Create policy
Create a policy. Without a placement, a deny goes to the top of the list and any other rule to the bottom.
Create a policy. Without a placement, a deny goes to the top of the list and any other rule to the bottom.
ApiKeyAuthAuthorizationBearer <token>Bearer token authentication. Send an API key (a saai_api_-prefixed token) as Authorization: Bearer <token>.
Required permission
manage_policiesapplication/json- body
One rule in an ordered list, for which agents and users and under what conditions. Tool and model rules decide what may run, and a request that matches none of them is denied by default; a model route decides where a model call runs, and a call that matches no route runs as requested. Rules are evaluated top to bottom and the first match wins.
placement?Where to put the rule. On create, omit it for the default: a deny goes to the top, any other rule to the bottom. On update, omit it to keep the current position.
scope*||Which calls a policy governs. A tool rule must name its tools in tool_patterns or tool_tag_ids; a model rule must name its models in model_patterns; a model route must name its model_patterns.
effect*PolicyEffectWhether a matching rule permits, blocks, gates, or routes the call.
"allow""deny""require_approval""route"status?RuleStatusWhether a rule takes part in evaluation.
"active""draft""disabled"approval_grace_seconds?|How long one approval stays reusable for later calls this rule matches, in seconds. Null uses the default. Only meaningful with the require_approval effect.
0 <= valueapproval_scope?|Which later calls one approval covers during its grace window. Null uses identical_call. Only meaningful with the require_approval effect, and ignored when the grace is 0.
route_targets?array<>|Where calls this route matches run: exactly one target today. Model routes only.
1 <= items <= 1name*stringShort label for the rule, shown in the policy list.
description?|Optional note explaining the rule's intent.
The request has succeeded and a new resource has been created as a result.
application/json- response
One rule in an ordered list, for which agents and users and under what conditions. Tool and model rules decide what may run, and a request that matches none of them is denied by default; a model route decides where a model call runs, and a call that matches no route runs as requested. Rules are evaluated top to bottom and the first match wins.
id*uuiduuidpriority*integerPosition in the evaluation order: lower numbers are checked first. Set it with placement on create or update.
int32scope*||Which calls a policy governs. A tool rule must name its tools in tool_patterns or tool_tag_ids; a model rule must name its models in model_patterns; a model route must name its model_patterns.
effect*PolicyEffectWhether a matching rule permits, blocks, gates, or routes the call.
"allow""deny""require_approval""route"status?RuleStatusWhether a rule takes part in evaluation.
"active""draft""disabled"approval_grace_seconds?|How long one approval stays reusable for later calls this rule matches, in seconds. Null uses the default. Only meaningful with the require_approval effect.
0 <= valueapproval_scope?|Which later calls one approval covers during its grace window. Null uses identical_call. Only meaningful with the require_approval effect, and ignored when the grace is 0.
route_targets?array<>|Where calls this route matches run: exactly one target today. Model routes only.
1 <= items <= 1name*stringShort label for the rule, shown in the policy list.
description?|Optional note explaining the rule's intent.
agent_instance_refs?array<>Running agent instances this rule's scope currently resolves to.
user_refs?array<>Users this rule's scope currently resolves to.
created_at*stringdate-timeupdated_at*stringdate-timecurl -X POST 'https://your-org.aisecurity.services.us.connect.secureauth.com/api/v1/policies' \ -H 'Content-Type: application/json' \ -d '{ "scope": { "type": "mcp_call" }, "effect": "allow", "name": "string"}'{ "id": "497f6eca-6276-4993-bfeb-53cbbbba6f08", "priority": 0, "scope": { "type": "mcp_call", "agent_slugs": [ "string" ], "agent_instance_ids": [ "bc3f18ff-0dcd-4cc7-8f61-6ca61814f10b" ], "user_ids": [ "2d98503d-4ab4-41bd-8fc6-78cc006fd2db" ], "groups": [ "string" ], "agent_slug_tag_ids": [ "e428595a-d9dd-4887-8327-2d9cad5b7425" ], "autonomous": true, "mcp_ids": [ "c2c9208c-a7ac-41cc-84f1-8c1caec92987" ], "mcp_tag_ids": [ "17d4a46c-d971-4171-9a66-88ec59ce6eeb" ], "tool_tag_ids": [ "41068329-9b4f-4cc3-b249-3ec2a7a2cee8" ], "tool_patterns": [ "string" ], "condition": "string", "condition_description": "string" }, "effect": "allow", "status": "active", "approval_grace_seconds": 0, "approval_scope": "identical_call", "route_targets": [ { "model": "string", "model_provider_id": "e8dd39c4-1070-4c00-8ab5-c345bdef77c9" } ], "name": "string", "description": "string", "agent_instance_refs": [ { "id": "497f6eca-6276-4993-bfeb-53cbbbba6f08", "display_name": "string", "user_display_name": "string" } ], "user_refs": [ { "id": "497f6eca-6276-4993-bfeb-53cbbbba6f08", "display_name": "string" } ], "created_at": "2019-08-24T14:15:22Z", "updated_at": "2019-08-24T14:15:22Z"}