Skip to content

docs(crowdstrike): name each tool and the API scope it needs - #464

Open
leet-c1 wants to merge 4 commits into
mainfrom
leet/crowdstrike-iom-scopes
Open

docs(crowdstrike): name each tool and the API scope it needs#464
leet-c1 wants to merge 4 commits into
mainfrom
leet/crowdstrike-iom-scopes

Conversation

@leet-c1

@leet-c1 leet-c1 commented Aug 20, 2026

Copy link
Copy Markdown
Contributor

What this fixes

The page told admins to grant "Read access to Detections, Incidents, and Hosts". The server exposes 14 read tools across nine different Falcon APIs, so that sentence leaves most of them unauthorized and names one product ("Detections") that no tool actually reads.

Changes

A Choose API scopes section listing every tool by name against the Falcon API scope it needs — list_hosts, list_alerts, list_incidents, list_ioc_indicators, list_device_control_policies, list_vulnerabilities, list_remediations, list_intel_actors, list_intel_indicators, list_intel_reports, list_zero_trust_assessments, list_discover_applications, list_ioms, list_iom_details. All 14 are covered.

The groupings are not editorial. They follow the CrowdStrike spec's own operation tags: list_vulnerabilities and list_remediations share a row because both endpoints are tagged spotlight_vulnerabilities, and the three intel tools share one because all three are tagged intel.

The CSPM registration point, in a <Note>. list_ioms and list_iom_details need CSPM registration; CrowdStrike's Cloud Security scope covers cloud groups and cloud risks — endpoints this server does not expose — so reaching for it by name will not make these two tools work. Both IOM endpoints are tagged cspm_registration in the spec, which is where the mapping comes from.

The write scopes, which the page previously implied were unnecessary. contain_host, lift_containment and hide_host need Write on Hosts; create_ioc_indicator needs Write on IOC Manager. Also states the useful inverse: leaving Write off is a deliberate way to keep the connection read-only, since those four tools then fail and the rest keep working.

Scope-change timing — edits apply on the next token request, so admins do not need to re-register the server or reconnect.

Where this came from, accurately

A customer asked for IOM coverage and mentioned they would grant the Cloud Security API Detections read scope. Separately, they reported not seeing the IOM tools at all after syncing.

Those were two different problems and only one of them is a docs problem. The missing tools were a generator bug — a resource-type collision that dropped the endpoints before they ever reached the surface — fixed in mcp-axiomatic and unrelated to scopes. The tools are also named list_ioms and list_iom_details, not the list_query_ioms/list_entity_ioms the request guessed at, which is worth knowing if someone is searching their tool list for those strings.

The scope guidance here is preventive rather than a post-mortem: the scope they named would not have authorized these tools, so it is documented before it costs anyone time. An earlier draft of this PR described an admin hitting 403s with every other tool working — that did not happen, and I have removed it.

Notes for review

  • The tool names are exact, read from the server's own surface. The Falcon console scope labels in the right-hand column are not verified — I have no tenant to check them against, and a REVIEWER NOTE in the page says so. The mapping behind each row is evidence-based (spec operation tags); it is the console's display strings that need a human with Falcon access.
  • npx mint broken-links reports 12 broken links in 5 files, all pre-existing (baton/bamboohr.mdx images, /docs-prefixed links in baton/sendgrid.mdx, baton/twilio-segment-v2.mdx and index.mdx, and the connector-template.mdx placeholder). None are on this page.
  • The page was already registered in docs.json, so the nav is unchanged. Both in-page links to #choose-api-scopes resolve.

The page told admins to grant "Read access to Detections, Incidents, and Hosts",
which does not map to the tools C1 actually exposes and leaves the most
confusing case undocumented.

Adds a Choose API scopes section mapping each tool group to its Falcon API
scope, covering hosts, alerts, incidents, indicators of compromise, device
control policies, Spotlight vulnerabilities, Falcon Intelligence, Zero Trust
Assessment, Falcon Discover, and cloud misconfigurations.

Calls out the trap behind a real support question: the cloud misconfiguration
tools need CSPM registration, not Cloud Security API Detections. The two scopes
authorize different Falcon APIs, so an admin who enables the latter sees 403s on
those tools while everything else works.

Also documents the write scopes the response tools need (Hosts and IOC Manager),
since the previous text implied Read was sufficient for the whole surface, and
notes that a read-only API client is a deliberate way to keep response actions
unavailable.

Scope edits apply on the next token request, so admins do not need to
re-register the server.
@mintlify

mintlify Bot commented Aug 20, 2026

Copy link
Copy Markdown
Contributor

Preview deployment for your docs. Learn more about Mintlify Previews.

Project Status Preview Updated (UTC)
conductorone 🟢 Ready View Preview Aug 20, 2026, 6:20 PM

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The scope tables labelled their rows with invented category names —
"Cloud misconfigurations (IOM)", "Threat actors, indicators, and
reports", "Indicators of compromise" — none of which appear in the
product, the server's tool list, or the Falcon console. A reader with 14
tools in front of them had no way to match a row to a tool, which is the
one thing the table exists to do.

Rows are now the tool names the server actually exposes, taken from the
impl's own surface: list_hosts, list_alerts, list_incidents,
list_ioc_indicators, list_device_control_policies, list_vulnerabilities,
list_remediations, list_intel_actors, list_intel_indicators,
list_intel_reports, list_zero_trust_assessments,
list_discover_applications, list_ioms, list_iom_details, and the four
write tools contain_host, lift_containment, hide_host and
create_ioc_indicator. All 14 read tools are covered.

The groupings are not editorial: they follow the CrowdStrike spec's own
operation tags, so list_vulnerabilities and list_remediations share a row
because both endpoints are tagged spotlight_vulnerabilities, and the
three intel tools share one because all three are tagged intel.

The note said the IOM tools need CSPM registration "not Cloud Security
API Detections", naming a console string I had not verified, and
referred to the tool as crowdstrike_list_ioms — a prefix this server does
not apply. It now names list_ioms and list_iom_details and says what is
actually checkable: the Cloud Security scope covers cloud groups and
cloud risks, endpoints C1 does not expose, so it cannot make these two
tools work. The CSPM registration mapping itself is evidence-based — both
IOM endpoints are tagged cspm_registration in the spec.

The reviewer note now says exactly which column is verified and which is
not: the tool names are exact, the Falcon console scope labels still need
a check against a real tenant.
@leet-c1 leet-c1 changed the title docs(crowdstrike): document the API scopes each tool group needs docs(crowdstrike): name each tool and the API scope it needs Aug 24, 2026
@leet-c1
leet-c1 requested a review from vipulgowda August 25, 2026 19:01
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

@mindymo mindymo left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed from a docs style perspective, and a few changes pushed. Good to go on my end, but not sure if the open questions need to be resolved before merge.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants