Access Grants

So “how do I give X access”, “who has access here”, “what can X do”, and “how is grant <id> scoped” are all answered from — and acted on through — the Space’s grants. One authorization error is not a grant question: a Space that was just created successfully and fails the moment anyone uses it has a space access role whose trust policy create-space never verified — see the first callout under Troubleshooting and answer it from spaces-and-domains.md.

This file covers both halves: configuring grants (create, narrow, change, revoke) and looking them up (who has access, at what level, how a specific grant is scoped).

Grants govern the Omni plane only: log groups a Space’s data came from stay readable under their own IAM (see Security considerations).

Always state these, in any answer drawn from this file:

• Only four permissions are requestable: READ, READ_WRITE_DELETE, SPACE_ADMIN, CUSTOM. There is no caller-usable WRITE or DELETE, and ADMIN cannot be granted on a Space through the API at all.
• SPACE_ADMIN is a delegation level — it lets the holder manage grants in the Space (grant access to others) on top of read, write, and delete. It is not a higher read level, nor a higher data-access tier — what it adds over READ_WRITE_DELETE is delegation.
• “Make X an admin” is answered in this order, every time, even when the customer asks you not to pause: (1) the first sentence is the question — what does X need to do: read and change data, or grant other people access?; (2) READ_WRITE_DELETE is the recommendation, because that is what “admin” almost always means; (3) SPACE_ADMIN only if X must manage others’ access, and it is delegation, not a higher data tier; (4) ADMIN cannot be granted on a Space through the API; (5) the caller creating the grant needs SPACE_ADMIN on the Space or an administrative grant on the Domain (see Permission levels).
• Resolve the principal before any action — grant, upgrade, revoke, or lookup. A person or team named by alias, username, or email is not a principalId. Run search-principals --domain-id … --search-query <name> and use the returned principalId (UUID) with principalType from the result (USER → IDC_USER, GROUP → IDC_GROUP). Stop if it does not resolve to exactly one principal. An IAM identity is named by its ARN, and the type follows the ARN (:user/ → IAM_USER, :role/ → IAM_ROLE, :root → IAM_ROOT). Never invent an ARN or a UUID, and never assume IAM_USER for a name you have not resolved.
• create-access-grant needs all of --domain-id, --space-id, --name, --principal, and --permission. --name is required and the call is rejected without it — never show a create command that omits it. A grant is not done when the create returns: read it back, then tell the customer in their own words who can now do what (“test-user can now read telemetry in this Space”), and that it applies at the next authorization-cache turnover, so a session already open may need to be re-established before it shows.
• Check before you create: list-access-grants --space-id … --principal-id … --principal-type … first. A second grant at the same named level conflicts.
• Access is deny-by-default for people and profiles: an Identity Center user, console session, or Access Profile with no matching grant can do nothing in the Space. The one exception is a direct IAM caller in the owning account, which falls through to its own IAM policy when no grant matches (see Grants, IAM, and Access Profiles). Access is opened deliberately, one grant at a time.
• Grants do not expire. There is no expiry field on a grant; it stays in effect until it is revoked. Revocation is not instantaneous: each Omni call is re-authorized against the grants, so a revoke takes effect at the next authorization-cache turnover (minutes) — even though credentials already issued remain valid until they expire (up to an hour) or the session is re-established.
• There is no update operation. For a level change (up- or downgrade), create the new-level grant first, then delete the old one — different levels coexist, so there is no gap and no lockout. For a same-level scope or name change, revoke first (a duplicate conflicts), accepting a brief gap. Either way the replacement has a new grant ID to update everywhere the old one was recorded.
• list-access-grants is paginated: follow nextToken until it is absent, and say how many grants you read. An empty list is a successful finding; an error (AccessDeniedException) is inconclusive, never “nobody has access”.
• ResourceNotFoundException from get-access-grant is a finding, not a failure: no grant has that ID — it was mistyped or already revoked. List the Space’s grants to find the right ID; never retry with a guessed one. Whenever you explain this error, also say that delete-access-grant is idempotent — deleting a grant that no longer exists succeeds rather than raising it — because “should I just delete it?” is the customer’s next question.
• Identity Center principals appear as UUIDs. Resolve names with search-principals in the Space’s Domain — not the Identity Center console or the identitystore CLI. An IDC_GROUP grant reaches every member and shows only the group ID.
• An alert principal is identified by its ARN, never its name (resolve with list-alerts), and a grant with principalId ALL covers every alert.
• Cross-account IAM grants are not supported at the Space grant level; the principal must be in the caller’s account. When the customer’s goal is cross-account access, lead with the organization Domain path (get-space-credentials-for-organization vends credentials for a target account — see org-domains.md); do not present an IAM assume-role hop or a different permission level as the answer.
• scopedActions works with any permission, and does a different job in each case: with CUSTOM it defines the grant (only the listed actions are allowed); with a named permission such as READ it narrows that permission to the listed resources, and every action named must be within that permission’s tier — nothing mutating can be scoped under READ.
• A scope’s shape and validation, all of which belong in any answer about a rejected scopedActions: resources is a plain list of scopes (not an object with a scopes member), each with a required resourceType and up to five optional resourceArns (DataSet is the one type that takes none — it is narrowed by rows); resourceType is a plain string validated per (action, resourceType) pair, case-sensitively — an unknown spelling such as Dataset is rejected with a message that lists the accepted values (DataSet), while a known type paired with an action that does not operate on it is rejected with a message that lists nothing; and both Space and Domain are invalid inside a scope — omit resources to cover the whole Space.
• Within a single scopedActions entry, all actions must share the same prefix; use a separate entry per prefix.
• Whenever you show a scoped grant, say that resources inside a scopedActions entry is a plain list of scopes, not an object with a scopes member. There is no top-level resources or customActions on a grant — scopedActions is the only scope field.
• Access questions are answered from the grants — never by pivoting to IAM identity or IAM policy APIs (not even sts get-caller-identity), and never by sending the customer to the console when the CLI answers directly.

