Everything an AI agent can do with the CrowdStrike API.

A reference guide for building AI agents: every method, how to authenticate, and the permissions each one needs.

Endpoints33
API versionv1
Last updated23 June 2026
Orientation

How the CrowdStrike Falcon API works.

The CrowdStrike API is how an app or AI agent works with a Falcon tenant: searching alerts, reading host details and containing a compromised machine, managing custom indicators, and running commands on a live endpoint. Access is granted through an API client with a secret and a fixed set of scopes, each read or write, so an agent is limited to the areas and actions that client was given. Falcon can also stream detections and audit events to a connection it holds open, rather than the agent repeatedly asking.

33Endpoints
9Capability groups
17Read
16Write
16Permissions
Authentication
Every call needs a bearer token obtained through OAuth 2.0 client credentials. An API client is created in the Falcon console with a client ID and secret, and the client posts both to the token endpoint at /oauth2/token to receive a token that lasts about 30 minutes. The token is then sent as 'Authorization: Bearer ' on each request. The token endpoint has a much lower rate limit than ordinary calls, so a token is reused until it expires.
Permissions
Access is controlled by API scopes ticked on the API client at creation, each set to read or write independently, such as Alerts read, Hosts write, IOC Manager write, or Real time response write. A call to a write endpoint with only read ticked returns 403. Scopes are fixed when the client is created and cannot be raised at runtime, so the client's scope list is the hard boundary on what an agent using it can do. Real Time Response splits across three scopes by command power: read-only, active responder, and admin.
Hosts and regions
The API answers at a region-specific host that matches the tenant's cloud, such as https://api.crowdstrike.com for US-1, https://api.us-2.crowdstrike.com for US-2, https://api.eu-1.crowdstrike.com for EU-1, and a separate host for US-GOV-1. The token endpoint and every call use that same regional host, and an integration must target the region its Falcon tenant lives in.
Data model
Most areas follow a two-step pattern: a queries endpoint returns the IDs that match a Falcon Query Language filter, then an entities endpoint takes those IDs and returns full records, with a combined endpoint sometimes doing both in one call. Errors come back as an errors array of code and message alongside the HTTP status. The Detects API was decommissioned on 30 September 2025 and its work moved into the unified Alerts API.
Connect & authenticate

Connection & authentication methods.

How an app or AI agent connects to CrowdStrike Falcon determines what it can reach. An API client is created in the Falcon console with a client ID and secret, and the API scopes ticked on that client, each set to read or write, fix what every call is allowed to do.

Ways to connect

REST API

The REST API answers at a region-specific host, such as https://api.crowdstrike.com for US-1 or https://api.eu-1.crowdstrike.com for EU-1. Each endpoint carries its own path version, like v1 or v2.

Best forConnecting an app or AI agent to CrowdStrike Falcon.
Governed byThe API client's ID and secret and the scopes ticked on it.
Docs ↗

Falcon MCP server (Model Context Protocol)

CrowdStrike's Falcon MCP server lets an agent call Falcon through the Model Context Protocol, exposing detections, incidents, hosts, vulnerabilities, and threat intelligence. It is maintained by CrowdStrike and is in public preview, so production use is not yet advised. The source is at github.com/CrowdStrike/falcon-mcp.

Best forConnecting an AI agent to CrowdStrike Falcon over the Model Context Protocol.
Governed byThe API client's ID and secret and the scopes ticked on it.
Docs ↗

Event Streams

Event Streams holds a long-lived connection open and pushes detections, audit records, authentication activity, and incident updates as they happen, resuming from an offset after a drop. It is started from the datafeed listing endpoint.

Best forReceiving Falcon events in near real time without polling.
Governed byThe API client's ID and secret and the Event streams scope.
Docs ↗

Falcon Fusion SOAR

Falcon Fusion SOAR workflows can post chosen alerts and detections to a receiver URL through a webhook plugin, and can call Falcon APIs from within the workflow without separate credentials.

