Package index

Agents

Create an agent, run a task, and inspect what happened.

Agent

Agent that runs tasks with tools

AgentResult()

Create an agent run result

result_is_success()

Check whether a run completed

result_n_turns()

Count the turns in a result

result_text_chunks()

Get the streamed text chunks from a result

result_tool_calls()

Get the tool calls from a result

result_tool_results()

Get the finished tool calls from a result

AgentEvent()

Create an agent event

AgentUsage()

Create a usage record

UsageLimits()

Set usage limits for a run

Conversations and context

Keep long conversations within the model’s context window.

ContextPolicy()

Configure automatic context management

DeputyCompaction()

Create a compaction result

Built-in tools

Tools for files, data, the web and code, and presets that bundle them.

tools_preset()

Get a tool preset by name

tools_file()

Get the basic file tools

tools_code()

Get the code execution tools

tools_data()

Get the data reading tools

tools_web()

Get web tools

tools_all()

Get all built-in tools

tool_read_file()

Read file contents

tool_read_markdown()

Convert a file to Markdown using MarkItDown

tool_write_file()

Write content to a file

tool_edit_file()

Edit file contents by replacing text

tool_multi_edit()

Apply multiple text edits to a file

tool_list_files()

List files in a directory

tool_glob_files()

Find files using a glob pattern

tool_grep_files()

Search file contents with grep-like matching

tool_read_csv()

Read a CSV file

tool_run_r_code()

Execute R code

tool_run_bash()

Execute bash commands

tool_web_fetch()

Fetch web page content

tool_web_search()

Search the web

tool_metadata()

Inspect a tool's origin and annotations

Running code

Persistent R sessions and OS-sandboxed interpreters.

RSession

Persistent R session for a conversation

tools_mcp_repl()

Load sandboxed R tools from mcp-repl

mcp_repl_connection()

Connect an agent to a sandboxed mcp-repl session

mcp_repl_control()

Interrupt or reset an mcp-repl session

mcp_console_connection()

Connect an agent to a sandboxed MCP Console session

mcp_console_control()

Interrupt or restart an MCP Console session

MCP servers

Use tools, resources and prompts from Model Context Protocol servers.

tools_mcp()

Get tools from MCP servers

McpConnection

MCP server connection

Permissions

Decide which tool calls an agent may make.

Permissions()

Create a permission policy

permissions_standard()

Create a standard permission policy

permissions_readonly()

Create a read-only permission policy

permissions_plan()

Create a planning permission policy

permissions_full()

Create a full access permission policy

PermissionMode

Permission modes

permissions_check()

Check a tool call against a permission policy

PermissionResultAllow()

Create an allow permission result

PermissionResultDeny()

Create a deny permission result

PermissionResultPending()

Pause a tool call for approval

Human input and approvals

Ask a person questions, and hold tool calls until someone approves them.

tools_interactive()

Create an ask_user tool with its own handler

tool_ask_user()

Ask the user questions

AskUserDeferred()

Defer answers to a later user turn

set_ask_user_callback()

Set a process-wide handler for ask_user

approval_read()

Read a pending approval

ApprovalContinuation()

Pending approval record

approval_review_ui() approval_review_server()

Review a pending tool call in Shiny

tool_input_review()

Tabulate a tool call's arguments for review

Trusted results

Make one tool the only source of each kind of result.

TrustedResults()

Designate trusted tools

result_trusted_results()

Get the trusted results from a run

Hooks

Run R code at fixed points in an agent’s lifecycle.

HookEvent

Hook events

HookMatcher()

Create a hook

hook_matches()

Test whether a hook matches a tool name

hook_log_tools()

Create a hook that logs tool calls

hook_block_dangerous_bash()

Create a hook that blocks dangerous bash commands

hook_limit_file_writes()

Create a hook that limits file writes to a directory

HookResultPreToolUse()

Create a PreToolUse hook result

HookResultPostToolUse()

Create a PostToolUse hook result

HookResultPreCompact()

Create a PreCompact hook result

HookResult() PermissionResult()

Hook and permission results

Skills

Package instructions and tools for reuse.

Skill()

Reusable instructions and tools

skill_create()

Create a skill in code

skill_load()

Load a skill from a directory

skills_list()

List available skills in a directory

skill_check_requirements()

Check a skill's requirements

Subagents

Delegate parts of a task from a lead agent to specialists.

LeadAgent

Agent that delegates to subagents

AgentDefinition() agent_definition()

Define a subagent

agent_definition_read() agent_definition_write() agent_definitions()

Read and write agent definition files

DelegationInput()

Describe a task for a subagent

DelegationPolicy()

Configure subagent tools, hooks and user input

DelegationResources()

Return tools from a resource factory

DelegationManifest()

Record what a subagent started with

DelegationOutcome()

Result of a delegation

DelegationDisclosure()

Control who can inspect subagents

delegation_history()

Replay saved subagent history

DelegationObservation()

Set limits for the subagent event buffer

DelegationSubscription

Subscription to subagent events

subagent_chat_ui() subagent_chat_server()

Show subagent conversations in a Shiny panel

Retained specialists

Keep a specialist’s conversation across several delegations.

adopt_chat()

Turn an ellmer chat into a retained agent

delegation_tool()

Create a tool that calls a retained agent

ContextFork()

Capture conversation turns for a fork

fork_agent()

Start a retained agent from a conversation fork

Background jobs

Queue work in one R process and run it in another.

job_create()

Create a background job

job_run()

Run a background job

job_read()

Read a background job

job_cancel()

Cancel a background job

AgentJob()

Background job record

Errors

Catch Deputy errors by class.

deputy-errors DeputyError

Deputy error classes

is_deputy_error()

Check whether a condition is a Deputy error