Kanaries · Knowledge Base
Practical tutorials, references, and fixes in this topic.
Claude Code mods (v2.1.287+) are plugins of TypeScript functions that draw panes, add instant slash commands and guard tool calls. Build and test three working mods, browse 15 community mods, and run the safety checks before you install one.
Install Pi 1.0 with one command, sign in with a subscription or API key, or point it at a local Ollama model. Tested on macOS with a local 8B model: real timings, a read-only mode, AGENTS.md loading, and the safety defaults that differ from Claude Code and Codex.
Claude Code v2.1.277 reads AGENTS.md when a project has no CLAUDE.md. Tested on 2.1.278: which file wins, why the first session ignores it, how CLAUDE.local.md blocks it, how to switch the Project instructions mode, and what Codex, OpenCode, and Cursor read.
Install DeepSeek Harness (dsh) with one command: npx @deepseek-ai/dsh web. Then connect a model key, pick a workspace, and run your first agent session. Includes the four runtime modes explained, a plugin starter pack, and what the official docs don't cover yet.
Run claude --permission-mode bypassPermissions in the CLI, or turn on Allow bypass permissions mode under Settings, Claude Code in Desktop. Includes a permission mode comparison table and fixes for 'Permission mode couldn't be changed'.
Install Codex CLI with npm i -g @openai/codex, then use the 5 tips that make it reliable: plan before implementing, set scope and non-goals, move repo rules into AGENTS.md, delegate to subagents, and always verify. Plus CLI vs App vs IDE and large-repo best practices.
Oh My OpenCode is an orchestration harness on top of OpenCode, not a replacement. Install OpenCode first, run /init, then add the harness with bunx oh-my-opencode install. Includes fixes for opencode not found and stale OpenAgent commands.
Learn how Claude Code routines work, what triggers they support, how they differ from Codex app Automations and OpenClaw cron jobs, and why AI agent cron jobs matter beyond chat-first workflows.
Learn how parallel code agents use worktrees and sandboxes, when executable feedback loops matter, and which desktop, IDE, cloud, or loop-native pattern fits.
Step by step tutorial of build a claude code agent by your self with claude code sdk in typescript.
Cursor has becaome a new popstar in the AI-assisted coding domain. But what exactly does it offer? Are people really using it or just intrigued by the hype? Let's dive deep into Cursor's capabilities, limitations, and potential impact on the software development ecosystem.