Repository navigation
Expand file tree
/
Copy pathconfig-baseline.json
More file actions
11247 lines (11247 loc) · 439 KB
/
Copy pathconfig-baseline.json
File metadata and controls
11247 lines (11247 loc) · 439 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
{
"generatedBy": "scripts/generate_config_baseline.py",
"entries": [
{
"path": "agent",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Agent",
"help": "Agent runtime configuration.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {
"approval_mode": "auto",
"streaming": true,
"model": "auto",
"role_models": {},
"role_efforts": {},
"model_context_tokens": {},
"fallback_model": "auto",
"refusal_fallback_model": "",
"reasoning_effort": "",
"provider": "acp",
"mcp_registry_mode": false,
"mcp_quarantine_after_failures": 3,
"acp_backend": "kas",
"member_acp_backend": "kas",
"acp_failover": {},
"kas_bundle": false,
"kas_persist_mcp_sign_in": true,
"kas_bundle_url": "https://kas.crew.kiro.dev",
"native_skill_projection": true,
"default_agent": "",
"crew_mode_agent": "kirocrew-conductor",
"deepseek_env": {},
"strands_env": {},
"strands_project": "",
"strands_project_sha256": "",
"strands_provider": "bedrock",
"strands_model": "auto",
"sweep_agents_backups": false,
"sandbox": "auto",
"sandbox_allow_no_isolation": false,
"sandbox_allow_unsandboxed_exec": false,
"apps_allow_third_party": false,
"apps_trusted": [],
"apps_trusted_local": [],
"apps_trusted_repositories": {},
"apps_default_installs_applied": [],
"apps_builtin_adoption": {},
"apps_ui_stream_timeout_secs": 30,
"jail": "auto",
"dangerously_skip_permissions": false,
"yolo_duration": "6h",
"notify_override_expiry": true,
"bot_name": "",
"tool_search": true,
"tool_search_min_pct": 5,
"tool_search_min_tokens": 50000,
"session_sharing": true,
"max_subagents": 0,
"max_stop_hook_nudges": 100,
"spawn_min_memory_gb": 2.0,
"resource_pressure_gb": 4.0,
"resource_critical_gb": 2.0,
"admission_gate": true,
"task_queue_enabled": true,
"task_dispatch_window": 64,
"task_store_journal_mode": "auto",
"admit_wait_secs": 30,
"subagent_queue_max_wait_secs": 1800,
"start_collect_timeout_secs": 300,
"lane_weights": {},
"child_reserve": 1,
"recovery_backoff_base_secs": 2.0,
"recovery_backoff_max_secs": 120.0,
"session_start_concurrency": "auto",
"child_env_defaults": {},
"env": {},
"cold_start_concurrency": "auto",
"runtime_spawn_concurrency": "auto",
"adaptive_concurrency": true,
"adaptive_concurrency_mode": "aimd",
"adaptive_floor": 1,
"adaptive_initial": 4,
"adaptive_slow_start": true,
"controller_sample_secs": 5,
"dependency_max_attempts": 20,
"dependency_wait_deadline_secs": 3600,
"dependency_wake_per_tick": 0,
"dependency_wake_spacing_secs": 1.0,
"interactive_command_policy": "cancel",
"workflow_run_timeout_secs": 3600,
"subagent_mem_buffer_pct": 20,
"chat_turn_timeout_secs": 14400,
"session_start_timeout_secs": 90,
"tool_approval_timeout_secs": 600,
"session_control": true,
"session_manager_chain": false,
"member_dispatch": true,
"crew_panel": true,
"subagent_cost_gb": 0.5,
"subagent_cpu_cost_cores": 1.0,
"subagent_auto_max": 32,
"subagent_spawn_stagger_secs": 0.25,
"subagent_max_turns": 1000,
"subagent_timeout_secs": 10800,
"subagent_stall_idle_secs": 120,
"completion_keep": "head",
"completion_keep_chars": 3000,
"subagent_result_ttl_secs": 3600,
"subagent_cwd_allowed_roots": [
"~/workspace",
"~/workspaces",
"~/workplace",
"~/workplaces"
],
"max_channels": 1,
"max_channel_agents": 3,
"log_level": "WARNING",
"soft_stop_budget_secs": 10.0
}
},
{
"path": "agent.approval_mode",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Approval Mode",
"help": "Tool approval mode. Every channel dispatcher resolves it once at start (with the CLI --approval override), so a change takes effect at the next restart.",
"hasChildren": false,
"enumValues": [
"auto",
"interactive"
],
"defaultValue": "auto",
"requiresRestart": true
},
{
"path": "agent.streaming",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Streaming",
"help": "Enable streaming responses.",
"hasChildren": false,
"enumValues": null,
"defaultValue": true
},
{
"path": "agent.model",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Model",
"help": "LLM model identifier. 'auto' resolves from agent config.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "auto"
},
{
"path": "agent.role_models",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Per-role models",
