Sweep 22 Sep 2026 · 17:19Z Build v2.1.280 501 read Stable v2.1.267 Latest v2.1.280 Next v2.1.280 Feeds RSS JSON llms.txt Unofficial
One change · api

Create Agent changed

api/beta/agents/create

Nearest release: v2.1.268, published under an hour after this site recorded the change. Shown because the two are within 24 hours of each other. Nothing here says the release caused the edit.

Recorded here
Lines+994added
Lines−893removed
From line 1 where the diff opens
First seen 14 Aug 2026 this site's first read of the page
Recorded edits9to this page, all time

The whole hunk

from line 1, old and new numbered
/
lines
from line 1
1---
2title: Create Agent
3url: https://platform.claude.com/docs/en/api/beta/agents/create
4---
5 
16# Create Agent
27 
38**POST** `/v1/agents`
from line 17
1217 
1318 - `string`
1419 
15 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 41 more`
20 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 42 more`
1621 
1722 - `"message-batches-2024-09-24"`
1823 
from line 65
6065 
6166 - `"user-profiles-2026-08-18"`
6267 
68 - `"user-profiles-2026-09-04"`
69 
6370 - `"advisor-tool-2026-03-01"`
6471 
6572 - `"managed-agents-2026-04-01"`
from line 109
102109 
103110 - `"mid-conversation-system-clear-at-2026-08-21"`
104111 
112- `"anthropic-workspace-id": optional string`
113 
105114## Body parameters
106115 
107116- `model: BetaManagedAgentsModel or BetaManagedAgentsModelConfigParams`
from line 273
264273 
265274 MCP servers this agent connects to. Maximum 20. Names must be unique within the array. Every server must be referenced by an `mcp_toolset` in `tools`; unreferenced servers are rejected. See the [MCP connector guide](https://platform.claude.com/docs/en/managed-agents/mcp-connector).
266275 
276 - `type: "url"`
277 
267278 - `name: string`
268279 
269280 Unique name for this server, referenced by mcp_toolset configurations. 1-255 characters.
from line 281
270281 
271282 minLength: 1, maxLength: 255
272283 
273 - `type: "url"`
274 
275284 - `url: string`
276285 
277286 Endpoint URL for the MCP server.
from line 295
286295 
287296 A coordinator topology: the session's primary thread orchestrates work by spawning session threads, each running an agent drawn from the `agents` roster.
288297 
298 - `type: "coordinator"`
299 
289300 - `agents: array of BetaManagedAgentsMultiagentRosterEntryParams`
290301 
291302 Agents the coordinator may spawn as session threads. 1–20 entries. Each entry is an agent ID string, a versioned `{"type":"agent","id","version"}` reference, or `{"type":"self"}` to allow recursive self-invocation. Entries must reference distinct agents (after resolving `self` and string forms); at most one `self`. Referenced agents must exist, must not be archived, and must not themselves have `multiagent` set (depth limit 1).
from line 307
296307 
297308 Specification for an Agent. Provide a specific `version` or use the short-form `agent="agent_id"` for the most recent version
298309 
310 - `type: "agent"`
311 
299312 - `id: string`
300313 
301314 The `agent` ID.
from line 315
302315 
303316 minLength: 1, maxLength: 128
304317 
305 - `type: "agent"`
306 
307318 - `version: optional number`
308319 
309320 The specific `agent` version to use. Omit to use the latest version. Must be at least 1 if specified.
from line 331
320331 
321332 Platform advisor roster entry: a model the session's primary thread may consult mid-turn. At most one per roster; the entry occupies the roster name `anthropic.advisor`.
322333 
334 - `type: "advisor"`
335 
323336 - `model: string`
324337 
325338 A Claude model id. The model must be permitted as an advisor for this agent's model — see the sessions/threads/advisor spec.
