Agent
A Claude Code tool, read out of the shipped bundle. It has been in the build since v2.1.138, including the newest one read.
Subagent launching gets a nesting-depth cap, a bypassable concurrency cap, and new teammate/model-inherit rules
Not Anthropic's. This is the line from the earliest changelog entry here that named it, v2.1.269.
In the changelogs
22Releases whose published page names Agent.
-
v2.1.307
A different check now decides who sees the 'Do not spawn agents' guidance
The Agent tool's guidance against starting helper agents no longer depends on being on the Pro plan, and its 'when to use' text is reorganised
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the security review prompt
The instructions telling Claude when to hand work to a helper agent were reworded in the security review prompt, including a warning about token cost
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the team onboarding prompt
The instructions on when Claude hands work to a helper agent were reworded in the team onboarding prompt, noting you can only wait or cancel
found in this entry's text
-
v2.1.307
Agent tool guidance reworded in the prompt after a hook blocks a tool
Claude's instructions for when to use a helper agent were reworded in the blocked-tool reminder prompt, now mentioning side quests
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the repeated-file-read reminder prompt
Claude's helper-agent instructions were rewritten in the repeated-read reminder prompt, telling it to do small jobs itself and not spawn when in doubt
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the empty-file reminder prompt
In the empty-file reminder prompt, Claude's helper-agent instructions now say to do short tasks and known lookups itself
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the file-offset reminder prompt
In the file-offset reminder prompt, Claude's helper-agent instructions now say to delegate reviews only for an independent view, giving code, not conclusions
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the skill-run reminder prompt
In the skill-run reminder prompt, Claude's helper-agent instructions now say to wait for delegated work instead of also doing it
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the to-do reminder prompt
In the to-do reminder prompt, Claude's helper-agent instructions now end with a default: when in doubt, don't start a helper
found in this entry's text
-
v2.1.307
Agent tool guidance rewritten in the general-purpose subagent prompt
In the general-purpose subagent prompt, the Agent tool text now tells Claude how to brief a helper: goal, what's ruled out, files to read, narrow scope
found in this entry's text
-
v2.1.299
Plugin evals can now take tools away from Claude with --disallow-tools
The plugin eval runner accepts
--disallow-toolsand a per-casedisallowed_toolslist, and scores runs that end on an API error as 0found in this entry's text
-
v2.1.296
Subagents can be asked to think one step harder or lighter than the main session
The Agent tool can take an effort option of "lower" or "higher", which moves a subagent's reasoning effort one step from the current level
found in this entry's text
-
v2.1.296
SendMessage explains who a message can go to
The recipient field of
SendMessagenow explains its allowed values, and theAgenttool gains a step that adjusts its inputfound in this entry's text
-
v2.1.293
Clearer guidance for Claude when it asks for a tool it cannot use
When Claude calls a tool that is not available, the error now says whether it is disabled or simply not offered here, with special advice for
Agentfound in this entry's text
-
v2.1.293
Background agent notices stop suggesting a message Claude cannot send
Background agent notices now tell Claude to message the agent only when it is able to; otherwise they just say it will be notified
found in this entry's text
-
v2.1.283
Subagent token totals now skip unbilled final messages
Token usage reported when a subagent finishes now comes from its last billed reply rather than whatever message came last
found in this entry's text
-
v2.1.271
Task/Agent tool description gains a 'delegation costs first' mode
Task/Agent tool's built-in instructions can now include a 'delegation costs first' framing, depending on the build
found in this entry's text
-
v2.1.269
Subagent/Task launch pipeline (Agent tool) substantially reworked with new caps and teammate rules
Subagent launching gets a nesting-depth cap, a bypassable concurrency cap, and new teammate/model-inherit rules
found in this entry's text
-
v2.1.178
Named-Teammate Agent Swarms (TeamCreate/TeamDelete Removed)
found in this entry's text
- v2.1.178
- v2.1.63
-
v2.1.63
Task Tool Renamed to Agent
found in this entry's text
First cited
0No release's published evidence quotes this name. The ledger indexes the Evidence lines of every changelog on this site, so this says nobody here has ever quoted it; it says nothing about how old the name is.
Presence across releases
2Build by build
164One row per release since the first build this name was read out of. Absent means the build was read and the name was not in it, never mined means no bundle for that release was ever archived, and a declared type or a default is only ever what that release's own bundle stated.
Read out of the published npm bundle release by release, and out of Anthropic's own documentation as this site captured it. Nothing on this page is a description anybody here wrote about what the tool does. All tools.