"help": "Optional per-task-class model overrides. Keys: 'background' (lite / heartbeat background workers) and 'subagent' (spawned sub-agents). An empty value or 'auto' defers to the chat default (agent.model) and then to the provider default, so an unpinned role stays usable on every subscription tier. Pin a cheaper model here to run background / sub-agent work on it without changing the interactive chat default. A 'subagent' pin runs each spawned sub-agent in its own runtime (its own agent process and MCP servers, about 0.55 GB of memory each) where it could otherwise share the parent chat's.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.role_models.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.role_efforts",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Per-role reasoning effort",
"help": "Optional per-task-class reasoning effort, paired with role_models (keys: 'background', 'subagent'). Empty for a role inherits the chat default (agent.reasoning_effort) and then the provider/model default. Only applies on reasoning-capable models. Like a model pin, a 'subagent' effort runs each spawned sub-agent in its own runtime (about 0.55 GB of memory each) where it could otherwise share the parent chat's.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.role_efforts.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.model_context_tokens",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Served context windows",
"help": "Context window, in tokens, that an exact model id is served with, for example {\"server-a/org/model-7b\": 32768}. The key is the id exactly as agent.model carries it, so two servers of the same weights stay distinct. A declared window budgets the first prompt of a new session, before the backend has reported one, and outranks the catalog and registry windows. A value is an integer from 1024 to 10000000, at most 64 entries are read, and an invalid entry is ignored with a warning. Empty by default: a model with no window of its own keeps the 1,000,000-token reference. Takes effect on the next config reload.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.model_context_tokens.*",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.fallback_model",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Fallback model",
"help": "Model tried when the active model's transient-retry budget is exhausted (throttle/capacity). Default 'auto' defers to the backend's availability-aware routing; a concrete model id (as advertised by the provider, e.g. 'claude-opus-4.8') is tried first with 'auto' as the final fallthrough; empty ('') disables fallback entirely (fail loudly, pre-feature behavior). A fallback swap is announced in chat, sticks until the primary recovers, and the serving model is recorded in every turn's stats — never silent.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "auto"
},
{
"path": "agent.refusal_fallback_model",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Refusal fallback model",
"help": "Model the current message is retried on ONCE when the active model's content filter declines it (refusal / CONTENT_FILTERED). Empty ('', the default) disables the retry: the refusal card surfaces exactly as before. A concrete model id (as advertised by the provider, e.g. 'claude-opus-4.8') retries that single message on it and restores the primary model on the next turn; 'auto' retries on the model the provider's refusal envelope recommends, when it names one. The retry is announced in chat — never silent — and a refusal from the fallback too is terminal.",
"hasChildren": false,
"enumValues": null,
"defaultValue": ""
},
{
"path": "agent.reasoning_effort",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Reasoning Effort",
"help": "Default reasoning effort for new sessions on models that support it. Empty defers to the provider/model default. Per-session overrides win.",
"hasChildren": false,
"enumValues": [
"",
"low",
"medium",
"high",
"xhigh",
"max"
],
"defaultValue": ""
},
{
"path": "agent.provider",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Provider",
"help": "LLM provider backend (KiroACP / kiro-cli).",
"hasChildren": false,
"enumValues": [
"acp"
],
"defaultValue": "acp"
},
{
"path": "agent.mcp_registry_mode",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Enterprise MCP Registry Mode",