Best forSending selected alerts to an external receiver on a workflow trigger.
Governed byThe workflow definition and the tenant context it runs under.
Docs ↗
Authentication

OAuth 2.0 client credentials

An API client is created in the Falcon console with a client ID and secret, and the API scopes ticked on it, each set to read or write. The client posts its ID and secret to the token endpoint to receive a bearer token that lasts about 30 minutes, then sends that token on each call.

TokenBearer token from client ID and secret
Best forConnecting an app, integration, or AI agent to Falcon
Docs ↗
Capability map

What an AI agent can do in CrowdStrike Falcon.

The Falcon API is split into areas an agent can act on, such as alerts, hosts, incidents, custom indicators, vulnerabilities, and Real Time Response. Each area maps to its own API scope, and some grant the ability to act on a live machine rather than only read data.

Alerts & detections

4 endpoints

Query alert IDs by filter, aggregate alerts, pull full alert details, and update an alert's status or assignment. This is the unified surface that replaced the older Detects API.

Writes here change the status and assignment of real alerts.
View endpoints

Hosts & devices

3 endpoints

Search managed devices by filter, read full device details, and run a containment action that cuts a host off from the network or lifts that containment.

Containment cuts a real machine off from the network.
View endpoints

Incidents

3 endpoints

Search incidents and their behaviours, read incident details, and perform incident actions like tagging, commenting, or renaming.

Writes here change real incident records, and these endpoints are being retired.
View endpoints

Custom indicators (IOCs)

5 endpoints

Query, read, create, update, and delete custom indicators of compromise, the hashes, domains, and addresses Falcon watches for and acts on.

Writes here change what Falcon detects and blocks across the estate.
View endpoints

Spotlight vulnerabilities

3 endpoints

Query vulnerability IDs by filter, read vulnerability details combined with affected hosts, and pull remediation records.

Reads here expose the unpatched weaknesses across the estate.
View endpoints

Real Time Response

4 endpoints

Open a remote session to a live host, run read-only or active-responder commands such as listing or deleting files, and manage scripts and put-files for those sessions.

These run commands directly on a live machine.
View endpoints

Prevention policies

4 endpoints

Query and read prevention policies, create and update them, assign or unassign host groups, and set policy precedence.

Writes here change how endpoints are protected across the estate.
View endpoints

Users

5 endpoints

Query user IDs, read user records, and create, update, or delete users and their role assignments.

Writes here change who can sign in to Falcon and what they can do.
View endpoints

Event streams

2 endpoints

List the available event feeds for the tenant and refresh an active streaming session so it stays open.

This feed carries the tenant's detection and audit events.
View endpoints
Endpoint reference

Every CrowdStrike Falcon API method.

Filter by method, access, or permission, or search any path. Select a row for version detail, rate limits, the related webhook event, and the source.

MethodEndpointWhat it doesAccessPermissionVersion

Alerts & detections

Query alert IDs by filter, aggregate alerts, pull full alert details, and update an alert's status or assignment. This is the unified surface that replaced the older Detects API.4

Falcon Query Language is CrowdStrike's filter syntax, used here to narrow alerts by fields like status, severity, or tactic. Console scope: Alerts (read).

Acts onalert
Permission (capability)Alerts: read
VersionAvailable since the API’s base version
Webhook eventDetectionSummaryEvent
Rate limitStandard limits apply

Returns counts grouped by a chosen field rather than individual alerts. Console scope: Alerts (read).

Acts onalert
Permission (capability)Alerts: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

The IDs come from the alerts query endpoint. This is a POST because the ID list travels in the request body. Console scope: Alerts (read).

Acts onalert
Permission (capability)Alerts: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Actions are applied to the listed alert IDs, for example setting status to in_progress or assigning an owner. Console scope: Alerts (write).

Acts onalert
Permission (capability)Alerts: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Hosts & devices

Search managed devices by filter, read full device details, and run a containment action that cuts a host off from the network or lifts that containment.3

Filters on device fields like hostname, platform, or last seen. Console scope: Hosts (read).