from line 339
326339 
327340 minLength: 1, maxLength: 256
328341 
329 - `type: "advisor"`
330 
331 - `type: "coordinator"`
332 
333342- `skills: optional array of BetaManagedAgentsSkillParams`
334343 
335344 Skills available to the agent.
from line 347
338347 
339348 An Anthropic-managed skill.
340349 
350 - `type: "anthropic"`
351 
341352 - `skill_id: string`
342353 
343354 Identifier of the Anthropic skill (e.g., "xlsx").
from line 355
344355 
345356 minLength: 1, maxLength: 64
346357 
347 - `type: "anthropic"`
348 
349358 - `version: optional string or null`
350359 
351360 Version to pin. Defaults to latest if omitted.
from line 365
356365 
357366 A user-created custom skill.
358367 
368 - `type: "custom"`
369 
359370 - `skill_id: string`
360371 
361372 Tagged ID of the custom skill (e.g., "skill_01XJ5...").
from line 373
362373 
363374 minLength: 1, maxLength: 64
364375 
365 - `type: "custom"`
366 
367376 - `version: optional string or null`
368377 
369378 Version to pin. Defaults to latest if omitted.
from line 403
394403 
395404 Configuration override for the bash tool.
396405 
406 - `type: optional "bash"`
407 
397408 - `name: "bash"`
398409 
399410 Must be "bash".
from line 413
402413 
403414 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
404415 
405 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
416 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
406417 
407418 Permission policy for tool execution.
408419 
from line 429
418429 
419430 - `type: "always_ask"`
420431 
421 - `type: optional "bash"`
432 - `BetaManagedAgentsAutoPolicy object`
422433 
434 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
435 
436 - `type: "auto"`
437 
423438 - `BetaManagedAgentsEditToolConfigParams object`
424439 
425440 Configuration override for the edit tool.
426441 
442 - `type: optional "edit"`
443 
427444 - `name: "edit"`
428445 
429446 Must be "edit".
from line 449
432449 
433450 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
434451 
435 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
452 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
436453 
437454 Permission policy for tool execution.
438455 
from line 461
444461 
445462 Tool calls require user confirmation before execution.
446463 
447 - `type: optional "edit"`
464 - `BetaManagedAgentsAutoPolicy object`
448465 
466 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
467 
449468 - `BetaManagedAgentsReadToolConfigParams object`
450469 
451470 Configuration override for the read tool.
452471 
472 - `type: optional "read"`
473 
453474 - `name: "read"`
454475 
455476 Must be "read".
from line 479
458479 
459480 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
460481 
461 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
482 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
462483 
463484 Permission policy for tool execution.
464485 
from line 491
470491 
471492 Tool calls require user confirmation before execution.
472493 
473 - `type: optional "read"`
494 - `BetaManagedAgentsAutoPolicy object`
474495 
496 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
497 
475498 - `BetaManagedAgentsWriteToolConfigParams object`
476499 
477500 Configuration override for the write tool.
478501 
502 - `type: optional "write"`
503 
479504 - `name: "write"`
480505 
481506 Must be "write".
from line 509
484509 
485510 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
486511 
487 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
512 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
488513 
489514 Permission policy for tool execution.
490515 
from line 521
496521 
497522 Tool calls require user confirmation before execution.
498523 
499 - `type: optional "write"`
524 - `BetaManagedAgentsAutoPolicy object`
500525 
526 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
527 
501528 - `BetaManagedAgentsGlobToolConfigParams object`
502529 
503530 Configuration override for the glob tool.
504531 
532 - `type: optional "glob"`
533 
505534 - `name: "glob"`
506535 
507536 Must be "glob".
from line 539
510539 
511540 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
512541 
513 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
542 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
514543 
515544 Permission policy for tool execution.
516545 
from line 551
522551 
523552 Tool calls require user confirmation before execution.
524553 
525 - `type: optional "glob"`
554 - `BetaManagedAgentsAutoPolicy object`
526555 
556 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
557 
527558 - `BetaManagedAgentsGrepToolConfigParams object`
528559 
529560 Configuration override for the grep tool.
530561 
562 - `type: optional "grep"`
563 
531564 - `name: "grep"`
532565 
533566 Must be "grep".
from line 569
536569 
537570 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
538571 
539 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
572 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
540573 
541574 Permission policy for tool execution.
542575 
from line 581
548581 
549582 Tool calls require user confirmation before execution.
550583 
551 - `type: optional "grep"`
584 - `BetaManagedAgentsAutoPolicy object`
552585 
586 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
587 
553588 - `BetaManagedAgentsWebFetchToolConfigParams object`
554589 
555590 Configuration override for the web_fetch tool.
556591 
592 - `type: optional "web_fetch"`
593 
557594 - `name: "web_fetch"`
558595 
559596 Must be "web_fetch".
from line 613
576613 
577614 format: int32
578615 
579 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
616 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
580617 
581618 Permission policy for tool execution.
582619 
from line 625
588625 
589626 Tool calls require user confirmation before execution.
590627 
591 - `type: optional "web_fetch"`
628 - `BetaManagedAgentsAutoPolicy object`
592629 
630 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
631 
593632 - `BetaManagedAgentsWebSearchToolConfigParams object`
594633 
595634 Configuration override for the web_search tool.
596635 
636 - `type: optional "web_search"`
637 
597638 - `name: "web_search"`
598639 
599640 Must be "web_search".
from line 651
610651 
611652 Whether this tool is enabled and available to Claude. Overrides the default_config setting.
612653 
613 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
654 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
614655 
615656 Permission policy for tool execution.
616657 
from line 663
622663 
623664 Tool calls require user confirmation before execution.
624665 
625 - `type: optional "web_search"`
666 - `BetaManagedAgentsAutoPolicy object`
626667 
668 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
669 
627670 - `user_location: optional BetaManagedAgentsUserLocation or null`
628671 
629672 Approximate user location for search result localization.
from line 705
662705 
663706 Whether tools are enabled and available to Claude by default. Defaults to true if not specified.
664707 
665 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
708 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
666709 
667710 Permission policy for tool execution.
668711 
from line 717
674717 
675718 Tool calls require user confirmation before execution.
676719 
720 - `BetaManagedAgentsAutoPolicy object`
721 
722 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
723 
677724 - `BetaManagedAgentsMCPToolsetParams object`
678725 
679726 Configuration for tools from an MCP server defined in `mcp_servers`.
680727 
728 - `type: "mcp_toolset"`
729 
681730 - `mcp_server_name: string`
682731 
683732 Name of the MCP server. Must match a server name from the mcp_servers array. 1-255 characters.
from line 733
684733 
685734 minLength: 1, maxLength: 255
686735 
687 - `type: "mcp_toolset"`
688 
689736 - `configs: optional array of BetaManagedAgentsMCPToolConfigParams`
690737 
691738 Per-tool configuration overrides.
from line 747
700747 
701748 Whether this tool is enabled. Overrides the `default_config` setting.
702749 
703 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
750 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
704751 
705752 Permission policy for tool execution.
706753 
from line 759
712759 
713760 Tool calls require user confirmation before execution.
714761 
762 - `BetaManagedAgentsAutoPolicy object`
763 
764 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
765 
715766 - `default_config: optional BetaManagedAgentsMCPToolsetDefaultConfigParams or null`
716767 
717768 Default configuration for all tools from an MCP server.
from line 771
720771 
721772 Whether tools are enabled by default. Defaults to true if not specified.
722773 
723 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or null`
774 - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`
724775 
725776 Permission policy for tool execution.
726777 
from line 783
732783 
733784 Tool calls require user confirmation before execution.
734785 
786 - `BetaManagedAgentsAutoPolicy object`
787 
788 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
789 
735790 - `BetaManagedAgentsCustomToolParams object`
736791 
737792 A custom tool that is executed by the API client rather than the agent. When the agent calls this tool, an `agent.custom_tool_use` event is emitted and the session goes idle, waiting for the client to provide the result via a `user.custom_tool_result` event.
738793 
794 - `type: "custom"`
795 
739796 - `description: string`
740797 
741798 Description of what the tool does, shown to the agent to help it decide when to use the tool.
from line 815
758815 
759816 minLength: 1, maxLength: 128
760817 
761 - `type: "custom"`
762 
763818## Returns
764819 
765820- `BetaManagedAgentsAgent object`
from line 821
766821 
767822 A Managed Agents `agent`.
768823 
824 - `type: "agent"`
825 
769826 - `id: string`
770827 
771828 - `archived_at: string or null`
from line 841
784841 
785842 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`
786843 
787 - `name: string`
788 
789844 - `type: "url"`
790845 
846 - `name: string`
847 
791848 - `url: string`
792849 
793850 - `metadata: map[string]`
from line 973
916973 
917974 Resolved coordinator topology with a concrete agent roster.
918975 
976 - `type: "coordinator"`
977 
919978 - `agents: array of BetaManagedAgentsAgentReference or BetaManagedAgentsAdvisor`
920979 
921980 Agents the coordinator may spawn as session threads, each resolved to a specific version.
from line 983
924983 
925984 A resolved agent reference with a concrete version.
926985 
927 - `id: string`
928 
929986 - `type: "agent"`
930987 
988 - `id: string`
989 
931990 - `version: number`
932991 
933992 format: int32
from line 995
936995 
937996 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.
