Codex to add an MCP server to its global config.toml and install a malicious npm package, and a crafted CLAUDE.md can get Claude Code to point ANTHROPIC_BASE_URL at an attacker's server. This guide gives macOS/Linux and PowerShell check commands, shows what a correct AICodeWith setup looks like, reports our trust_level test results, and explains how to rotate your Key if you were hit.

AI Code With · 10/27/202610 min
Migrating from Claude Code to Codex: moving rules, skills, MCP and hooks with /import — does model_provider survive the import? (tested with a third-party API Key)
Tutorials

Migrating from Claude Code to Codex: moving rules, skills, MCP and hooks with /import — does model_provider survive the import? (tested with a third-party API Key)

Codex CLI's /import can move Claude Code's CLAUDE.md, settings.json, hooks, skills, MCP, subagents and recent conversations into Codex. In our test, the existing model_provider = "aicodewith" was not changed and calls worked normally after import; but the Key in settings.json gets written into Codex's shell environment and has to be removed by hand.

AI Code With · 10/27/20267 min
Sharing one AGENTS.md between Claude Code and Codex: not read before 2.1.281 with a third-party API, plus a skillshare skill sync test
Tutorials

Sharing one AGENTS.md between Claude Code and Codex: not read before 2.1.281 with a third-party API, plus a skillshare skill sync test

One AGENTS.md for both Claude Code and Codex. We tested with a Key from this site: Claude Code 2.1.280 does not read AGENTS.md, while 2.1.281 and later do; older versions can also read it if you add one line, @AGENTS.md, to CLAUDE.md. Skills are synced to both sides in one go with skillshare.

AI Code With · 10/27/20267 min
How to use GPT-6.1 Terra in Codex and the API: config.toml setup, reasoning effort values, measured timing and common errors
Tutorials

How to use GPT-6.1 Terra in Codex and the API: config.toml setup, reasoning effort values, measured timing and common errors

OpenAI released GPT-6.1 Terra early on October 27 Beijing time; the API model name is gpt-6.1-terra. This article shows the config.toml setup for switching to Terra in Codex and a curl command for calling the Responses API directly. A small fix-a-bug-and-add-tests task took about 18 seconds in our test, and we list the valid reasoning effort values plus common errors such as ultra returning 400.

AI Code With · 10/27/20265 min
How to use Codex /review: uncommitted changes, comparing against main, and codex review in CI (checked against the official rubric)
Tutorials

How to use Codex /review: uncommitted changes, comparing against main, and codex review in CI (checked against the official rubric)

Tested with Codex CLI 0.164.0 and an API Key from this site: codex review --uncommitted flagged two [P1] issues on a clear regression (about 15.7s); minor flaws can produce an empty result; --base main took about 11.8s. Checked against the rubric, and with auto_review clearly separated from code review.

AI Code With · 10/26/20264 min
How to stop Codex from editing /etc/fstab: 0.166 sandbox and /permissions tested (untrusted retired, workspace-write, common errors)
Tutorials

How to stop Codex from editing /etc/fstab: 0.166 sandbox and /permissions tested (untrusted retired, workspace-write, common errors)

According to GitHub issue #59060, a user asked Codex CLI 0.166.0 only to reboot a server, and it edited /etc/fstab and touched systemd and containers. We tested Codex 0.166.0 writing files outside the workspace under read-only, workspace-write and danger-full-access, showed the on-request approval prompt and how codex exec differs, and give a config.toml that keeps an agent away from system files plus recovery steps for a broken fstab. The untrusted retirement error and migration are rechecked on 0.166.0.

AI Code With · 10/26/202611 min
How do you accept AI-written code? Beyond reviewing the diff, drive the real app with a verification skill (Cursor pstack)
Tutorials

How do you accept AI-written code? Beyond reviewing the diff, drive the real app with a verification skill (Cursor pstack)

Accepting AI changes shouldn't stop at "it compiles" or a diff review. This article separates pre-PR AI code review from a verification skill (driving the real app and leaving evidence), walking through Cursor pstack's /create-verification-skill: the install chain, Launch/Doctor/Drive/Evidence/Cleanup, the feature map and Prove. No timing figures we did not reproduce.