Where to look, by kind of question

Every access question is one of five kinds. Find the kind, read that section, and apply the rules in the block above — they hold for all five.

Contents

• Where to look, by kind of question
• What an access grant is
  • Grants, IAM, and Access Profiles
  • What a grant reaches
• Permission levels
  • When the customer asks for an “admin”
• Operations you will call
• Prerequisites
• Granting access
  • Cross-account access
  • Step 1 — Interview the customer first
  • Step 2 — Identify the principal
  • Step 3 — Create the grant
  • Step 4 — Narrow the grant with scoped actions
  • Step 5 — Verify, then report
• Changing a grant
• Revoking a grant
• Looking up access
  • Who has access to the Space
  • What a given principal holds
  • Finding a person by name or alias
  • How a specific grant is scoped
  • Reporting a lookup
• Troubleshooting
• Gotchas — expectations that look like faults
• Security considerations
• Additional resources

What an access grant is

The model is Domain → Space → grant. A Domain is the identity boundary (how members sign in — IAM and/or IAM Identity Center); a Space is the account-scoped boundary for Omni resources; grants are written inside a Space. See aws-observability → references/cloudwatch-omni/concepts.md for how the pieces connect and spaces-and-domains.md for creating a Domain and Space.

Every grant names three things, and every lookup answers a question about one of them:

• A principal — principalType plus, normally, principalId. What principalId holds depends on the type:

  principalType	principalId
  IDC_USER, IDC_GROUP	The Identity Center user or group ID (a UUID, not an alias). Verified against the Domain’s identity store
  IAM_USER, IAM_ROLE, IAM_ROOT	The IAM ARN. Must be in the caller’s account
  ACCESS_PROFILE	The Access Profile’s ID — see access-profiles.md
  ALERT	An alert’s ARN — the alertArn that list-alerts returns — or the reserved, case-sensitive token ALL meaning every alert in the Space. Never the alert’s name or its alertId. ALL is accepted only for ALERT, and only on a CUSTOM grant whose single action is AssumeAccessProfile; any other use is rejected (“Wildcard principal identifier is not supported for this principal type”). A profile that alerts created later must assume needs ALL: CreateAlert authorizes against the wildcard alert/* ARN, which only ALL matches
  AGENT	An agent workload principal — a service-defined identifier for the agent, not an IAM ARN

• A permission — one of READ, READ_WRITE_DELETE, SPACE_ADMIN, or CUSTOM. See Permission levels.

• A scope — the actions and resources a CUSTOM grant allows, or the narrowing applied to a named permission. Scope is not on a list summary; it is only on a single grant’s detail (get-access-grant).

A grant also carries a required name, and its detail records who created it and when — createdBy, createdAt, updatedAt. It carries no expiry: there is no expiry field on either the summary or the detail. Never tell a customer a grant expires or report an expiry time; if asked when access ends, say grants do not expire and are removed by revoking them.

Grants, IAM, and Access Profiles

Three mechanisms sit near each other and are routinely confused. Keep them apart:

• Grants are not IAM, and they bind the two caller types differently. A human identity — an IAM Identity Center user, a console session, or a workload that has assumed an Access Profile — must hold a matching grant: with none, the Space denies the call regardless of IAM. An IAM principal calling the API directly with its own credentials (no Omni session context) is treated as a machine identity, for which a grant is optional: grants that match it are enforced (their row and column scopes apply, and an explicit deny wins), but when none match the Space does not deny — the request falls through to the caller’s IAM policy, which alone decides, with no row restriction applied. Direct IAM callers reach only Spaces owned by their own account. So “who has access to this Space” is the grants for people and profiles, plus IAM policy for direct IAM callers that hold no grant — check both.
• Grants are not Access Profiles. A profile bounds what a workload may do; it is itself the principal of the grants that describe its permissions, and the target of trust grants that say which workloads may assume it. Give a person or a team access with a grant directly — never by creating a profile for them.
• Deny-by-default. A human principal or profile with no matching grant can do nothing in the Space (a direct IAM caller with no grant is governed by IAM, above). There is nothing to “lock down” after creating a Space; there is only access to open, one grant at a time.

What a grant reaches

A grant is scoped to one principal in one Space. It does not span Spaces, Regions, or accounts.

Limits worth stating before the customer designs their access model, because they surface as conflicts rather than as advice:

• 50 grants per principal, and 500 grants per Space.
• 10 CUSTOM grants per principal per Space. Named-permission grants are additionally one per principal per Space per level — a second READ grant for the same principal conflicts rather than replacing the first.
• A grant enforces at most 10 scoped-action entries, 25 actions per entry, and 50 actions per grant in total, with up to 10 resource scopes per entry and 5 resourceArns per scope — report these as the effective limits. (The API schema’s ceilings are higher, but the service rejects anything above the effective limits, so only the effective limits are worth telling a customer.)

Constraints:

• You MUST tell the customer that a grant covers one Space only. A customer with Spaces in several Regions needs a grant per Space, and will otherwise read the missing access as a bug.
• You MUST NOT create a second grant at the same named permission for a principal already holding one. Read the existing grant instead — see Step 2.
• You MUST NOT create a Space grant when the customer describes organization-wide administration. A grant here names exactly one Space; a Domain grant is a different operation in org-domains.md.

Permission levels

The permission decides the shape of the whole create call:

Those four are the only values a caller may request. The service rejects anything else with a message naming the supported set. ADMIN exists only through Domain-level administration and cannot be granted on a Space.

SPACE_ADMIN is a delegation level, not a higher read level. It conveys the ability to manage grants in the Space — to give other principals access. The other three levels confer no delegation ability. Someone who simply needs broad access to the data wants READ_WRITE_DELETE instead.