Acts ondevice
Permission (capability)Hosts: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Returns each device's operating system, agent version, network details, and policy state. Console scope: Hosts (read).

Acts ondevice
Permission (capability)Hosts: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

The action_name parameter selects contain or lift_containment. Containment cuts the host off from the network while keeping the Falcon agent connected. Console scope: Hosts (write).

Acts ondevice
Permission (capability)Hosts: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Incidents

Search incidents and their behaviours, read incident details, and perform incident actions like tagging, commenting, or renaming.3

The legacy incident API is deprecated and CrowdStrike plans to remove it in March 2026, with incidents surfacing through the Alerts API instead. Console scope: Incidents (read).

Acts onincident
Permission (capability)Incidents: read
VersionDeprecated 2024-10-01 · Sunset 2026-03-31
Webhook eventNone
Rate limitStandard limits apply

Part of the deprecated incident API, planned for removal in March 2026. Console scope: Incidents (read).

Acts onincident
Permission (capability)Incidents: read
VersionDeprecated 2024-10-01 · Sunset 2026-03-31
Webhook eventNone
Rate limitStandard limits apply

Part of the deprecated incident API, planned for removal in March 2026. Console scope: Incidents (write).

Acts onincident
Permission (capability)Incidents: write
VersionDeprecated 2024-10-01 · Sunset 2026-03-31
Webhook eventNone
Rate limitStandard limits apply

Custom indicators (IOCs)

Query, read, create, update, and delete custom indicators of compromise, the hashes, domains, and addresses Falcon watches for and acts on.5

An indicator of compromise is a hash, domain, IP address, or URL that Falcon watches for. Console scope: IOC Manager (read).

Acts onindicator
Permission (capability)IOC Management: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Console scope: IOC Manager (read).

Acts onindicator
Permission (capability)IOC Management: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

The action set on an indicator decides whether Falcon only alerts on it or blocks it across the estate. Console scope: IOC Manager (write).

Acts onindicator
Permission (capability)IOC Management: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Console scope: IOC Manager (write).

Acts onindicator
Permission (capability)IOC Management: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Removing an indicator stops Falcon detecting or blocking on it. Console scope: IOC Manager (write).

Acts onindicator
Permission (capability)IOC Management: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Spotlight vulnerabilities

Query vulnerability IDs by filter, read vulnerability details combined with affected hosts, and pull remediation records.3

Spotlight is Falcon's vulnerability-management module. Console scope: Vulnerabilities (read).

Acts onvulnerability
Permission (capability)Vulnerabilities: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Combines the query and detail steps, and can enrich each vulnerability with its affected host. Console scope: Vulnerabilities (read).

Acts onvulnerability
Permission (capability)Vulnerabilities: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Console scope: Vulnerabilities (read).

Acts onremediation
Permission (capability)Vulnerabilities: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Real Time Response

Open a remote session to a live host, run read-only or active-responder commands such as listing or deleting files, and manage scripts and put-files for those sessions.4

Real Time Response opens a remote shell to a live machine. A session is the precondition for running any command. Console scope: Real time response (write).

Acts onsession
Permission (capability)Real time response: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

This is the read-only command tier of Real Time Response, covering commands that inspect a host without changing it. Console scope: Real time response (write).

Acts onsession
Permission (capability)Real time response: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

The active-responder tier runs commands that change a live host, like deleting a file or killing a process. Console scope: Real time response (write).

Acts onsession
Permission (capability)Real time response: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

The admin tier covers the most powerful commands, including running custom scripts, and uses a separate scope from the other two tiers. Console scope: Real time response admin (write).

Acts onsession
Permission (capability)RTR admin: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Prevention policies

Query and read prevention policies, create and update them, assign or unassign host groups, and set policy precedence.4

A prevention policy is the set of protection settings applied to a group of hosts. Console scope: Prevention Policies (read).