AI Code With · 10/26/20266 min
Don't rush to stack sub-Agents: fold workflows into Claude Code Skills as Anthropic recommends (with a minimal side-by-side)
Tutorials

Don't rush to stack sub-Agents: fold workflows into Claude Code Skills as Anthropic recommends (with a minimal side-by-side)

Anthropic points out that many companies' multi-Agent orchestration is slower and loses more context. This article maps that advice onto Claude Code: when to use a Skill and when to keep a Subagent. It gives a minimal Skill vs Agent file comparison for a pre-PR check, and explains why workflows that need independent refutation, such as security-audit, should not be forced into one loop.

AI Code With · 10/26/20266 min
How to use Claude Code's Monitor tool: since 2.1.310 it watches logs on Bedrock and with telemetry off too, plus a hands-on pipe buffering test
Tutorials

How to use Claude Code's Monitor tool: since 2.1.310 it watches logs on Bedrock and with telemetry off too, plus a hands-on pipe buffering test

Claude Code 2.1.310 makes Monitor and plugin monitors available on Bedrock, Vertex and Foundry, and when telemetry is turned off. We compare the changelog with older docs, check the SDK fields in a local 2.1.310 install (timeout capped at 1800000ms), and test the same log pipe both ways: without --line-buffered the first event takes 1.213 s, with it 0.009 s.

AI Code With · 10/26/20265 min
How to use Cloudflare's security-audit skill: a six-phase code security audit in Claude Code, with every candidate vulnerability handed to a fresh Agent to refute (tested with a third-party API Key)
Tutorials

How to use Cloudflare's security-audit skill: a six-phase code security audit in Claude Code, with every candidate vulnerability handed to a fresh Agent to refute (tested with a third-party API Key)

Cloudflare open-sourced security-audit, the skill its vulnerability discovery system started from. We tested it in Claude Code with an API Key from this site: a standard run on DVNA, a Node practice project with deliberate vulnerabilities, took 52 min 57 s and produced 14 candidates, 4 confirmed and 10 needs_validation; two runs on the mature libraries serve-static + send found nothing. Includes the install command, the three profiles, and terminal output from the refutation step.

AI Code With · 10/26/202613 min
Does AI-built web design scream AI at first glance? Install Hallmark v2 for Claude Code / Codex to build pages from 48 themes and self-check them (tested with a third-party API Key)
Tutorials

Does AI-built web design scream AI at first glance? Install Hallmark v2 for Claude Code / Codex to build pages from 48 themes and self-check them (tested with a third-party API Key)

Why does AI-generated web design give itself away at a glance? Hallmark v2 is a web design skill for Claude Code and Codex: it first picks 3 of its 48 themes for you to choose from, lays out the page around the landing page's job, then self-checks in a real browser at 8 widths. In this post we test installation with this site's API Key, build a landing page for an expense-tracking app, and run /hallmark audit and /hallmark redesign, with before-and-after comparisons and timing for each step.

AI Code With · 10/25/202610 min
What's New in Codex CLI 0.164? /presets Prompt Presets, One-Line Reasoning Effort Switching with /model, Windows Corporate Proxy (Tested with a Third-Party Key)
Tutorials

What's New in Codex CLI 0.164? /presets Prompt Presets, One-Line Reasoning Effort Switching with /model, Windows Corporate Proxy (Tested with a Third-Party Key)

Codex CLI 0.164.0 adds /presets with 10 numbered slots, inline arguments for /model and /permissions, a fix so resume restores session settings, and Negotiate/NTLM authentication for Windows corporate proxies. This post quotes the official release notes to explain each change, then uses a third-party API Key and a custom provider to test cache hits and resume after switching reasoning effort with /presets and /model.

AI Code With · 10/24/20267 min
How to Use Claude Code MCP Tasks: Run Long Jobs in the Background, Confirm Mid-Task with Elicitations, Supported Since 2.1.304 (Tested with a Third-Party API Key)
Tutorials

How to Use Claude Code MCP Tasks: Run Long Jobs in the Background, Confirm Mid-Task with Elicitations, Supported Since 2.1.304 (Tested with a Third-Party API Key)

Claude Code v2.1.304 introduces MCP Tasks in public beta: the MCP server takes over a long job and runs it itself, so Claude doesn't have to wait, and remote tasks keep running after you exit. This post follows the official docs to explain the requirements for enabling it, how to use /tasks, the two variables for waiting on results, and how it differs from "calls over 2 minutes automatically move to the background." We also tested auto-backgrounding, /tasks, and Elicitation mid-task confirmation on v2.1.309 with a third-party API Key.