"help": "Set true when this Kiro account is governed by an enterprise MCP registry (Kiro console -> Shared settings -> MCP Registry URL, which applies to IAM Identity Center and API-key sign-ins). In registry mode the client connects ONLY to mcpServers entries carrying 'type': \"registry\" that resolve to a catalog entry of the same name, so Kiro Crew stamps that marker on the servers it manages. Leave false on a personal account: with no registry configured the filter inverts and registry-marked entries are the ones dropped. The administrator must also allow-list, by those exact names, every managed kirocrew-* server Kiro Crew mounts: kirocrew-core, kirocrew-cron and kirocrew-computer by default, plus any opt-in server (e.g. kirocrew-dashboard, kirocrew-work) an agent is granted.",
"hasChildren": false,
"enumValues": null,
"defaultValue": false
},
{
"path": "agent.mcp_quarantine_after_failures",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Failing-Probe Threshold",
"help": "Consecutive failed probes before an MCP server is reported as persistently failing on its dashboard row. A probe verdict is otherwise forgotten between rounds, so a server that failed once on a cold cache looked identical to one that has failed forty times. Counts only 'error' and 'timeout': a server asking for OAuth sign-in is working correctly and is never counted, and one success clears the count. The server stays mounted, but while its streak is at or over this threshold MCP discovery stops spawning it for probes; the count survives a gateway restart and holds until it is cleared (the dashboard offers a one-click reset). 0 turns it off.",
"hasChildren": false,
"enumValues": null,
"defaultValue": 3
},
{
"path": "agent.acp_backend",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "ACP Backend",
"help": "Which ACP agent to drive. Omitted or '' means no choice: KAS is the default. 'kiro' selects kiro-cli; any other value must be a registered selectable backend id. The live list is served by GET /api/config/schema.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "kas"
},
{
"path": "agent.member_acp_backend",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Crew member ACP backend",
"help": "Backend for crew-member DM sessions: 'kas' (default when omitted or '') or another dispatch-capable backend (claude, codex, opencode, goose). Crewmates run on KAS as their Kiro backend, so 'kiro' here also selects KAS. Members dispatch work into worker sessions through session-control tools mounted per session over the wire.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "kas"
},
{
"path": "agent.acp_failover",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "ACP backend failover",
"help": "Opt-in, off-by-default policy that starts a sub-agent on another backend when the one it asked for cannot start. An object with 'enabled' (bool), 'chain' (ordered list of distinct backend names, spelled as spawn_run spells them, e.g. 'codex', 'claude', 'kiro') and optional 'triggers' (which refusals cause a hop: backend_not_installed, backend_restart_required; both when omitted). A hop only goes to a later chain member that can start, never a silent Kiro fallback, and the spawn result names the backend that ran and the one it replaced. A hop moves the prompt and its context to that backend's vendor and bill. A name that is not selectable is skipped. Absent means no failover.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.kas_bundle",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "KAS: run Kiro Crew's own engine bundle",
"help": "When true, the 'kas' backend spawns the @kiro/agent build that 'kirocrew kas install' laid down under the data home (engines/kas/) with its own Node runtime, instead of reaching the engine through kiro-cli's ACP relay. kiro-cli is not needed on this route, and the engine always takes its credential from Kiro Crew's own sign-in (the dashboard's Sign in), never from kiro-cli's store. With nothing installed the gateway installs the pinned engine at the first spawn. With false, KAS still takes this route when kiro-cli's relay cannot run on the host (no kiro-cli, or one that is not signed in).",
"hasChildren": false,
"enumValues": null,
"defaultValue": false,
"requiresRestart": true
},
{
"path": "agent.kas_persist_mcp_sign_in",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "KAS: keep remote MCP sign-ins across sessions",
"help": "When true (the default), the 'kas' backend lets Kiro Crew keep a remote MCP server's OAuth sign-in (client registration and tokens) in a dedicated encrypted vault under the data home (kas/mcp-grants/), so new sessions, subagents and crons reuse it until the refresh token expires or is revoked, as the 'kiro' backend does with its own sign-in cache. The vault is not the Secrets vault: it is not listed there and cannot be granted to a cron, and the agent cannot read it. When false, the engine keeps each sign-in in memory and a new session asks again. Only a JSON true turns it on: any other stored value, or a config file that cannot be read, keeps sign-ins in memory.",