Acts onprevention policy
Permission (capability)Prevention policies: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Console scope: Prevention Policies (read).

Acts onprevention policy
Permission (capability)Prevention policies: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Changing a policy changes how every host group assigned to it is protected. Console scope: Prevention Policies (write).

Acts onprevention policy
Permission (capability)Prevention policies: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

The action_name parameter selects add-host-group, remove-host-group, enable, or disable. Console scope: Prevention Policies (write).

Acts onprevention policy
Permission (capability)Prevention policies: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Users

Query user IDs, read user records, and create, update, or delete users and their role assignments.5

Console scope: User Management (read).

Acts onuser
Permission (capability)User management: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Returns names, email addresses, and role assignments. Console scope: User Management (read).

Acts onuser
Permission (capability)User management: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Creating a user grants someone access to the Falcon console. Console scope: User Management (write).

Acts onuser
Permission (capability)User management: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Console scope: User Management (write).

Acts onuser
Permission (capability)User management: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Removing a user revokes their access to the Falcon console. Console scope: User Management (write).

Acts onuser
Permission (capability)User management: write
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply

Event streams

List the available event feeds for the tenant and refresh an active streaming session so it stays open.2

This is the entry point to the Event Streams feed. The returned feed URL is held open to receive detections, audit records, and incident updates as they happen. Console scope: Event streams (read).

Acts onstream
Permission (capability)Event streams: read
VersionAvailable since the API’s base version
Webhook eventDetectionSummaryEvent
Rate limitStandard limits apply

A streaming session has to be refreshed periodically or it closes. This uses the same Event streams scope as listing the feeds. Console scope: Event streams (read).

Acts onstream
Permission (capability)Event streams: read
VersionAvailable since the API’s base version
Webhook eventNone
Rate limitStandard limits apply
No endpoints match those filters.
Webhooks

Webhook events.

CrowdStrike can stream events to an app or AI agent instead of the app repeatedly asking. The Event Streams feed holds a long-lived connection open and pushes detections, audit records, and incident updates as they happen, and Falcon Fusion SOAR workflows can post chosen alerts to a receiver URL.

EventWhat it signalsTriggered by
DetectionSummaryEventCarried on the Event Streams feed when Falcon raises a new detection, with the detection's severity, tactic, technique, and the host involved./alerts/queries/alerts/v2
/sensors/entities/datafeed/v2
IncidentSummaryEventCarried on the Event Streams feed when an incident is created or updated, grouping related detections./sensors/entities/datafeed/v2
AuthActivityAuditEventCarried on the Event Streams feed for authentication activity, such as a console sign-in or an API token request./sensors/entities/datafeed/v2
UserActivityAuditEventCarried on the Event Streams feed for administrative changes inside Falcon, such as a policy edit or a user being added./sensors/entities/datafeed/v2
No events match that search.
Rate limits & pagination

Rate limits, pagination & request size.

CrowdStrike limits how fast an app or AI agent can call, per API client and per endpoint, and reports the current state in response headers. Going over returns a 429, and the token endpoint itself is held to a much lower rate than ordinary calls.

Request rate

CrowdStrike rate-limits per API client and per endpoint rather than across the whole API, and does not publish a single fixed number for every endpoint. Each response carries X-RateLimit-Limit, the ceiling per minute, and X-RateLimit-Remaining, the calls left in the current window. When a limit is exceeded the call returns 429 and an X-RateLimit-RetryAfter header gives a UTC timestamp for when the pool will have capacity again. The token endpoint is held to a much lower limit than ordinary calls, so a token should be reused across requests until it expires rather than fetched per call.

Pagination

List endpoints use offset-based pagination through the offset and limit parameters, with a default page size of 100 and a maximum of 500 on most endpoints. Some high-volume endpoints, such as the combined alerts feed, use an after token instead, where each page returns the token for the next page until no token is returned. The total count of matches is returned in the response metadata.