AI Code With · 10/23/202611 min
Kimi K3.1 Pro Launches: Claude Code Setup Guide, How It Differs from K3, and Choosing Highspeed
Tutorials

Kimi K3.1 Pro Launches: Claude Code Setup Guide, How It Differs from K3, and Choosing Highspeed

Kimi released K3.1 Pro and K3.1 Pro Highspeed at 01:15 Beijing time on October 24. This post covers the key points from the official announcement, how K3.1 Pro differs from K3, the Claude Code settings.json configuration (the official K3.1 Pro model ID and the kimi-k3 model currently available on our platform), and a coding task run with kimi-k3 inside Claude Code.

AI Code With · 10/23/20265 min
The Claude API Has a Stable Release, 2026-10-21: Upgrade Your SDK with `/claude-api` in Claude Code, and Sending temperature Now Returns a 400
Tutorials

The Claude API Has a Stable Release, 2026-10-21: Upgrade Your SDK with `/claude-api` in Claude Code, and Sending temperature Now Returns a 400

Anthropic has released 2026-10-21, the first formal API version since 2023-06-01. Messages requests that still send temperature now return a 400. Claude Code can migrate your code with the `/claude-api` skill. This post covers the upgrade steps, setting up your key on this site, and side-by-side test results.

AI Code With · 10/21/20268 min
Claude Code / Codex Build a Game with Godot: Install the Game Dev Skill Pack and Test Double Jump and Inventory
Tutorials

Claude Code / Codex Build a Game with Godot: Install the Game Dev Skill Pack and Test Double Jump and Inventory

The game dev skill pack awesome-gamedev-agent-skills ships 74 skills and a router, and it installs into both Claude Code and Codex. On Linux with Godot 4.5.1, we created a new 2D platformer project and gave each tool two prompts: "add a double jump to the character" and "add an inventory." Codex automatically read 4 Godot skills and finished both steps in 322 seconds. Claude Code won't use skills unless you name them; after typing /router, it recognized Godot 4.5 on its own, loaded 2 skills, and finished the double jump in 82 seconds with a self-test. Includes install commands, config, and real errors.

AI Code With · 10/21/20269 min
How to Build a WordPress Site with Claude Code / Codex? Hands-On with the WordPress Studio 2.0 Official Plugin, from Install to Local Site to Publishing
Tutorials

How to Build a WordPress Site with Claude Code / Codex? Hands-On with the WordPress Studio 2.0 Official Plugin, from Install to Local Site to Publishing

On October 19, Automattic released WordPress Studio 2.0 and the wp-studio plugin, and both Claude Code and Codex can install it. We installed it on each side using our site's Key: the 4 plugin install commands took about 4 seconds in total. Claude Code built the site, made a Cover homepage, and published a post from a single prompt in 119 seconds. Codex got stuck on the plugin's built-in MCP, so we wired it up manually with the npm version of the Studio CLI and finished in 155 seconds. Configs, commands, and common errors included.

AI Code With · 10/20/20269 min
Claude Code Blender Test: Install 22 Blender Skills and Have AI Build Game-Ready Low-Poly 3D Models (Works with Codex Too)
Tutorials

Claude Code Blender Test: Install 22 Blender Skills and Have AI Build Game-Ready Low-Poly 3D Models (Works with Codex Too)

Can AI make game 3D models if you don't know Blender? We installed the open-source Blender Agent Skills (22 skills) into Claude Code and Codex, ran them in the background with blender -b, and had each build a low-poly treasure chest. Claude Code took 341 seconds and produced 1028 triangles. Codex took 219 seconds and produced 1836 triangles. Both have UVs and PBR materials, and the exported .glb files showed 0 errors in the Khronos glTF Validator.