When the customer asks for an “admin”

“Admin” is not a permission, and the most common mistake in this file is granting SPACE_ADMIN because the customer used the word. Find out what the person needs to do — even when answering in a single message, open with that question and then answer both branches:

• Read, write, and delete telemetry and resources in the Space. This is what “admin” almost always means, and it is the recommendation: READ_WRITE_DELETE.
• Give other people access (manage grants). Only then SPACE_ADMIN, which includes read, write, and delete and adds grant management on top. What it adds is delegation, not more data.
• ADMIN cannot be granted on a Space through the API; only READ, READ_WRITE_DELETE, SPACE_ADMIN, and CUSTOM are requestable.

A good opening reads: “It depends on what test-user needs to do. If they need to read, write, and delete in the Space — which is what admin usually means — grant READ_WRITE_DELETE. Grant SPACE_ADMIN only if they also need to give other people access.”

Constraints:

• You MUST put the question — what does the person need to do? — before any permission name, even when the customer asked you not to pause. Asking it and then answering both branches in the same message satisfies both. An answer that opens with SPACE_ADMIN, or that silently assumes one branch, has skipped the step.
• You MUST recommend READ_WRITE_DELETE as the default reading of “admin”, and offer SPACE_ADMIN only for the case where the person must manage others’ access.
• You MUST state that ADMIN is not grantable on a Space, and that the caller creating the grant needs SPACE_ADMIN on the Space or an administrative grant on the Domain.
• The person still has to be resolved to a principalId and principalType before any command is written (Step 2).

Any of the four can additionally be narrowed to particular resources with scopedActions — see Step 4. That is a separate decision from the permission itself.

What a caller may delegate is bounded by what the caller holds:

• A caller with Domain administration may grant any of the four requestable permissions. No caller can grant ADMIN — it is not requestable through the API at all, whatever the caller holds.
• A caller with SPACE_ADMIN on the Space may grant SPACE_ADMIN, READ, READ_WRITE_DELETE, and CUSTOM within it.
• A caller with a lesser grant may not manage grants.
• No caller, at any level, can grant to a principal outside the Space’s own account — cross-account IAM grants are not supported at the Space grant level.

Least privilege is per-grant narrowing: prefer several narrow grants over one broad one, and the narrowest named level over CUSTOM.

Constraints:

• You MUST NOT offer a write-only or delete-only grant. There is no caller-usable WRITE or DELETE permission — READ_WRITE_DELETE is the only value that conveys mutation.
• You SHOULD start from the narrowest named permission that satisfies the request and reach for CUSTOM only when no named level fits. CUSTOM grants carry more ways to be wrong and count against a tighter quota.
• You MUST NOT attempt to grant a permission broader than the caller holds. The rejection names both levels, so read it rather than retrying.

Operations you will call

Every operation is an aws cloudwatchomni <operation> CLI command, with the operation’s input-shape members passed as --flags. Run it through the aws___call_aws tool when that tool is available; otherwise run the same command in a shell. See aws-observability → references/cloudwatch-omni/programmatic-access.md for the service name, signing name, and what to do when the local CLI does not know cloudwatchomni.

aws cloudwatchomni (Omni control-plane):

There is no update operation for a grant. Changing a grant’s level, scope, or name means revoking it and creating a new one — see Changing a grant.

If the API is not reachable from where you are running — aws cloudwatchomni is unknown to the installed CLI, call_aws/boto3 reports the service as unsupported, or the CLI knows the service but cannot resolve its endpoint — say so in one sentence and answer from this file: give the exact commands, in order, with the customer's identifiers filled in, and what each result means. Do not spend the answer hunting for service models, endpoint rule sets, or alternative SDKs on the local machine; that is never what the customer asked.

These five are the only grant operations an access question needs; resolving a principal’s name may additionally use list-alerts or list-access-profiles (Step 2). Do not reach for IAM or STS (iam get-role, iam simulate-principal-policy, sts get-caller-identity, identitystore describe-user) to answer who has access or why a lookup was denied — the grants are the authorization records, and an AccessDeniedException on a grant lookup is answered by the caller’s own grant, not by their IAM identity.