Request size

Requests and responses are JSON. Many list and detail endpoints take IDs in batches, commonly up to a few hundred or a thousand IDs per call, and exceeding a batch limit returns a 400 with the limit named in the errors array. Specific per-endpoint batch ceilings are documented on each endpoint.

Errors

Status codes & error handling.

The status codes an agent should handle, and what to do about each.

StatusCodeMeaningWhat to do
400Bad RequestThe request was malformed, such as an invalid Falcon Query Language filter or a missing required field. The body holds an errors array, where each entry has a code and a message.Read the message in the errors array, correct the named field or filter, and resend.
401UnauthorizedThe bearer token is missing, invalid, or expired. Tokens last about 30 minutes.Request a fresh token from the token endpoint with the client ID and secret, then retry.
403ForbiddenThe token is valid but the API client does not hold the scope the endpoint needs, for example calling a write endpoint with only read ticked.Add the required scope to the API client in the Falcon console, then request a new token.
404Not FoundThe resource does not exist, or the requested ID is not visible to this API client.Confirm the ID and the path, and that the client has access to the resource.
429Too Many RequestsA rate limit was exceeded. The X-RateLimit-RetryAfter header gives a UTC timestamp for when the request pool will have capacity again. The token endpoint has a much lower limit than ordinary calls.Wait until the X-RateLimit-RetryAfter time, then retry with exponential backoff.
500Internal Server ErrorFalcon hit an unexpected error handling the request. The body still carries an errors array with a code and message.Retry with backoff, and if it persists, contact CrowdStrike support with the trace ID from the response.
Versioning & freshness

Version history.

The Falcon API is not versioned as one whole. Each endpoint carries its own version in its path, such as v1 or v2, and a new version ships as a new path while the old one is deprecated on a published timeline before it is removed.

Version history

What changed, and when

Latest versionv1
v1Current version
Per-endpoint path versioning

The Falcon API is not versioned as one whole. Each endpoint carries its own version in its path, most commonly v1 with newer v2 and v3 endpoints alongside, and a new version ships as a new path while the old one is deprecated before removal. There is no single account-wide version to pin.

What changed
  • Alerts: PATCH /alerts/entities/alerts/v3 is the current alert-update endpoint, with v1 and v2 deprecated.
  • Hosts: GET and POST /devices/entities/devices/v2 return device details, replacing the v1 endpoint.
  • Spotlight: remediation details moved to /spotlight/entities/remediations/v2.
2025-09-30Requires migration
Detects API decommissioned

The legacy Detects API was decommissioned on 30 September 2025, having been deprecated on 1 October 2024. Its detection search, aggregate, and detail operations were merged into the unified Alerts API. Integrations had to update their API client scopes and endpoint paths from the detects paths to the alerts paths to avoid a break in data.

What changed
  • Removed the /detects query, aggregate, and entities endpoints.
  • Detection data now served through the /alerts endpoints across all Falcon products.
  • Existing API clients needed the Alerts scope added in place of Detections.
2025-07-01Feature update
Falcon MCP server enters public preview

CrowdStrike published the Falcon MCP server, letting AI agents reach Falcon over the Model Context Protocol for detections, incidents, hosts, vulnerabilities, and threat intelligence. It entered public preview, governed by the same API client scopes as the REST API, with production use advised against until a stable release.

What changed
  • Falcon MCP server released as open source at github.com/CrowdStrike/falcon-mcp.
  • Exposes Falcon detections, incidents, hosts, and intelligence to MCP-aware agents.
  • Reuses OAuth 2.0 client credentials and the existing API scopes.
Earlier
Incident API retirement announced

CrowdStrike announced that the legacy incident API, covering CrowdScore, incidents, and behaviours, is deprecated and planned for removal in March 2026, with incident data surfacing through the Alerts API. Integrations on the /incidents endpoints need to move before then.