"hasChildren": false,
"enumValues": null,
"defaultValue": true,
"requiresRestart": true
},
{
"path": "agent.kas_bundle_url",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "KAS bundle base URL",
"help": "https:// base the KAS engine manifest is fetched from (<base>/kas/<version>/manifest.json). Every artifact is sha256-pinned by that manifest and the manifest by this release, so the url chooses WHERE the pinned build is fetched from and never what is trusted. Only an https:// url is accepted; redirects that leave the host are refused.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "https://kas.crew.kiro.dev"
},
{
"path": "agent.native_skill_projection",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Native skill views for kiro-cli",
"help": "When true (the default), the 'kiro' backend launches each agent under a Kiro Crew managed view (kirocrew-skill-view-*.json in the kiro agents directory) that serves skills through Kiro Crew's skill search, and overlays the workspace's .kiro/settings/cli.json while it runs. When false, agents launch under their authored names, no view is written, the workspace overlay is rolled back, and kiro-cli loads skill metadata natively again. Same effect as KIROCREW_NATIVE_SKILL_PROJECTION=0, which still disables projection whatever this says.",
"hasChildren": false,
"enumValues": null,
"defaultValue": true,
"requiresRestart": true
},
{
"path": "agent.default_agent",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Default Agent",
"help": "Default agent name for new sessions.",
"hasChildren": false,
"enumValues": null,
"defaultValue": ""
},
{
"path": "agent.crew_mode_agent",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Crew Mode conductor",
"help": "The conductor template a chat runs when Crew Mode is turned on. Empty falls back to the default. Takes effect live for new chats; no gateway restart.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "kirocrew-conductor"
},
{
"path": "agent.deepseek_env",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "DeepSeek Harness provider keys",
"help": "Provider keys handed to the DeepSeek Harness ('deepseek' backend) as environment variables at spawn, mapping an environment-variable NAME to a 'secret://<vault name>' reference. The harness resolves a provider credential from its inherited environment above its own credential files, so this is how it reaches a hosted model without Kiro Crew leaving those files readable inside the sandbox. Store the key under Settings > Secrets, then map it here, e.g. {\"DEEPSEEK_API_KEY\": \"secret://my-dsh-key\"}. Any provider name the harness knows works (ANTHROPIC_API_KEY, OPENAI_API_KEY, ...). A plaintext value is REFUSED at write time -- the config write itself fails, so no credential is ever stored in config.json -- and so, at spawn, is a name the harness would forward to its own shell children, a name Kiro Crew owns, or one Kiro Crew's agent environment scrub strips; the DeepSeek session is refused before it starts with a message naming the offending key. Empty means no key: a model served locally on this machine needs none. Ignored by every other backend.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.deepseek_env.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.strands_env",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Strands provider keys",
"help": "Model-provider settings handed to the Strands harness ('strands' backend) as environment variables at spawn, mapping an environment-variable NAME to a 'secret://<vault name>' reference, the same shape as agent.deepseek_env. Store the value under Settings > Secrets, then map it here, e.g. {\"ANTHROPIC_API_KEY\": \"secret://my-anthropic-key\"}. Only the names the Strands CLI reads for its providers are accepted: AWS_REGION, AWS_DEFAULT_REGION and AWS_BEARER_TOKEN_BEDROCK (Bedrock and Bedrock Mantle), ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY, OLLAMA_HOST, OLLAMA_MODEL, LITELLM_BASE_URL, LITELLM_API_KEY and LITELLM_MODEL. Kiro Crew withholds every mapped name from the shells and code the Strands agent runs. A plaintext value is REFUSED at write time, and any other name refuses the Strands session before it starts with a message naming the key. The harness runs sandboxed and cannot read ~/.aws, so Bedrock needs AWS_BEARER_TOKEN_BEDROCK (a Bedrock API key) and AWS_REGION mapped here. Ignored by every other backend.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.strands_env.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.strands_project",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Strands project",