AI Code With · 10/20/20267 min
Migrating Legacy COBOL to Java/Python with Codex: GPT-6.1 Sol Hands-On Steps, Choosing Reasoning Effort, and the Money Rounding Trap
Tutorials

Migrating Legacy COBOL to Java/Python with Codex: GPT-6.1 Sol Hands-On Steps, Choosing Reasoning Effort, and the Money Rounding Trap

On October 19, OpenAI reported that GPT-6.1 Sol scores 86.6% at low reasoning effort and 97.9% at high reasoning effort on its COBOL migration eval. In this post, we connect Codex to gpt-6.1-sol and migrate a COBOL program with fixed-point amounts and missing records to Python, report the measured time and test results at both low and high settings, and explain the truncation and rounding pitfalls.

AI Code With · 10/19/20269 min
How do you use Claude Code and Codex together? claudex-loop has the two models review each other's plans and code (one Key for both ends)
Tutorials

How do you use Claude Code and Codex together? claudex-loop has the two models review each other's plans and code (one Key for both ends)

Want Claude Code and Codex working together, one writing the plan and the other picking it apart? The open-source skill pack claudex-loop turns that into a fixed workflow: one side drafts the plan, the other reviews it, code gets written after approval, and then the other side checks the result. We connected both ends with a single AICodeWith Key and ran the full flow twice in a real git repo: once in 125 seconds, once in 314 seconds. During the planning stage, Codex caught problems like "exporting to the database file would overwrite the database."

AI Code With · 10/19/20269 min
How to Edit Earlier Messages in the Codex Desktop App: Does Your Code Roll Back After You Change the Prompt? (Plus CLI Double-Esc Rewind and a git Safety Net)
Tutorials

How to Edit Earlier Messages in the Codex Desktop App: Does Your Code Roll Back After You Change the Prompt? (Plus CLI Double-Esc Rewind and a git Safety Net)

Editing an earlier message and rewinding to a previous turn to start over is one of the most common moves when requirements change. But does it undo your code too? We tested it in Codex CLI 0.163.0: after three rounds of changes, we double-pressed Esc to rewind to the first message and resent it. The prompt said "File changes are unchanged" outright, and every file change from rounds two and three was still there. This article covers how to rewind in the CLI, what the desktop app's official docs say, and how to use git as a safety net to undo code.

AI Code With · 10/19/20267 min
How Does Claude Code Control Unreal Engine? Epic's Official Plugin, MCP Connection, and Security Settings (Hands-On with UE5)
Tutorials

How Does Claude Code Control Unreal Engine? Epic's Official Plugin, MCP Connection, and Security Settings (Hands-On with UE5)

Epic Games offers a Claude Code Unreal Engine plugin, unreal-engine-skills-for-claude-code, in the official Anthropic plugin marketplace. Paired with Unreal MCP in the editor, it lets Claude Code operate the UE5 editor directly. This article covers which two plugins to enable in the editor, how to start the MCP server, how to install the plugin and verify the connection on the Claude Code side, and security settings such as execute_tool_script. We tested with an AICodeWith Key: the plugin installed in 2.5 seconds, and no Anthropic account login was needed at any point.

AI Code With · 10/18/20269 min
How to Install Claude Code Role Plugins? Install knowledge-work-plugins for Sales, Support, and Finance in One Step, No CRM Required
Tutorials

How to Install Claude Code Role Plugins? Install knowledge-work-plugins for Sales, Support, and Finance in One Step, No CRM Required

Anthropic's open-source knowledge-work-plugins bundles plugins by role, including sales, customer support, finance, and legal, and you can install them in Claude Code with two commands. We tested with an AICodeWith Key: the install never asked for an Anthropic account login and finished in under 3 seconds. With no CRM connected, /sales:call-prep generated a complete pre-meeting briefing in 45 seconds and spelled out which information it couldn't get.

AI Code With · 10/18/20268 min
Cleaning Up Dead Code and Stale Feature Flags with Codex: 3 Prompts from OpenAI Engineers, Tested Locally
Tutorials

Cleaning Up Dead Code and Stale Feature Flags with Codex: 3 Prompts from OpenAI Engineers, Tested Locally

OpenAI engineers used Codex to clean up dead code and stale feature flags: 5 people merged more than 1,400 PRs in two weeks and deleted over 800,000 lines of code. This post adapts their 3 prompts into a general-purpose version and tests them on the open-source project jmespath.py with an AICodeWith Key: 133 seconds for report-only, 43 seconds for direct cleanup, and no mistaken deletions after checking every item by hand.

AI Code With · 10/17/20266 min
Claude Code Running Slow? gzip Request Compression Since 2.1.300, the Toggle, and What to Know About Third-Party APIs
Tutorials

Claude Code Running Slow? gzip Request Compression Since 2.1.300, the Toggle, and What to Know About Third-Party APIs

Starting with 2.1.300, Claude Code compresses large requests sent to api.anthropic.com with gzip. You can turn it off with CLAUDE_CODE_GZIP_REQUEST_BODIES=0. We tested with an AICodeWith Key: when you connect to a third-party API, requests aren't compressed in the first place, so the toggle has no effect on time to first token. If responses are slow, start with session length, the model, and Fast Mode.

AI Code With · 10/16/20264 min
How Does Claude Code Do Code Review? Using /code-review, /security-review, and REVIEW.md
Tutorials

How Does Claude Code Do Code Review? Using /code-review, /security-review, and REVIEW.md

The Claude Code team's own code review process works like this: verify your own work first, run every PR through AI review, let security review block merges, and set the number of approvers based on risk. Individuals and small teams can build the same setup with the local /code-review and /security-review commands plus a REVIEW.md file. We tested it with an AICodeWith Key, and both commands ran without problems.

AI Code With · 10/16/20267 min
How to Hand Off a Half-Finished Claude Code Task to Codex: Pass Context with handoff.md (Including Switching Between Both Tools with One Key)
Tutorials

How to Hand Off a Half-Finished Claude Code Task to Codex: Pass Context with handoff.md (Including Switching Between Both Tools with One Key)

Claude Code and Codex don't share sessions, so context doesn't follow you when you switch tools. The fix is to have Claude Code write its progress to handoff.md in the project root, then have Codex read only that file and pick up the work. This post gives you a copy-ready template and two prompts, and tests switching between both tools with the same AICodeWith Key.

AI Code With · 10/16/20265 min
How to Write SKILL.md: Build Your Own Claude Skill to Capture Repeated Workflows
Tutorials

How to Write SKILL.md: Build Your Own Claude Skill to Capture Repeated Workflows

If you keep pasting the same instructions to an AI, it's time to turn them into a Skill. This article covers which tasks are worth turning into a Skill, how to write the name, description, and body of SKILL.md, and walks through a complete example from scratch. It also shows how to write a description that triggers accurately, then test and iterate.

AI Code With · 10/16/20266 min
How Do You Use Agent Skills? A Hands-On Guide to Install Directories and Triggers in Claude Code / Codex (Now Handed to the Linux Foundation's AAIF)
Tutorials

How Do You Use Agent Skills? A Hands-On Guide to Install Directories and Triggers in Claude Code / Codex (Now Handed to the Linux Foundation's AAIF)

On October 22, Agent Skills was handed to the Linux Foundation's AAIF, so a single skill can work across multiple clients. This article covers the install directories and trigger methods for Claude Code Skill and Codex Skill, the results of testing the same SKILL.md on both, and what to check before installing a third-party skill.

AI Code With · 10/16/202612 min
How to Use Claude Opus 5.5 Fast Mode: Enable /fast in Claude Code, Pricing, and Fixing the "disabled by your organization" Error
Tutorials

How to Use Claude Opus 5.5 Fast Mode: Enable /fast in Claude Code, Pricing, and Fixing the "disabled by your organization" Error

Opus 5.5 Fast Mode is the same Opus 5.5, just up to 2.5x faster on output, priced at $8/$40 per million tokens. This article covers how to turn on /fast in Claude Code, how to add speed:"fast" to API calls, and why you see "Fast mode has been disabled by your organization" when connecting with a Base URL + Token.

AI Code With · 10/15/20268 min
What to Do About Codex's "Selected model is at capacity" Error? It's Not a Quota Issue: 503 Causes, Auto-Retry, and Switching Models
Tutorials

What to Do About Codex's "Selected model is at capacity" Error? It's Not a Quota Issue: 503 Causes, Auto-Retry, and Switching Models

When Codex says Selected model is at capacity, you haven't run out of quota. The server is overloaded (HTTP 503). This post walks you through telling the errors apart, waiting for auto-resume, switching models, and raising the retry count, with a table of common errors.