What changed
  • The /incidents query, entities, and action endpoints marked deprecated.
  • Incident data consolidating into the unified Alerts surface.
  • Removal targeted for March 2026.

An integration pins each endpoint by its path version and moves up as newer ones ship.

CrowdStrike Developer Center ↗
Questions

CrowdStrike Falcon API, answered.

How does an agent authenticate to the Falcon API?+
Through OAuth 2.0 client credentials. An API client is created in the Falcon console under API clients and keys, which produces a client ID and secret. The client posts both to the token endpoint at /oauth2/token and receives a bearer token that lasts about 30 minutes, then sends that token on each call. The token endpoint is rate-limited far more tightly than ordinary calls, around 15 requests per minute on some clouds, so a token is reused until it expires rather than fetched per request.
How do API scopes work, and why am I getting a 403?+
Each API client is created with a fixed list of scopes, and every scope is set to read or write independently, such as Alerts read or Hosts write. A 403 means the token is valid but the client does not hold the scope the endpoint needs, most often calling a write endpoint such as containing a host with only read ticked. The fix is to add the missing scope to the API client in the Falcon console and request a new token. Scopes cannot be raised at runtime.
What happened to the Detects API?+
CrowdStrike deprecated the Detects API on 1 October 2024 and decommissioned it on 30 September 2025. Its detection search, aggregate, and detail operations moved into the unified Alerts API, which covers alerts from across Falcon, not only endpoint detections. An integration that called the old Detects endpoints needs to move to the Alerts endpoints, which usually means updating the API client's scopes and the endpoint paths.
How does a host get contained through the API?+
Network containment runs through the device actions endpoint, with the action_name parameter set to contain, and lifting it uses lift_containment. Containment cuts the host off from the network while keeping the Falcon agent connected, so the host can still be investigated and released remotely. The call needs the Hosts scope with write, and it acts on the device IDs passed in, so it is one of the highest-impact actions an agent can take.
How do I receive events instead of polling?+
Through Event Streams. Listing the datafeed returns a feed URL and token, and the integration holds that connection open to receive detections, incident updates, authentication activity, and administrative changes as they happen, resuming from an offset after a drop. The session has to be refreshed periodically or it closes. This needs the Event streams scope. Falcon Fusion SOAR can also post chosen alerts to a receiver URL on a workflow trigger.
What are the rate limits?+
CrowdStrike limits per API client and per endpoint rather than across the whole API, and does not publish one fixed number for every endpoint. Each response carries X-RateLimit-Limit and X-RateLimit-Remaining, and exceeding a limit returns 429 with an X-RateLimit-RetryAfter header giving a UTC timestamp for when capacity returns. The token endpoint is the tightest limit, so tokens should be reused rather than minted per call. Honour the retry-after time with exponential backoff.
Does CrowdStrike have an official MCP server for AI agents?+
Yes. CrowdStrike maintains the Falcon MCP server, which lets an agent reach Falcon over the Model Context Protocol for detections, incidents, hosts, vulnerabilities, and threat intelligence. It is in public preview, and CrowdStrike advises against production use until a stable release. It is governed by the same API client scopes as the REST API, so an agent through the MCP server can do only what its client's scopes allow.
Related

More security API guides for agents

What is Bollard AI?

Control what every AI agent can do in CrowdStrike Falcon.

Bollard AI sits between a team's AI agents and CrowdStrike Falcon. Grant each agent exactly the access it needs, read or write, scope by scope, and every call is checked and logged.

  • Set read, write, or full access per agent, never a shared Falcon API client secret.
  • Denied by default, so an agent reaches only what has been explicitly allowed.
  • Every call recorded in plain English: who, what, where, and the decision.
CrowdStrike
SOC Triage Agent
Read alerts and detections ResourceOffReadFull use
Read host details ResourceOffReadFull use
Contain a host ActionOffReadFull use
Run Real Time Response commands ActionOffReadFull use
Per-agent access, set in Bollard AI, not in CrowdStrike