"help": "The Strands Python project the 'strands' backend runs: the path of its agent.py, or of a folder holding one. Empty runs Kiro Crew's bundled minimal project. Either way Kiro Crew loads it through its own gate module, which makes every Strands Agent the project builds ask Kiro Crew before each tool call, so a project needs no change of its own. Its imports must be installed in the Python environment Kiro Crew runs Strands with. A project of yours runs only as the version recorded in agent.strands_project_sha256. Takes effect for new Strands sessions. Ignored by every other backend.",
"hasChildren": false,
"enumValues": null,
"defaultValue": ""
},
{
"path": "agent.strands_project_sha256",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Strands project fingerprint",
"help": "The fingerprint of the agent.strands_project version you confirmed: a sha256 over every file in the project's folder. The project's code runs with your provider keys and outside the tool gate, so a Strands session on your project starts only when the folder still matches it; a refused session prints the current value and the command that records it. Set it yourself after reviewing the files (kirocrew config set agent.strands_project_sha256 <value>); Kiro Crew never records it on its own. Packages the project imports from outside its folder are not covered. Ignored for the bundled project and by every other backend.",
"hasChildren": false,
"enumValues": null,
"defaultValue": ""
},
{
"path": "agent.strands_provider",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Strands provider",
"help": "The model provider a new Strands session ('strands' backend) runs on. With agent.strands_model it builds the Strands CLI's --model '<provider>/<model>' argument; with agent.strands_model 'auto' no --model is passed. A model picked for the chat wins over both. Any value outside the list refuses the session before it starts. Its key comes from agent.strands_env. Ignored by every other backend.",
"hasChildren": false,
"enumValues": [
"bedrock",
"bedrock-mantle",
"anthropic",
"openai",
"google",
"ollama",
"litellm"
],
"defaultValue": "bedrock"
},
{
"path": "agent.strands_model",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Strands model",
"help": "The model a new Strands session runs when the chat picks none, as an id for agent.strands_provider: Kiro Crew launches the Strands CLI with --model '<provider>/<model>'. 'auto' (or empty) passes no --model, so the provider's CLI default runs. Ignored by every other backend.",
"hasChildren": false,
"enumValues": null,
"defaultValue": "auto"
},
{
"path": "agent.sweep_agents_backups",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Sweep foreign agent backups",
"help": "When true, the agents-directory janitor also deletes aged backup files (*.bak-<digits> / *.json.bak.<digits>, older than 14 days) from the shared kiro agents directory. OFF by default: Kiro Crew does not author those backups, so every one it would delete belongs to another tool whose retention policy is not ours to decide. The orphaned atomic-write TEMP sweep (24h) always runs and reclaims most of the growth at near-zero risk; enable this only if you also want foreign backups in that directory reaped.",
"hasChildren": false,
"enumValues": null,
"defaultValue": false
},
{
"path": "agent.sandbox",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Sandbox",
"help": "Sandbox mode for ACP provider. Default 'auto' engages OS-level isolation (namespace on Linux, sandbox-exec on macOS) at the standard tier and automatically defers to kiro-cli's internal sandbox on macOS when it is enabled (kiro-cli >= 2.13; nested seatbelt causes EPERM). The standard tier deliberately leaves ~/.aws, ~/.ssh and ~/.kube visible to the agent's shell so the aws CLI, boto3 credential_process, git-over-SSH and kubectl keep working; the file tools still refuse those paths. Set to 'strict' to also hide ~/.aws (including ~/.aws/sso/cache, kiro-cli's grant store for OAuth-connected remote MCP servers), ~/.ssh (except known_hosts), ~/.kube and ~/.config/gh, plus the credential files ~/.npmrc, ~/.pypirc, ~/.netrc and ~/.git-credentials, from every agent subprocess -- opt-in, and inside the agent it breaks the aws CLI, boto3, git-over-SSH, gh, kubectl, npm/pip registry auth, .netrc HTTPS auth, the git credential store and remote-MCP OAuth for the same reason. Like every value of this key, a change applies to sessions started after it; a session already running keeps the tier it was spawned with until it ends. 