The fenbs blogAll articles
Page 15 of 48
- RACI Matrix: Meaning, Example, and Where AI Agents FitRACI stands for Responsible, Accountable, Consulted and Informed: a table that says, for each task or decision, who does it, who answers for it, who is asked and who is told. What each letter means, an example, the rules, and why an AI agent can be Responsible but never Accountable. · 7 min read
- Punch List: Meaning, Template and How to Close One OutA punch list is the list of work still to finish or put right when a building job is nearly complete. What it is, who does what, the fields every item needs, a template you can copy, and a process for closing it out before handover. · 7 min read
- Custom Slash Commands in Claude Code: Templates You Can CopyA custom command is one Markdown file that becomes a /name you can type. Where the files go now that commands have merged into skills, how arguments work, five command files to copy, and when a command should become a skill instead. · 7 min read
- Claude Code Plugins: How to Install Them and Which to TrustInstall Claude Code plugins from Anthropic’s marketplace, a GitHub repository or a folder on disk, manage them from the /plugin panel or the shell, and check what a plugin will run on your machine before you let it. · 8 min read
- Why Claude Code Ignores Your CLAUDE.md, and How to Fix ItEither the file never loaded, or it loaded and lost out to something else. How to tell which in one command, then a fix for each cause: location, start folder, imports, path rules, memory, compaction, and the rules that need a hook instead. · 7 min read
- Claude Code With Local Models: Ollama, LM Studio and OpenRouterClaude Code can be pointed at a model other than Claude by changing one environment variable. How Ollama, LM Studio and OpenRouter document it, what Anthropic does and does not support, what breaks, and what stays private. · 7 min read
- Claude Code in JetBrains IDEs: Setup, Diffs and LimitsIn IntelliJ IDEA, PyCharm, WebStorm and the rest, Claude Code is the CLI in the IDE’s terminal plus a plugin that connects the two. How to install both, how diffs and selections reach the IDE, the settings that matter, the usual failures, and how it differs from the VS Code extension. · 7 min read
- Using Claude Code From Your Phone: Remote Control and the WebThere is no separate Claude Code phone app. The Code tab in the Claude app either steers a session still running on your computer, through Remote Control, or runs one in the cloud. How to set up each, what you can do from the phone, what stays on the desk, and the security model. · 8 min read
- Rolling Out Claude Code to a Team: Settings, Seats and GuardrailsGiving a team Claude Code is a handful of decisions: who gets a seat, which rules the organization enforces, which the repository shares, which servers and plugins are allowed, how you see usage, and where the work is recorded. The order that works, with the settings for each. · 7 min read
- Claude Code Not Working? Fixes for the Common ErrorsWhen Claude Code will not install, will not start or will not sign in, the error text nearly always names the cause. A fix list by symptom: command not found, failed installs, login loops, proxies and VPNs, terminal quirks, old versions, and where the logs are. · 8 min read
- Claude Code MCP Server Failed to Connect: A Fix ListAn MCP server that will not connect in Claude Code nearly always tells you why, if you know where to look. Read the status first, then work through scopes, .mcp.json syntax, project approval, stdio commands, HTTP sign-in, timeouts and the debug log. · 8 min read
- The Claude Agent SDK: What It Gives You Beyond Claude CodeThe Agent SDK is Claude Code’s agent loop as a TypeScript or Python library. What it adds over the CLI, the options that matter, three short examples of different shapes, and when claude -p is the better tool. · 9 min read
- Does Claude Code Read AGENTS.md? Sharing One File With Every AgentYes: current Claude Code reads AGENTS.md on its own, but only when no CLAUDE.md is in the way. Which files block it, the setting that changes that, when to use an @AGENTS.md import or a symlink instead, and a layout that serves Claude Code, Codex, Copilot and Cursor from one file. · 7 min read
- What Claude Code Can Read: PDFs, Images, Spreadsheets and the WebClaude Code reads PDFs, images, notebooks and any text file natively, reaches spreadsheets and Word files through scripts, and searches and fetches the web with your permission. Format by format: how to hand it over, and the limit that applies. · 8 min read
- Claude Code on the Web: Cloud Sessions vs the TerminalClaude Code on the web runs each task in its own cloud VM against a GitHub repository, pushes a branch and keeps going with your laptop shut. How the GitHub connection, environments, setup scripts and network levels work, how to review and open pull requests, how to pull a session into your terminal, and what it cannot do. · 8 min read
- Gemini CLI vs Claude Code vs Codex CLI: The Terminal Agents ComparedThree vendors, three agents that live in your terminal. How Gemini CLI, Claude Code and Codex CLI differ on install, sign-in, instruction files, approvals and sandboxing, MCP, subagents, scripting and open source, and who each one suits. · 7 min read