AI Code With · 10/15/20265 min
Claude Code Invoice Skill Tested: Organize Invoices Automatically with invoice-organizer
Tutorials

Claude Code Invoice Skill Tested: Organize Invoices Automatically with invoice-organizer

With Claude Code and the invoice-organizer skill, one sentence turns a folder of XML e-invoices into a ledger, and it automatically flags duplicate invoices and invoices whose amounts don't add up. In our test it processed 6 invoices in 63 seconds and left every original file untouched.

AI Code With · 10/15/20265 min
How to Remove the AI Flavor: Use Claude Code to Rewrite AI Articles So They Read Like a Human Wrote Them
Tutorials

How to Remove the AI Flavor: Use Claude Code to Rewrite AI Articles So They Read Like a Human Wrote Them

Use Claude Code with the humanize-ai-text skill to detect AI-sounding phrasing in an article, rewrite it into natural prose, and re-test. In our run it finished in 42 seconds, and by the skill's built-in rules the style score dropped from 20 to 3.

AI Code With · 10/15/20266 min
AI for Excel: Clean a Messy Spreadsheet with One Prompt and Auto-Generate SUMIFS Summary Formulas
Tutorials

AI for Excel: Clean a Messy Spreadsheet with One Prompt and Auto-Generate SUMIFS Summary Formulas

Use Claude Code with the official xlsx Skill to clean up a messy sales sheet in one prompt and generate SUMIFS summary formulas that recalculate automatically. In our test it finished in 61.5 seconds, and every summary by month and by salesperson is a real Excel formula.

AI Code With · 10/15/20267 min
Local Q&A on Your Own Docs: A Zero-Dependency Python Build with Claude Code
Tutorials

Local Q&A on Your Own Docs: A Zero-Dependency Python Build with Claude Code

Have Claude Code write a local document Q&A script that uses only the Python standard library. It searches by paragraph, returns the original sentence and its source, and answers "not found in the documents" when the answer isn't there. In testing, it took 67 seconds to write and answered all 4 questions correctly.

AI Code With · 10/15/20265 min
Check npm Packages for Vulnerabilities Before Installing: Run npm audit + OSV with Claude Code
Tutorials

Check npm Packages for Vulnerabilities Before Installing: Run npm audit + OSV with Claude Code

Before installing npm packages, have Claude Code generate only a lockfile, run npm audit, and cross-check with OSV to get safe versions and dependency syntax. In our test, 4 packages turned up 30 vulnerabilities, both sources agreed, and after switching versions npm audit reported found 0 vulnerabilities.

AI Code With · 10/15/20266 min
Use Claude Code to Search and Deep-Read arXiv Papers: Chinese Notes from a Single Prompt
Tutorials

Use Claude Code to Search and Deep-Read arXiv Papers: Chinese Notes from a Single Prompt

Use Claude Code to call the official arXiv API for the latest papers, pick 3 to read closely, and automatically write Chinese notes with links. In our test, one run took 281 seconds, and all 5 paper links opened.

AI Code With · 10/15/20266 min
Making Slides with Claude: Generate an Editable .pptx from One Sentence
Tutorials

Making Slides with Claude: Generate an Editable .pptx from One Sentence

Use Claude Code with Anthropic's official pptx Skill to generate an editable .pptx from one sentence, then revise it in plain language. In our test, one sentence produced 6 slides, one revision made it 7, and every slide had speaker notes.

AI Code With · 10/15/20266 min
What Is Claude Code Idle Compaction? How to Turn It Off, Adjust the Threshold, and What Third-Party API Users Should Know
Tutorials

What Is Claude Code Idle Compaction? How to Turn It Off, Adjust the Threshold, and What Third-Party API Users Should Know

When a session has been idle for about 54 minutes and the context has hit 200K tokens, Claude Code automatically compacts it before the 1-hour cache expires. This post covers the trigger conditions, how to turn it off, how to change the threshold, and how API and third-party users can set a 1-hour cache.

AI Code With · 10/15/20267 min
Skip n8n: One cron Job and codex exec Produce a Daily GitHub Trending Digest in Chinese at 8 AM
Tutorials

Skip n8n: One cron Job and codex exec Produce a Daily GitHub Trending Digest in Chinese at 8 AM

79 lines of bash: curl pulls GitHub Trending and each README, codex exec writes the Chinese commentary, and cron runs it at 8 AM every day. Measured at about 43 seconds, with the full script and the pitfalls I hit.