938997 
998 - `type: "advisor"`
999 
9391000 - `model: string`
9401001 
9411002 The advisor model id.
9421003 
943 - `type: "advisor"`
944 
945 - `type: "coordinator"`
946 
9471004 - `name: string`
9481005 
9491006 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`
from line 1009
9521009 
9531010 A resolved Anthropic-managed skill.
9541011 
955 - `skill_id: string`
956 
9571012 - `type: "anthropic"`
9581013 
1014 - `skill_id: string`
1015 
9591016 - `version: string`
9601017 
9611018 - `BetaManagedAgentsCustomSkill object`
from line 1019
9621019 
9631020 A resolved user-created custom skill.
9641021 
965 - `skill_id: string`
966 
9671022 - `type: "custom"`
9681023 
1024 - `skill_id: string`
1025 
9691026 - `version: string`
9701027 
9711028 - `system: string or null`
from line 1031
9741031 
9751032 - `BetaManagedAgentsAgentToolset20260401 object`
9761033 
1034 - `type: "agent_toolset_20260401"`
1035 
9771036 - `configs: array of BetaManagedAgentsAgentToolConfig`
9781037 
9791038 - `BetaManagedAgentsBashToolConfig object`
from line 1039
9801039 
9811040 Configuration for the bash tool.
9821041 
1042 - `type: "bash"`
1043 
9831044 - `enabled: boolean`
9841045 
9851046 - `name: "bash"`
9861047 
987 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1048 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
9881049 
9891050 Permission policy for tool execution.
9901051 
from line 1061
10001061 
10011062 - `type: "always_ask"`
10021063 
1003 - `type: "bash"`
1064 - `BetaManagedAgentsAutoPolicy object`
10041065 
1066 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1067 
1068 - `type: "auto"`
1069 
10051070 - `BetaManagedAgentsEditToolConfig object`
10061071 
10071072 Configuration for the edit tool.
10081073 
1074 - `type: "edit"`
1075 
10091076 - `enabled: boolean`
10101077 
10111078 - `name: "edit"`
10121079 
1013 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1080 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
10141081 
10151082 Permission policy for tool execution.
10161083 
from line 1089
10221089 
10231090 Tool calls require user confirmation before execution.
10241091 
1025 - `type: "edit"`
1092 - `BetaManagedAgentsAutoPolicy object`
10261093 
1094 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1095 
10271096 - `BetaManagedAgentsReadToolConfig object`
10281097 
10291098 Configuration for the read tool.
10301099 
1100 - `type: "read"`
1101 
10311102 - `enabled: boolean`
10321103 
10331104 - `name: "read"`
10341105 
1035 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1106 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
10361107 
10371108 Permission policy for tool execution.
10381109 
from line 1115
10441115 
10451116 Tool calls require user confirmation before execution.
10461117 
1047 - `type: "read"`
1118 - `BetaManagedAgentsAutoPolicy object`
10481119 
1120 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1121 
10491122 - `BetaManagedAgentsWriteToolConfig object`
10501123 
10511124 Configuration for the write tool.
10521125 
1126 - `type: "write"`
1127 
10531128 - `enabled: boolean`
10541129 
10551130 - `name: "write"`
10561131 
1057 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1132 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
10581133 
10591134 Permission policy for tool execution.
10601135 
from line 1141
10661141 
10671142 Tool calls require user confirmation before execution.
10681143 
1069 - `type: "write"`
1144 - `BetaManagedAgentsAutoPolicy object`
10701145 
1146 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1147 
10711148 - `BetaManagedAgentsGlobToolConfig object`
10721149 
10731150 Configuration for the glob tool.
10741151 
1152 - `type: "glob"`
1153 
10751154 - `enabled: boolean`
10761155 
10771156 - `name: "glob"`
10781157 
1079 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1158 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