'strict' changes nothing where Kiro Crew applies no sandbox of its own: Windows has no OS backend, and a macOS spawn delegated to kiro-cli's internal sandbox is confined by that profile instead. Set to 'off' to skip Kiro Crew's own OS-level sandbox -- delegation to kiro-cli's internal sandbox still fires on macOS if it is enabled, and a SECURITY warning is logged when neither layer is active.",
"hasChildren": false,
"enumValues": [
"auto",
"strict",
"off"
],
"defaultValue": "auto"
},
{
"path": "agent.sandbox_allow_no_isolation",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Allow No-Isolation Fallback",
"help": "Acknowledge running the agent subprocess WITHOUT OS-level credential isolation when no sandbox backend is available (e.g. macOS >= 26, or Linux without user namespaces). When false (default), that fallback is logged as a loud SECURITY warning. When true, the operator has accepted the risk and it is logged at info level.",
"hasChildren": false,
"enumValues": null,
"defaultValue": false
},
{
"path": "agent.sandbox_allow_unsandboxed_exec",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Allow Unsandboxed Execution",
"help": "When true, allow agent subprocesses to execute without any sandbox backend (fail-open). When false, wrap_argv raises if no sandbox backend is available and mode is not 'off', preventing unsandboxed execution entirely (fail-closed). This is distinct from sandbox_allow_no_isolation which only controls warning severity — this field controls whether execution proceeds at all. A LOADED config carries the effective policy: the loader resolves an undeclared key through config.loader.unsandboxed_exec_platform_default() — allow on Windows, which has no backend that any operator action could install, and fail-closed everywhere else, where a missing backend is broken or one profile away from working. A declared value always wins in both directions, and a governance sandbox.min_level floor outranks the declaration. save() omits the key unless the operator declared it, so a 'never decided' host stays undeclared on disk. This dataclass default stays platform-independent so the committed config-schema snapshot is identical on every platform. `kirocrew setup` surfaces the decision on a backend-less host, offering the opt-in where the default is fail-closed and stating the exposure plus offering the opt-out where it is allow, and writes nothing unless the operator answers yes.",
"hasChildren": false,
"enumValues": null,
"defaultValue": false
},
{
"path": "agent.apps_allow_third_party",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Allow Third-Party Apps",
"help": "Explicitly allow executable code from third-party (non-builtin) apps. Defaults to false. Only the JSON boolean true admits in-process Python hooks, backend processes, lifecycle/install scripts, and openCommand. App code can access the filesystem, network, and in-memory credentials; enable this only for apps you trust (CSE SEC-012). Prefer apps_trusted, which grants the same admission to ONE named app.",
"hasChildren": false,
"enumValues": null,
"defaultValue": false
},
{
"path": "agent.apps_trusted",
"kind": "core",
"type": "array",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Trusted Apps",
"help": "Per-app grants for third-party execution — the narrow form of apps_allow_third_party. An app whose manifest name appears here is admitted to run Python hooks, its backend, lifecycle scripts, and openCommand; every other third-party app stays blocked. Only a JSON array of app-name strings is honoured, and no wildcard entry is accepted (use apps_allow_third_party to trust all).",
"hasChildren": true,
"enumValues": null,
"defaultValue": []
},
{
"path": "agent.apps_trusted.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.apps_trusted_local",
"kind": "core",
"type": "array",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Trusted Local Apps",
"help": "App names whose per-app execution grant was explicitly reviewed as local, repository-less code. This internal grant-kind marker distinguishes current local consent from legacy name-only grants; it is effective only with the matching apps_trusted entry.",
