Skip to main content

CrowdStrike Falcon

Triage Falcon detections, investigate hosts and vulnerabilities, pull threat intelligence, and track cases - through your own falcon-mcp deployment.

Connect the CrowdStrike Falcon platform through falcon-mcp, CrowdStrike's official MCP server, which you deploy and hold the Falcon API credentials for. Search and triage detections across EPP, IDP, XDR, OverWatch and NG-SIEM; look up hosts and their Spotlight vulnerabilities; research threat actors, indicators and reports; and read, create and update Falcon case-management records. Yekar.AI never sees your Falcon API keys - the deployment's own credentials and API scopes are the permission ceiling, and requests never leave its host.

Connect

Credential fieldRequiredWhere it comes from
Access token (ACCESS_TOKEN)NoToken for the proxy fronting your falcon-mcp deployment. This is NOT a Falcon API credential - those live in the deployment's own environment. Leave empty for an internal endpoint with no auth.
SettingRequiredWhat it is
Server URL (SERVER_URL)YesYour falcon-mcp deployment's Streamable HTTP endpoint, e.g. https://falcon-mcp.example.com/mcp - every request stays on its host
Auth header name (AUTH_HEADER)NoHeader the access token is sent under (default Authorization, as “Bearer <token>”; a custom name sends the token verbatim)
Request timeout (ms) (TIMEOUT_MS)No

Tools

ToolAccessWhat it does
Check connectivity (yekar.crowdstrike-falcon.check-connectivity)ReadVerify the connection end to end: reaches the falcon-mcp deployment and confirms ITS Falcon API credentials authenticate against CrowdStrike. Run this first - it separates a bad server URL or token (Yekar.AI's side) from bad or unscoped Falcon API credentials (the deployment's side), which are different problems with different owners.
List enabled modules (yekar.crowdstrike-falcon.list-enabled-modules)ReadWhich falcon-mcp modules the connected deployment actually loaded (detections, hosts, spotlight, intel, cases…). A deployment started with --modules exposes only a subset, and every other tool here fails with a not-found when its module is missing - this is how you confirm that diagnosis instead of guessing at an outage.
Search detections (yekar.crowdstrike-falcon.search-detections)ReadTHE TRIAGE QUEUE - when someone asks for open incidents, open alerts, or what needs triage right now, this is the tool. Searches detections across every Falcon product (EPP, IDP, XDR, OverWatch, NG-SIEM) by severity, status, hostname, tactic or time range, returning full alert records with process context, device info and threat classification. (Falcon's separate case-management records are the formal tickets a triaged detection is promoted into - see Search cases.)
Get detection details (yekar.crowdstrike-falcon.get-detection-details)ReadFull records for detection ids you already hold - process tree, device, tactic and technique, resolution state. Use Search detections to find ids by criteria; use this once you have them.
Update detections (yekar.crowdstrike-falcon.update-detections)WriteChange the triage state of one or more detections: status, assignee, resolution tags, analyst comment, console visibility. This WRITES to the Falcon console - the change is visible to every analyst on the tenant and there is no undo beyond setting the fields back.
Search hosts (yekar.crowdstrike-falcon.search-hosts)ReadFind devices in the Falcon estate by hostname, platform, IP, OS version, sensor version, tags or last-seen time. Returns the device ids the detection and vulnerability tools take.
Get host details (yekar.crowdstrike-falcon.get-host-details)ReadFull device records for host ids you already hold - OS and sensor version, agent state, first/last seen, network addresses, groups and tags.
Search vulnerabilities (yekar.crowdstrike-falcon.search-vulnerabilities)ReadFalcon Spotlight vulnerability findings - which hosts are exposed to which CVEs, at what severity, in what remediation state. Filter by CVE, severity, status or host, and request the host_info / remediation / cve facets when you need the detail rather than just the counts.
Search threat actors (yekar.crowdstrike-falcon.search-actors)ReadCrowdStrike intelligence on adversary groups - motivation, origin, targeted industries and countries, known aliases and capabilities. Use it to attribute activity or to brief on an actor named in a detection.
Search threat indicators (yekar.crowdstrike-falcon.search-indicators)ReadCrowdStrike intelligence IOCs - domains, IPs, hashes and URLs with malicious-confidence ratings, first/last seen, and the actors and reports they belong to. This is the intel corpus for enrichment; it is not your tenant's own detections.
Search intelligence reports (yekar.crowdstrike-falcon.search-reports)ReadCrowdStrike's published threat reports and intelligence publications - titles, summaries, targeted sectors and the actors they cover. Use it for background on a campaign; the report bodies themselves may require a higher intel subscription tier on the deployment's API client.
Get MITRE ATT&CK report (yekar.crowdstrike-falcon.get-mitre-report)ReadThe MITRE ATT&CK tactics and techniques CrowdStrike attributes to one threat actor - the actor's tradecraft laid out against the framework, for mapping observed behaviour to known TTPs.
Search cases (yekar.crowdstrike-falcon.search-cases)ReadFalcon case-management records - the formal tracked entity (name, severity 1–100, status, assignee, tags, attached evidence) that a triaged detection is promoted into. Use it for investigation and workload bookkeeping: what is being worked, by whom, at what severity. This is NOT the live alert queue - for what needs triage right now, use Search detections.
Get cases (yekar.crowdstrike-falcon.get-cases)ReadFull case-management records for case ids you already hold - status, severity, assignee, tags, version, and the alert and event evidence attached to each.
Create case (yekar.crowdstrike-falcon.create-case)WritePromote triaged work into a tracked Falcon case: create a case-management record and attach the detections (alertIds) or LogScale events that justify it. This WRITES to the Falcon console - the case is visible to every analyst on the tenant, and calling twice creates two cases.
Update case (yekar.crowdstrike-falcon.update-case)WriteChange a Falcon case-management record's fields: title, description, status, severity, assignee. This WRITES to the Falcon console. Pass expectedVersion from the case you read to take the optimistic lock - without it, a concurrent edit by another analyst is overwritten with no warning.
Get FQL guide (yekar.crowdstrike-falcon.fql-guide)ReadThe filter grammar for one search tool - the fields it accepts, their types, and worked examples, straight from the deployment rather than from memory. FQL is where these searches go wrong; read the guide before writing a non-trivial filter.

Notes

  • This integration needs a running falcon-mcp deployment (github.com/CrowdStrike/falcon-mcp) - CrowdStrike does not host one for you. Start it with --transport streamable-http and give it FALCON_CLIENT_ID, FALCON_CLIENT_SECRET and FALCON_BASE_URL for your region.
  • It must be reachable at a public HTTPS hostname: falcon-mcp binds 127.0.0.1 by default, and private addresses are refused, so put it behind a reverse proxy or tunnel and point Server URL at the endpoint (usually ending in /mcp).
  • The deployment's Falcon API client is the permission ceiling - Yekar.AI never holds those credentials, so a missing API scope is fixed in the Falcon console and by restarting the deployment, not here. Protect the endpoint itself with a proxy token and paste it as the Access token.
  • One deployment serves one CID: falcon-mcp fixes member_cid at startup, so an MSSP parent covering several child CIDs needs one deployment and one connection per CID.
  • Run Check connectivity first, then List enabled modules - a deployment started with --modules exposes only a subset, and tools whose module is missing fail with a not-found naming it. A deployment running --dynamic exposes no named tools at all and will not work with this integration.
  • For the live alert queue - what needs triage now - use Search detections. Cases are the formal records a triaged detection is promoted into, not the queue itself.