10801159 
10811160 Permission policy for tool execution.
10821161 
from line 1167
10881167 
10891168 Tool calls require user confirmation before execution.
10901169 
1091 - `type: "glob"`
1170 - `BetaManagedAgentsAutoPolicy object`
10921171 
1172 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1173 
10931174 - `BetaManagedAgentsGrepToolConfig object`
10941175 
10951176 Configuration for the grep tool.
10961177 
1178 - `type: "grep"`
1179 
10971180 - `enabled: boolean`
10981181 
10991182 - `name: "grep"`
11001183 
1101 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1184 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
11021185 
11031186 Permission policy for tool execution.
11041187 
from line 1193
11101193 
11111194 Tool calls require user confirmation before execution.
11121195 
1113 - `type: "grep"`
1196 - `BetaManagedAgentsAutoPolicy object`
11141197 
1198 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1199 
11151200 - `BetaManagedAgentsWebFetchToolConfig object`
11161201 
11171202 Configuration for the web_fetch tool.
11181203 
1204 - `type: "web_fetch"`
1205 
11191206 - `enabled: boolean`
11201207 
11211208 - `name: "web_fetch"`
11221209 
1123 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1210 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
11241211 
11251212 Permission policy for tool execution.
11261213 
from line 1219
11321219 
11331220 Tool calls require user confirmation before execution.
11341221 
1135 - `type: "web_fetch"`
1222 - `BetaManagedAgentsAutoPolicy object`
11361223 
1224 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1225 
11371226 - `allowed_domains: optional array of string`
11381227 
11391228 - `blocked_domains: optional array of string`
from line 1235
11461235 
11471236 Configuration for the web_search tool.
11481237 
1238 - `type: "web_search"`
1239 
11491240 - `enabled: boolean`
11501241 
11511242 - `name: "web_search"`
11521243 
1153 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1244 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
11541245 
11551246 Permission policy for tool execution.
11561247 
from line 1253
11621253 
11631254 Tool calls require user confirmation before execution.
11641255 
1165 - `type: "web_search"`
1256 - `BetaManagedAgentsAutoPolicy object`
11661257 
1258 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1259 
11671260 - `allowed_domains: optional array of string`
11681261 
11691262 - `blocked_domains: optional array of string`
from line 1297
12041297 
12051298 - `enabled: boolean`
12061299 
1207 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1300 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
12081301 
12091302 Permission policy for tool execution.
12101303 
from line 1309
12161309 
12171310 Tool calls require user confirmation before execution.
12181311 
1219 - `type: "agent_toolset_20260401"`
1312 - `BetaManagedAgentsAutoPolicy object`
12201313 
1314 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1315 
12211316 - `BetaManagedAgentsMCPToolset object`
12221317 
1318 - `type: "mcp_toolset"`
1319 
12231320 - `configs: array of BetaManagedAgentsMCPToolConfig`
12241321 
12251322 - `enabled: boolean`
from line 1323
12261323 
12271324 - `name: string`
12281325 
1229 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1326 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
12301327 
12311328 Permission policy for tool execution.
12321329 
from line 1335
12381335 
12391336 Tool calls require user confirmation before execution.
12401337 
1338 - `BetaManagedAgentsAutoPolicy object`
1339 
1340 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
1341 
12411342 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`
12421343 
12431344 Resolved default configuration for all tools from an MCP server.
from line 1345
12441345 
12451346 - `enabled: boolean`
12461347 
1247 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy`
1348 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`
12481349 
12491350 Permission policy for tool execution.
12501351 
from line 1357
12561357 
12571358 Tool calls require user confirmation before execution.
12581359 
1259 - `mcp_server_name: string`
1360 - `BetaManagedAgentsAutoPolicy object`
12601361 
1261 - `type: "mcp_toolset"`
1362 The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.
12621363 
1364 - `mcp_server_name: string`
1365 
12631366 - `BetaManagedAgentsCustomTool object`
12641367 
12651368 A custom tool as returned in API responses.
12661369 
1370 - `type: "custom"`
1371 
12671372 - `description: string`
12681373 
12691374 - `input_schema: BetaManagedAgentsCustomToolInputSchema`
from line 1382
12771382 - `required: optional array of string or null`
12781383 
12791384 - `name: string`
1280 
1281 - `type: "custom"`
1282 
1283 - `type: "agent"`
12841385 
12851386 - `updated_at: string`
12861387