Prerequisites

• The Space’s spaceId and its Domain’s domainId. Both are required on the create. See spaces-and-domains.md if the Space does not exist yet. Lookups need the spaceId; search-principals needs the domainId.
• A SPACE_ADMIN grant on that Space, or an administrative grant on the Domain. A caller holding neither cannot manage grants at all.
• For Identity Center principals, the Domain must use Identity Center, and the principal must exist in the Domain’s identity store. The service verifies this at create time.
• Cross-account IAM grants are not supported at the Space grant level. No caller, at any permission level, can create a Space access grant for a principal outside their own account. Say this plainly rather than implying it by describing the principal as being “in your account”. Scope the claim to Space grants — it is not true of Omni as a whole. The organization Domain path reaches across accounts by a different mechanism: get-space-credentials-for-organization vends temporary credentials for a target account (see org-domains.md). If a customer’s actual goal is cross-account access, point them there instead of telling them it is impossible — and lead with it, rather than mentioning it as a footnote.

Constraints:

• You MUST NOT create grants for principals in another account. Cross-account IAM grants are not supported, and the rejection says so explicitly.
• You MUST NOT present a workaround as the answer to a cross-account ask — neither retrying create-access-grant at a different permission level (the check is on the principal’s account, not the level) nor an IAM assume-role hop through a local role. The organization Domain path is the mechanism that reaches across accounts; describe it first, and describe anything else only if the customer rules that path out.

Granting access

First, confirm the grant belongs at the Space level at all:

Cross-account access

A Space access grant cannot name a principal outside the Space’s own account. The service rejects the attempt with a ValidationException saying cross-account grants are not supported, at every permission level — there is no level, flag, or retry that changes this. Say that plainly, and scope it to Space grants: it is not true of Omni as a whole.

The mechanism that reaches across accounts is the organization Domain. In a Domain shared across an AWS Organization, get-space-credentials-for-organization vends temporary credentials for a target account’s Space to the organization’s management account or a delegated administrator, and create-domain-access-grant-for-organization grants Domain-level administration. When the customer’s goal is cross-account access, point them there — org-domains.md — as the answer, not as a footnote.

Do not offer an IAM assume-role hop through a local role, or a retry at a different permission level, as the way to do this. The first moves the problem into IAM and gives the customer a same-account principal; the second fails identically.

Step 1 — Interview the customer first

Ask these in one message and wait for answers. Do not ask piecemeal, and do not default silently.

1. Which Space? The spaceId, and the domainId it belongs to.
2. Who is getting access? A person or group in Identity Center, an IAM role or user, or a workload such as an alert or an Access Profile.
3. What do they need to do? Map the answer to one of the four permissions in Permission levels. Ask what they need to do, not which permission they want — customers routinely ask for more than the task requires.
4. If CUSTOM: exactly which actions. Wildcards are not accepted, so the list has to be explicit.
5. Should the grant be limited to particular resources? Optional, and only worth raising if the customer has a reason — see Step 4.
6. A name for the grant? Required. Letters, digits, underscores and hyphens; 1–64 characters. Pick one that tells the grant apart when a principal holds several.

Confirm the choices back in one line, then execute.

Constraints:

• You MUST ask what the principal needs to do before proposing a permission. A customer asking for “admin” usually needs READ_WRITE_DELETE, and SPACE_ADMIN lets them re-grant access to others.
• You MUST NOT assume CUSTOM. Reach for it only when the customer’s answer to question 3 does not fit a named level.

Step 2 — Identify the principal

This step is not optional and not specific to granting: every action that names a principal — grant, upgrade, downgrade, revoke, “what does X hold” — starts by turning the name the customer used into an exact principalId and principalType.

If a name resolves to zero or to more than one principal, stop and say so. Do not substitute a principal from the Space’s grant list, do not guess a UUID, and do not fabricate an IAM ARN from the name.