"hasChildren": true,
"enumValues": null,
"defaultValue": []
},
{
"path": "agent.apps_trusted_local.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.apps_trusted_repositories",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Trusted App Repositories",
"help": "Repository coordinates captured by the per-app trust endpoint. Each key is an app name from apps_trusted and each value is the normalized repository shown at consent. Registry installation refuses if that name later resolves to a different repository. Legacy repository-backed grants without an entry require one-time re-consent before code execution.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.apps_trusted_repositories.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.apps_default_installs_applied",
"kind": "core",
"type": "array",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Applied Default App Installs",
"help": "App names already installed once because their registry marked them defaultInstall. A name here is never installed again automatically, so an app you uninstall or disable stays that way. Remove a name to let it be installed again on the next start.",
"hasChildren": true,
"enumValues": null,
"defaultValue": []
},
{
"path": "agent.apps_default_installs_applied.*",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": false,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.apps_builtin_adoption",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Adopted Built-in Apps",
"help": "Former built-in apps this gateway moved to their own repository. A name here is never registered as a built-in again, so uninstalling the app sticks. Written by the gateway; do not edit.",
"hasChildren": true,
"enumValues": null,
"defaultValue": {}
},
{
"path": "agent.apps_builtin_adoption.*",
"kind": "core",
"type": "object",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "*",
"help": "",
"hasChildren": true,
"enumValues": null,
"defaultValue": null
},
{
"path": "agent.apps_ui_stream_timeout_secs",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "App UI Stream Timeout",
"help": "Total transfer deadline, in seconds, for ONE response body on the unauthenticated /apps/{name}/ui/ route. Clamped to 5..600; a value that is not a whole number loads as 30. Read per request, so an edit applies without a restart. There is no off switch: the route holds eight descriptor permits, and this deadline is what stops a client that quits draining its socket from holding one for as long as it stays connected — eight such clients would otherwise stop every app UI on the host. Raise it only when a real transfer on this host needs longer than the default allows.",
"hasChildren": false,
"enumValues": null,
"defaultValue": 30
},
{
"path": "agent.jail",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Jail",
"help": "Process-isolation jail mode for agent-bearing commands. 'auto' uses a jail when the active edition supplies a working backend (the public edition has none, so 'auto' and 'on' are no-ops there); 'off' disables it. Disable per-invocation with --no-jail or KIROCREW_NO_JAIL=1.",
"hasChildren": false,
"enumValues": [
"auto",
"on",
"off"
],
"defaultValue": "auto",
"requiresRestart": true
},
{
"path": "agent.dangerously_skip_permissions",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Dangerously Skip Permissions",
"help": "Skip EVERY tool approval confirmation, permanently. Declaring it here is a standing instruction: the grant does not expire and is re-established on every startup. This is the advanced, config-file-only escape hatch — there is deliberately no dashboard toggle for it. An enterprise policy can forbid it, which falls back to the ad-hoc duration below.",
"hasChildren": false,
"enumValues": null,
"defaultValue": false,
"requiresRestart": true
},
{
"path": "agent.yolo_duration",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Ad-hoc Auto-approve Duration",
"help": "How long auto-approve (YOLO) lasts when it is enabled AD HOC — from the dashboard picker, Slack, or the API. Every one of those surfaces uses this same duration. Accepts 30m / 1h / 6h / 12h / 24h, or until_shutdown to keep it on with no timed expiry until Kiro Crew restarts. Timed values are capped at 24h. Does NOT apply to a grant declared via 'dangerously_skip_permissions' above, which persists.",
"hasChildren": false,
"enumValues": [
"30m",
"1h",
"6h",
"12h",
"24h",
"until_shutdown"
],
"defaultValue": "6h"
},
{