- Cline, Roo Code and Kilo Code vs Claude CodeCline and Kilo Code are open-source agents that run in your editor with whichever model you bring; Claude Code is Anthropic’s own agent for Claude models. How they differ on models, modes, approvals, checkpoints, rules files and MCP, what happened to Roo Code, and who each suits. · 7 min read
- Kiro vs Claude Code: Spec-First IDE or Terminal AgentKiro, from AWS, is built around specs: requirements, design and tasks written and approved before the code. Claude Code is Anthropic’s agent that starts from your prompt and plans when you ask. How their specs and plan mode, steering and CLAUDE.md, hooks, MCP and autonomy compare, and how to borrow Kiro’s specs in Claude Code. · 7 min read
- OpenCode vs Claude Code: Open-Source Agent or Anthropic’s OwnOpenCode is an MIT-licensed coding agent that runs any of 75-plus model providers from a terminal, desktop app, IDE or browser. Claude Code is Anthropic’s agent for Claude models. How they differ on models, config, agents, permissions, MCP and clients, what the docs say about using Claude in OpenCode, and who each suits. · 7 min read
- Google Antigravity vs Claude Code, and Where Cursor FitsAntigravity is Google’s agent-first platform: a desktop command center for many agents, a terminal client, an IDE and editor extensions, with plans, walkthroughs and browser recordings you review as artifacts. Claude Code is Anthropic’s agent, one engine in a terminal, editors, a desktop app and the web. How they differ on models, parallel agents, review and where the work is recorded, with Cursor in brief. · 8 min read
- Cursor vs Codex: Editor Agent or OpenAI’s Coding AgentCursor is an editor built around its agent, with a CLI and cloud agents. OpenAI Codex is one agent with a CLI, an IDE extension, a view in the ChatGPT desktop app and a cloud. How they compare on instruction files, approvals and sandboxing, pull request review and models, why you can run Codex inside Cursor, and who each suits. · 6 min read
- OpenAI Codex vs GitHub CopilotCodex and Copilot now meet at every level: a cloud agent that returns a pull request, an agent in the editor, a CLI, a desktop app and a pull request reviewer. How they compare at each, which instruction files each reads, which models each runs, and how Codex can run inside Copilot. · 7 min read
- Codex CLI Commands and Config: A ReferenceEvery Codex CLI command worth knowing on one page: subcommands, flags, slash commands grouped by job, keyboard shortcuts, the config.toml keys people change most, and how to exit, update and uninstall. · 7 min read
- Codex CLI Best Practices: AGENTS.md, Approval Modes and ReviewThe habits that make a Codex CLI run something you can check in five minutes: a short AGENTS.md, the right permissions for each task, a plan before code, one task per session, a real review, worktrees for parallel work, secrets kept out, and a record of what was done. · 7 min read
- Codex Cloud: Delegating Tasks to OpenAI’s Cloud AgentCodex cloud runs coding tasks in containers on OpenAI’s side while you do something else. How to connect GitHub, set up an environment, start tasks from the web, your editor, the terminal, GitHub, Slack or Linear, review the result, and keep it safe. · 7 min read
- ChatGPT Work vs Claude Cowork: Two Desktop Agents ComparedChatGPT agent has been replaced by ChatGPT Work, and Claude Cowork is being folded into ordinary Claude conversations. Where each runs, what it can reach, when it stops to ask, how it repeats work, and who each one suits. · 8 min read
- Do ChatGPT Projects Share Memory? How Project Memory WorksWhether one ChatGPT project can see another depends on a single setting. What default and project-only memory let through, how saved memories and chat history fit in, why shared projects are always walled off, and how to check and reset what ChatGPT remembers. · 7 min read
- ChatGPT Apps SDK: How Apps Inside ChatGPT WorkThe ChatGPT Apps SDK now lives in OpenAI’s plugin documentation, but the idea is unchanged: an app is an MCP server, with optional UI that renders in the chat. The architecture, a minimal example, testing in developer mode, sign-in, review for the directory, and how an app differs from a connector and a custom GPT. · 8 min read
- The OpenAI Agents SDK in Practice: Agents, Handoffs and GuardrailsThe OpenAI Agents SDK runs the agent loop for you: an agent calls tools, hands off to another agent and is checked by guardrails, with sessions and tracing built in. Here are the primitives, a working Python example with one tool and one handoff, and when to use the plain Responses API instead. · 7 min read
- OpenAI Agent Builder: What It Builds, Its Shutdown, and When to Use Code InsteadAgent Builder is OpenAI’s visual canvas for agent workflows, and it shuts down on November 30, 2026. What it built, how its workflows were deployed, what its evals and guardrails did, and how to move a workflow to the Agents SDK, ChatGPT workspace agents or a tool such as n8n. · 7 min read
- Lovable, Bolt and v0 vs Claude Code: App Builders or a Coding AgentLovable, Bolt and v0 build and host an app from a chat in the browser. Claude Code works inside a repository you already have. What each one generates, where the code lives, who each suits, how to hand a builder project over to a coding agent, and the security checks a generated app needs. · 7 min read
- Why GitHub Copilot Ignores Your Instructions FileEither Copilot never saw the file, or it saw it and something else won. How to tell which in one click, then a fix for each cause: the wrong path, a setting that is off, the wrong agent in VS Code, an applyTo glob that never matches, conflicting files, a file that is too long or vague, and surfaces that never read it. · 7 min read