AI Code With · 10/15/202617 min
Using Codex to Automatically Fix a Real Bug and Verify It by Running Tests
Tutorials

Using Codex to Automatically Fix a Real Bug and Verify It by Running Tests

Using codex exec to fix real issue #239 in the open-source library humanize: write a failing test first, then let Codex find the root cause. It finished in 246 seconds, and the source change matches upstream PR #240. Includes config, prompts, a usage breakdown, and the pitfalls I hit.

AI Code With · 10/15/20265 min
How to Write CLAUDE.md for Claude Code: Turn Project Rules into Instructions It Can Act On
Tutorials

How to Write CLAUDE.md for Claude Code: Turn Project Rules into Instructions It Can Act On

Spell out project commands, code boundaries, and verification standards, check whether CLAUDE.md is loaded, and handle rule conflicts and path scoping.

AI Code With · 10/10/20267 min
How to Connect Claude Code to a Third-Party API: Base URL, API Key, and Model Configuration Guide
Tutorials

How to Connect Claude Code to a Third-Party API: Base URL, API Key, and Model Configuration Guide

Configure the third-party API address, credentials, and model for Claude Code, then verify in the terminal and check the call records.

AI Code With · 10/10/20267 min
Troubleshooting Haiku 5.5 API Errors: 400, thinking, and Response Parsing
Tutorials

Troubleshooting Haiku 5.5 API Errors: 400, thinking, and Response Parsing

Locate Haiku 5.5 API problems by HTTP status, request parameters, and response blocks, with complete fix examples.

AI Code With · 10/10/20268 min
How Do You Configure a Haiku 5.5 Subagent in Claude Code?
Tutorials

How Do You Configure a Haiku 5.5 Subagent in Claude Code?

Set up a Haiku 5.5 subagent, tell default, inherited, and forced modes apart, and verify which model actually ran.

AI Code With · 10/10/20267 min
What Changed in DeepSeek V4.1 Flash 1017? Claude Code / Codex Setup, and Web Search Billing Starting November 1
Tutorials

What Changed in DeepSeek V4.1 Flash 1017? Claude Code / Codex Setup, and Web Search Billing Starting November 1

DeepSeek V4.1 Flash was upgraded to DeepSeek-V4.1-Flash-1017 on October 17. The API model name is still deepseek-flash, and the calling method and pricing are unchanged. This post covers the officially announced updates, how to set up Claude Code and Codex, whether the per-request Search/Fetch charges starting November 1 affect you, and a hands-on test of fixing a bug with deepseek-flash in Claude Code.

AI Code With · 10/09/20267 min
How Do You Use Codex in VS Code? A Complete Guide to Windows Installation, Login, Context, Diffs, and Proxy Troubleshooting
Tutorials

How Do You Use Codex in VS Code? A Complete Guide to Windows Installation, Login, Context, Diffs, and Proxy Troubleshooting

From VS Code installation and login to context selection and Diff review, this guide walks through the full Codex IDE workflow, and adds a layered approach to troubleshooting Windows networking, proxies, Providers, and common errors.

AI Code With · 09/03/20269 min
How to Configure Multiple Codex Tools Without Crossing Wires: A Setup and Verification Guide for CC Switch, OpenCode, and WorkBuddy
Tutorials

How to Configure Multiple Codex Tools Without Crossing Wires: A Setup and Verification Guide for CC Switch, OpenCode, and WorkBuddy

How to keep Provider, Endpoint, and Key separate when you use Codex alongside CC Switch, OpenCode, and WorkBuddy, and how to use call records to keep your configs from crossing wires.

AI Code With · 08/26/202612 min
Getting Started with the Codex SDK: TypeScript Calls, Streaming Events, and Thread Resumption
Tutorials

Getting Started with the Codex SDK: TypeScript Calls, Streaming Events, and Thread Resumption

Use @openai/codex-sdk to run a read-only task, handle streaming events, and resume threads. Covers Node.js requirements, the Git working directory, authentication, and troubleshooting failures.

AI Code With · 08/25/20265 min