Claude Code reference

Claude Code slash commands: reference and custom commands The built-in commands that matter, the bundled skills hiding among them, and how to add your own in a markdown file. 11 min Claude Code plan mode: how to use it and when it pays off Read-only research first, edits only after you approve. The single best habit for work that matters. 9 min CLAUDE.md: the highest-leverage file in your repository Every location, the load order that surprises people, the 200-line target, and a complete example you can adapt. 8 min MCP servers in Claude Code: setup, scopes, and security The three scopes, every transport, the timeouts nobody documents in one place, and the security question people skip. 9 min Claude Code subagents: when they help and when they cost you The built-ins you already use without knowing, every frontmatter field, and the isolation cost that makes delegation a bad default. 9 min Claude Code hooks: deterministic control over an agent The events worth wiring, the exit code that blocks a tool call, the JSON contract, and two hooks every repo should have. 9 min Claude Code settings.json: the complete reference Every file location, the precedence chain that decides conflicts, how arrays merge, and a complete team config. 6 min Claude Code permissions: what to allow and what to never allow Six modes, a deny list every repo should copy, the path syntax that trips people up, and an honest account of the limits. 8 min Claude Code headless: scripting, CI, and automation Print mode, structured output, bare mode, both CI authentication routes, and the guard rails an unattended agent needs. 8 min Claude Code custom commands: turning a good prompt into a tool Where they live now that commands are skills, the argument syntax that changed, and four commands worth copying into your repo today. 9 min Claude Code context window: compaction, clearing, and what fills it What actually fills it, compact against clear, what survives a compaction, and how to move the auto-compact threshold. 8 min Choosing a Claude Code model: Opus, Sonnet, or Haiku The August 2026 lineup with real rates, every alias and what it resolves to, and a routing table you can apply per task. 8 min Claude Code best practices that survive contact with real work Eight habits ranked by what they are worth, the verification ladder that lets a run finish without you, and the five failure modes that waste the most time. 9 min Claude Code skills: packaged expertise the agent loads on demand The SKILL.md format, the full frontmatter table, why progressive disclosure is the whole point, and the three reasons a skill never fires. 8 min Claude Code in GitHub Actions: the safe patterns The official action and the raw CLI, both auth paths, the trigger that will get you compromised, and how to keep the bill bounded. 8 min The Claude Agent SDK: building on the Claude Code loop The exact packages, a working agent in both languages, the options that matter, and the four things that surprise people in production. 8 min Extended thinking and reasoning effort: when to spend it Every level per model, the five places you can set it and which one wins, the ultrathink keyword, and the four tasks that justify the top of the dial. 8 min Claude Code plugins: install, compare, and build A working map of plugins, marketplaces, skills, slash commands, and subagents, with the exact install commands and no guessed package names. 6 min Claude Code session management: resume and restore The full session reference: resume flags, picker shortcuts, session IDs, JSONL transcripts, forks, checkpoints, and cross-device handoff. 7 min Claude Code cheat sheet: commands, flags, shortcuts, and config The complete one-page Claude Code reference: launch flags, in-session commands, shortcuts, configuration, hooks, MCP, worktrees, and cost controls. 13 min Use Continuum’s hosted inference in Claude Code Four environment variables run Claude Code on your Continuum plan. Two of them are the ones people get wrong. 8 min Use Continuum’s hosted inference in the Claude desktop app The desktop app has no custom-endpoint setting, but the same binary has a third-party inference mode that does — and Continuum speaks the wire it wants. 9 min Use Continuum’s hosted inference in T3 Code T3 Code drives CLIs rather than calling models itself, so you point the CLI, not T3 Code — through a provider’s environment-variable pane. 8 min Use Continuum’s hosted inference in OpenCode One provider block in opencode.json puts every model on your Continuum plan into OpenCode’s model picker — as long as you list each one. 7 min Can you run Continuum’s models in Devin? The answer is no, at every tier including Enterprise, and Cognition says so in one sentence. What that leaves you is a model selector, not an endpoint. 7 min Use Continuum’s hosted inference with the ChatGPT desktop app’s Codex The engine reads your config; the GUI fights you. A profile in the shared config home is the part that genuinely works today. 10 min Free AI coding tools: what costs $0, and what moves the bill A zero-dollar stack for completion, code generation, terminal agents, and multi-agent work, with every hidden model, quota, and hardware cost named. 13 min Devin pricing: plans, quotas, ACUs, and the real cost The current plan ladder, the quota system that replaced public ACU billing, and a defensible way to price autonomous work. 12 min Devin Review: what it reviews, what it misses, and whether Devin is worth it Two intents in one name: Cognition’s pull-request reviewer, and an honest assessment of the autonomous engineer behind it. 12 min Cognition AI: the company behind Devin and the former Windsurf The company, the autonomous engineer, the acquired editor, and how two product lines became one Devin system. 12 min Windsurf pricing: what Devin Desktop costs now Windsurf is billed as Devin Desktop now. Here is the current plan ladder, what free means, and why older credit prices are stale. 12 min Windsurf IDE is Devin Desktop now: what it is and where it goes The editor still exists. Its name, bundled agent, billing home, and strategic role changed. Here is the current product without pretending the future is settled. 13 min Factory AI review: where Droid wins and what to test Factory reviewed as an agent platform: strong harness design, broad execution surfaces, persistent compute, real enterprise controls, and material platform overhead. 12 min Factory AI pricing: Pro, Plus, Max, and team costs The current Factory plan ladder, rolling usage windows, Droid Core fallback, prepaid overflow, computer entitlement, and team procurement math. 12 min Factory Droid CLI: install, commands, and safe automation A working install path plus the command surface that matters: sessions, spec mode, autonomy, JSON output, worktrees, Missions, MCP, and computers. 11 min Switch from Factory AI to Continuum without losing work A reversible migration from Droids to a free workbench over Claude Code, Codex, Cursor agent, and peers, with the reasons to stay on Factory stated plainly. 11 min Qodo Merge: what the PR review product became The legacy PR review product, its slash-command workflow, how it differed from PR-Agent, and what moved into Qodo Git and Qodo 2.x. 11 min Qodo PR-Agent: the open-source reviewer, clearly separated The MIT reviewer that began at Qodo, now community-owned: how to install it, run the core commands, choose a model, and operate it safely. 11 min