- Adding MCP Servers to GitHub Copilot CLIHow to use Copilot CLI with MCP: the built-in GitHub server, adding local and remote servers with /mcp add or copilot mcp add, the config file, OAuth sign-in, tool permissions, fixes by symptom, and what an organization can enforce. · 7 min read
- GitHub Copilot Hooks: Checks Around the AgentCopilot hooks run your own commands when an agent session starts, before and after each tool call, and when the agent stops. Where they work, the file format, three worked hooks, the security rules, and how they differ from Claude Code hooks. · 7 min read
- How to Write Instructions for an AI Agent: A System Prompt TemplateStanding instructions tell an agent what is true for every task; the task card tells it what to do now. What belongs in each, a six-part template to copy, the mistakes that make agents ignore instructions, and how to test a prompt before you trust it. · 7 min read
- Cursor Cloud Agents (Formerly Background Agents): Setup, Limits and ReviewCursor’s background agents are now called cloud agents. How to set up the environment they run in, keep secrets out of the transcript, start one from the editor, Slack, GitHub or the terminal, watch it work, and review what it hands back. · 8 min read
- Cursor Agent vs Ask vs Plan Mode: Which to Use WhenCursor’s agent has four built-in modes: Agent changes things, Ask answers, Plan proposes, and Debug investigates. Choose by how big the task is and what a wrong first attempt would cost, and know what happens to the context when you switch. · 7 min read
- Cursor Rules Examples: .mdc Files for Five StacksFive annotated Cursor rule files, for a Next.js app, a Python API, a monorepo package, writing tests and writing docs, each with the frontmatter for its rule type, plus when a plain AGENTS.md is the better choice and the lines that never belong in a rule. · 8 min read
- Gemini CLI vs Gemini Code Assist: Which Google ToolGemini CLI is Google’s open-source agent for the terminal. Gemini Code Assist is its assistant inside VS Code and JetBrains IDEs, and its agent mode in VS Code runs on Gemini CLI. What each does, how they share settings and quotas, who can still sign in since June 2026, and where Antigravity now fits. · 7 min read
- What Is Kiro? Amazon’s Spec-Driven IDE ExplainedKiro is AWS’s agentic coding environment: an IDE, a CLI, a web app and more, all running one agent that turns a prompt into requirements, a design and a task list before it writes code. What specs, steering files, hooks, MCP and Autopilot are, how to get started, and who it suits. · 7 min read
- What Is OpenCode? The Open-Source Terminal Coding AgentOpenCode is an MIT-licensed coding agent from Anomaly that runs in a terminal, a desktop app, an editor or a browser, with whichever model provider you connect. What it is, how to install it, its Build and Plan agents, config, MCP, session sharing, and a first session step by step. · 6 min read
- Confluence MCP: Connecting AI Assistants to Your DocsAtlassian’s Rovo MCP server gives Claude Code, Cursor, VS Code and ChatGPT the same Confluence tools: search with CQL, read pages, create and update them, and comment. What the tools cover, whose permissions they use, what admins control, the options for Data Center, and what an assistant should never publish without a person. · 7 min read
- Google Drive and Docs MCP: Official Options and Safe SetupThere are four ways to give an AI assistant your Google Drive today: Claude’s Google Workspace connectors, Google’s own Drive and Docs MCP servers, ChatGPT’s Google Drive app, and community servers. What each one can do, which OAuth scopes they ask for, what an admin controls, and how to prompt so nothing gets shared by accident. · 8 min read
- Figma MCP Server: What It Does and How to Set It UpFigma runs its own MCP server, in a remote version and a desktop one. What it gives a coding agent, which seats can use it, how to add it to Claude Code, Cursor and VS Code, how it differs from the Figma REST API, and the prompts and limits that decide how good the output is. · 8 min read
- Sentry MCP: Let Your Coding Agent Read the Error Before It Fixes ItSentry hosts an official MCP server that signs in with OAuth and can be pinned to one project. How to add it to Claude Code, Cursor and VS Code, what its tools and skills allow, how it compares with Sentry’s two command-line tools, and how to make the stack trace the first thing your agent reads. · 7 min read
- VS Code mcp.json: Adding MCP Servers Step by StepThree ways to add an MCP server to VS Code, what an mcp.json entry looks like for a local and a remote server, how to start, stop and pick tools, and which agent harness actually sees the servers you add. · 8 min read
- MCP Client, Host and Server: What Each One DoesThe host is the AI app you use, a client is the connection it opens for each server, and a server offers the tools. What each part is responsible for under the current specification, examples of each, and why the split decides who can stop what. · 7 min read
- The MCP Registry: What It Is and How Clients Use ItThe official MCP Registry is a public catalog of server metadata, not of code. What an entry contains, how names are verified, the API, how to publish, and how GitHub, VS Code and private registries build on it. · 8 min read