> ## Documentation Index > Fetch the complete documentation index at: https://agents.craft.do/docs/llms.txt > Use this file to discover all available pages before exploring further. # Introduction > AI Agent's coding and computer use capabilities, extended for your entire digital workflow Craft Agents brings Claude's coding capabilities to your entire digital workflow. It massively improves both coding and non-coding workflows via enhanced UI, review, and auto execution. Built on the same foundation as Claude Code (the [Claude Agent SDK](https://github.com/anthropics/claude-agent-sdk)), it extends the model's reasoning power to work across all your tools: GitHub, Linear, Slack, Craft documents, local files, and any API you can connect. ## What Makes Craft Agents Different ### Familiar interface The inbox works like a combination of email and task manager. Each conversation has a status, persists between sessions, and can be organized however you want. **Review agent plans before execution** in document-like interface, just like reviewing work from a colleague. ### Connect to anything **Sources** give the agent access to your tools—MCP servers, REST APIs, or local files: | Source Type | Examples | What You Can Do | | ----------- | --------------------------- | ------------------------------------------------- | | MCP servers | GitHub, Linear, Craft | Full integration with issue tracking, docs, repos | | REST APIs | Any service with an API | Query and mutate data from internal tools | | Local files | Obsidian vaults, code repos | Work with files on your machine | A single conversation might pull issues from Linear, reference code from GitHub, check Slack threads, and summarize everything in a Craft document. ### Multitask naturally Start multiple tasks. Jump between conversations with full history preserved. Run long tasks in the background while you work on something else. ### Explore, then execute Start in **Explore mode**—read-only by default. Let the agent research, analyze, and draft a plan. Review proposals before any changes happen. Once aligned, switch to **Execute** and let it work autonomously. ### Sessions become documentation Every conversation captures decisions, trade-offs, and implementation details. Share sessions with your team. Attach them to Linear issues, GitHub PRs, or Confluence pages so others understand not just what changed, but why. ### Work from anywhere Run a [remote server](/server/headless) and access your workspaces from any machine — desktop app, browser, or CLI. Transfer sessions between local and remote workspaces. Deploy a standalone server on Linux or macOS with a single install script. ### Fully customizable Themes, skills, statuses, permissions—everything is a config file the agent can edit. Modify behavior through conversation, not complex settings UIs. | What | How It Works | Customization Level | | --------------------------------------------- | --------------------------- | ------------------------------------------------------- | | [**Skills**](/skills/overview) | YAML frontmatter + markdown | Reusable instructions invoked with `@mention` | | [**Themes**](/go-further/themes) | 6-color semantic system | App or workspace, with dark mode and scenic backgrounds | | [**Statuses**](/statuses/overview) | Config for workflow states | Custom labels, colors, icons, inbox/archive categories | | [**Permissions**](/core-concepts/permissions) | Layered rules | App → workspace → source cascade | ## Craft Agents vs Claude Code Craft Agents and Claude Code share the same foundation—the [Claude Agent SDK](https://github.com/anthropics/claude-agent-sdk)—but serve different purposes: | Aspect | Claude Code | Craft Agents | | ----------------- | ----------------------- | ------------------------------------------------------------------------ | | **Interface** | Terminal CLI | Easy to use desktop app | | **Data sources** | MCP servers | MCP + REST APIs + local files | | **Session org** | Continue with `-c` flag | Inbox with custom workflow statuses | | **Config scope** | Per-project `.claude/` | Multi-workspace with full isolation | | **Permissions** | Fixed mode behaviors | [Customizable rules](/core-concepts/permissions) per workspace/source | | **Remote access** | SSH + terminal | [Remote server](/server/headless) with desktop, browser, and CLI clients | **What's the same:** Claude model, core tools (Read, Write, Edit, Bash, Glob, Grep), MCP support, skills format, and thinking level control (Off / Low / Medium / High / Max). ## Key Concepts [**Sources**](/sources/overview) connect Craft Agents to external data—GitHub repos, Linear projects, Craft spaces, or your local filesystem. Each source gives the agent new capabilities. [**Workspaces**](/go-further/workspaces) are isolated environments with their own sources, skills, and sessions. Use different workspaces for different projects or contexts. [**Permission modes**](/core-concepts/permissions) control what the agent can do. Cycle through them with **SHIFT+TAB**: | Mode | Behavior | | --------------- | --------------------------------------------- | | **Explore** | Read-only. Research and analyze without risk. | | **Ask to Edit** | Prompts before changes. Review each action. | | **Execute** | Full autonomy. The agent works uninterrupted. | [**Sessions**](/core-concepts/conversations) are persistent conversations organized like an inbox. Start multiple tasks, switch between them, and pick up where you left off. [**Skills**](/skills/overview) are custom instructions that define specialized agent behaviors—like a code review workflow or a content generation template. Same format as Claude Code skills. ## Open Source Craft Agents is open source under Apache 2.0. The codebase lives at [github.com/lukilabs/craft-agents-oss](https://github.com/lukilabs/craft-agents-oss). Fork it, extend it, make it yours. ## Next Steps Desktop app for macOS, Windows, and Linux. Connect GitHub, Linear, APIs, and local files.