Tools Overview

Claude Code Built-in Tools

Per-section reference for every built-in tool dispatched by Claude Code.

Table of Contents

1. How tools work
2. Filesystem tools
3. Shell & code execution tools
4. Network & web tools
5. Orchestration & agent tools
6. Scheduling & tasks
7. Session & user interaction tools
8. MCP (Model Context Protocol) tools
9. Meta & search tools
10. Remote & entrypoint-specific tools
11. Tools present in source but not in public docs
12. Feature-flag-gated tools summary
13. Discrepancies & notes
14. MCP tools — dynamic registration
15. Permissions & tool access control
16. Related references

Overview

16 sections, 44 entries across tool catalog, narrative, and reference subsections.

#	Section	Description	Entries
1	How tools work	How Claude dispatches tool calls, permissions, entrypoint gating, and MCP vs built-in distinction.	0
2	Filesystem tools	Read, Write, Edit, NotebookEdit, Glob, Grep — the file I/O and search primitives.	6
3	Shell & code execution tools	Bash, PowerShell, LSP, Monitor — running commands and interacting with code intelligence.	4
4	Network & web tools	WebFetch and WebSearch for retrieving external content.	2
5	Orchestration & agent tools	Agent, Skill, SendMessage, TeamCreate/Delete — spawning and coordinating sub-agents.	5
6	Scheduling & tasks	CronCreate/Delete/List plus TodoWrite and the Task* family for work tracking and cron-style scheduling.	10
7	Session & user interaction tools	EnterPlanMode, ExitPlanMode, EnterWorktree, ExitWorktree, AskUserQuestion, Config — session-state primitives.	6
8	MCP (Model Context Protocol) tools	ListMcpResources and ReadMcpResource — primitives for reading MCP server resources.	2
9	Meta & search tools	ToolSearch, BriefTool, SleepTool — meta operations on the tool catalog itself.	3
10	Remote & entrypoint-specific tools	RemoteTrigger and other tools available only in specific entrypoints (SDK, remote).	1
11	Tools present in source but not in public docs	Internal / experimental tools from the source tree that aren’t on the public tools-reference page.	5
12	Feature-flag-gated tools summary	Tools gated by ENV-level feature flags, mapped to their enabling env vars.	0
13	Discrepancies & notes	Known mismatches between public docs, source snapshot, and runtime behavior.	0
14	MCP tools — dynamic registration	How MCP servers register tools dynamically as mcp__<server>__<tool>.	0
15	Permissions & tool access control	Allow/ask/deny rules, the permission model, and per-tool gating via settings.json.	0
16	Related references	Cross-links to other docs in this repo (Commands, Skills, ENV, Settings).	0

Quick reference — built-in tools

Tool	Class	Side effect	Gating	Purpose
Read	Filesystem	Read-only	Always	Read file contents, including images and notebooks
Write	Filesystem	Writes	Permission	Create or overwrite files
Edit	Filesystem	Writes	Permission	Make targeted edits to specific lines
NotebookEdit	Filesystem	Writes	Permission	Modify Jupyter notebook cells
Glob	Filesystem	Read-only	Always	Find files by glob pattern matching
Grep	Filesystem	Read-only	Always	Search for patterns in file contents
Bash	Shell	Executes	Permission	Run shell commands (POSIX/Bash)
PowerShell	Shell	Executes	Permission	Run PowerShell commands (opt-in)
LSP	Code intelligence	Read-only	Always	Jump to definitions, find references, type info
WebFetch	Network	Read-only	Permission	Fetch and parse content from URLs
WebSearch	Network	Read-only	Permission	Perform web searches (US only)
Agent	Orchestration	Spawns	Always	Spawn a subagent with isolated context
Skill	Orchestration	Spawns	Permission	Execute a skill from the skill registry
CronCreate	Scheduling	Spawns	Always	Schedule a one-time or recurring prompt
CronDelete	Scheduling	Spawns	Always	Cancel a scheduled task
CronList	Scheduling	Read-only	Always	List all scheduled tasks
TaskCreate	Task management	Spawns	Always	Create a task in the task list
TaskGet	Task management	Read-only	Always	Get details for a specific task
TaskList	Task management	Read-only	Always	List all tasks with status
TaskUpdate	Task management	Writes	Always	Update task status, details, or delete
TaskStop	Task management	Executes	Always	Kill a running background task
TaskOutput	Task management	Read-only	Always	(Deprecated) Get output from a background task
EnterPlanMode	Session	None	Always	Switch to plan mode (design before code)
ExitPlanMode	Session	Writes	Permission	Present plan and exit plan mode
EnterWorktree	Session	Spawns	Always	Create or switch to a git worktree
ExitWorktree	Session	None	Always	Exit a worktree and return to main
Monitor	Shell	Executes	Permission	Run command in background, feed output back
SendMessage	Orchestration	None	Always	Message a teammate or resume a subagent
TeamCreate	Orchestration	Spawns	Feature-flag	Create an agent team with multiple teammates
TeamDelete	Orchestration	Executes	Feature-flag	Disband a team and clean up resources
AskUserQuestion	Session	None	Always	Ask multiple-choice questions
Config	Session	Writes	Always	Get or modify settings
ToolSearch	Meta	Read-only	Feature-flag	Search for and load deferred tools
ListMcpResources	MCP	Read-only	Always	List resources exposed by MCP servers
ReadMcpResource	MCP	Read-only	Always	Read a specific MCP resource by URI
TodoWrite	Task management	Writes	Always	Manage session task checklist (non-interactive)
RemoteTrigger	Remote	Executes	Entrypoint	Trigger a remote session or API call
BriefTool	Session	None	Feature-flag	Enable brief/compact output mode
SleepTool	Meta	None	Internal	Sleep for a duration (testing/internal)

See Also

• ../Commands/README.md — companion reference for built-in slash commands.
• ../Skills/README.md — companion reference for bundled skills.
• ../Skills/FRONTMATTER.md — frontmatter schema for skills, agents, and commands.
• ../ENV/README.md — environment variables (many tools are feature-flag gated here).
• Official docs: https://code.claude.com/docs/en/tools-reference