"path": "agent.notify_override_expiry",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Notify on Override Expiry",
"help": "DM the Slack owner when a time-limited safety override (YOLO) expires. Disable to silence the recurring expiry DM; the dashboard banner still shows.",
"hasChildren": false,
"enumValues": null,
"defaultValue": true
},
{
"path": "agent.bot_name",
"kind": "core",
"type": "string",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Bot Name",
"help": "Custom name the bot identifies as in conversations. Leave empty for default.",
"hasChildren": false,
"enumValues": null,
"defaultValue": ""
},
{
"path": "agent.tool_search",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "MCP Tool Search",
"help": "Load MCP tool specs on demand (search-and-call) instead of sending every tool definition each turn, keeping the context window clear when many MCP servers are configured. kiro-cli applies it through the agent overlay; KAS receives it in the ACP initialize request, enabled only when the spawn agent's spec grants the tool_search loader. Deferral only starts once the specs cross tool_search_min_pct or tool_search_min_tokens; disabling reverts to sending full tool specs. Crew's servers defer only when the spawn runs the pinned kiro-cli install or its kiro-cli-chat, and both are >= 2.27.0; any other executable, an older or unknown version keeps them resident. To override the never-defer list, set ASBX_KIRO_MANDATORY_MCPS (comma-separated server names) in the gateway's environment. No effect on any other ACP backend.",
"hasChildren": false,
"enumValues": null,
"defaultValue": true
},
{
"path": "agent.tool_search_min_pct",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Tool Search threshold (% of context)",
"help": "Start deferring MCP tool specs once they exceed this percentage of the context window. Paired with tool_search_min_tokens — whichever is crossed first wins. Below both thresholds every spec is sent directly, so the agent never pays a tool_search round-trip for a small tool set. 0 with tool_search_min_tokens 0 defers always. Clamped to 0-100; matches the kiro-cli default.",
"hasChildren": false,
"enumValues": null,
"defaultValue": 5
},
{
"path": "agent.tool_search_min_tokens",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Tool Search threshold (tokens)",
"help": "Start deferring MCP tool specs once they exceed this many tokens. Paired with tool_search_min_pct — whichever is crossed first wins. 0 with tool_search_min_pct 0 defers always. Matches the kiro-cli default.",
"hasChildren": false,
"enumValues": null,
"defaultValue": 50000
},
{
"path": "agent.session_sharing",
"kind": "core",
"type": "boolean",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Session Sharing",
"help": "Subagents reuse a shared ACP runtime instead of spawning a fresh backend process per subagent. Reduces startup from ~3-5s to ~200ms and memory from ~400MB to near-zero per subagent. Default ON for the eligible backends (kiro-cli and codex); every other ACP backend ignores it and runs a dedicated process per subagent. Set false to opt an eligible backend back onto per-subagent processes.",
"hasChildren": false,
"enumValues": null,
"defaultValue": true
},
{
"path": "agent.max_subagents",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Max SubAgents",
"help": "Maximum amount of subagents at one time. 0 = auto: the subagent_auto_max ceiling, with free memory bounding each start through spawn_min_memory_gb long before that on most hosts (sized from host memory when spawn_min_memory_gb is 0, 3 when host memory cannot be read; see dynamic-subagent-sizing docs). Default; set a fixed cap by pinning an integer >= 3 (values of 1 or 2 are raised to 3). A live edit moves the subagent cap at once; the MCP gateway's spawn-gate ceiling, raised to this figure, follows at the next gateway restart.",
"hasChildren": false,
"enumValues": null,
"defaultValue": 0
},
{
"path": "agent.max_stop_hook_nudges",
"kind": "core",
"type": "integer",
"required": false,
"deprecated": false,
"sensitive": false,
"tags": [],
"label": "Max Stop-hook nudges",
"help": "Maximum consecutive Stop-hook block continuations before the run halts and surfaces a halt card instead of dispatching another turn. Bounds a buggy always-block hook in an unattended session. 0 = uncapped (opt-in for genuinely unbounded feedback loops).",
"hasChildren": false,
"enumValues": null,
"defaultValue": 100
},
{
"path": "agent.spawn_min_memory_gb",
"kind": "core",
"type": "number",
"required": false,
"deprecated": false,
"sensitive": false,