Find an Identity Center principal

If the customer knows the person or group by name rather than by ID, look it up rather than asking them to find it:

Use the returned principalId (the Identity Center UUID) as the grant’s principalId, and map the result’s principalType — USER or GROUP — to IDC_USER or IDC_GROUP. Each result also carries displayName and, for users, userName; there is no email field. A name search returns at most 10 results and does not paginate. To pull the whole directory (for resolving many UUIDs from a grant list), search for *:

Only the * search accepts --max-results (1–50) and --next-token. This operation searches Identity Center only, and only works when the Domain uses Identity Center.

Check what the principal already has

Pass both principalId and principalType — the same identifier can be valid for more than one type, and filtering on both keeps the result unambiguous.

--principal-type also decides which grants you see. For an Identity Center user, --principal-id alone returns the person’s effective grants — their own plus the ones inherited through IDC_GROUP membership (those keep the group as principal); adding --principal-type IDC_USER narrows to the grants made to the user directly. Use both when the question is “does this person already hold their own READ grant”; drop the type when the question is “what can this person do here”.

Read the result before creating anything. If a grant at the requested named permission already exists, report it and stop — a duplicate conflicts rather than replacing it. If the customer wants a different level, see Changing a grant.

Constraints:

• You MUST run this check before create-access-grant. A conflict after the fact is avoidable and confusing.
• You MUST prefer IDC_GROUP over IDC_USER when the customer is describing a team or a role rather than one person. Group membership changes without touching grants.
• If the check cannot be completed — access denied, or an ambiguous response — you MUST treat the result as inconclusive, not negative. Report that and stop. You MUST NOT create a grant on that basis.

Step 3 — Create the grant

A named permission

Substitute READ_WRITE_DELETE or SPACE_ADMIN as the interview decided. --name is required; the call is rejected without it.

A CUSTOM permission

CUSTOM requires scopedActions, which names the actions the principal may perform. Nothing outside that list is permitted.

Action names are prefix:Action — a lowercase service prefix, a colon, then the action name in upper camel case with no underscores or hyphens. Wildcards are not permitted, so every action is named explicitly. All actions within one scopedActions entry must share the same prefix; use separate entries for separate prefixes.

Capture the grant ID from the response. It is what get-access-grant and delete-access-grant take.

Constraints:

• permission is always required. There is no default, and CUSTOM is never inferred from the presence of scopedActions.
• You MUST NOT attempt a wildcard action. Expand the customer’s intent into an explicit list, and if they cannot enumerate it, a named permission is the right answer instead.
• The action prefix MUST match the service’s vendor code. A prefix that does not is rejected, and the message names the expected one.
• With a named permission, every action you name in scopedActions MUST be within that permission’s tier. Naming a mutating action under READ is rejected, and the message names both the action and the permission.
• You MUST NOT grant administrative actions to an ACCESS_PROFILE principal. The service rejects it, and a profile is not the right place for administration.

Step 4 — Narrow the grant with scoped actions

scopedActions does two different jobs, and which one applies depends on the permission:

• With CUSTOM, it defines the grant. The principal may perform exactly the actions named and nothing else.
• With a named permission, it narrows that permission. The actions named are restricted to the resources listed in the same entry, rather than applying across the whole Space.

Each entry requires actions and may add resources. The shape of a scope, which any answer about a rejected scope should restate:

Every scope requires a resourceType and may add:

• resourceArns — up to 5 ARNs. Omit to cover every resource of that type. (One exception: a DataSet scope takes no resourceArns — the service rejects a dataset ARN with Invalid observe resource type in ARN — and is narrowed by rows instead, with signalTypes + rowScopeGroups below.)
• signalTypes and rowScopeGroups — row-level filtering, and only on the DataSet resource type. See below.

resourceType is a plain string, not an enum, and the service validates it per (action, resourceType) pair: a type is accepted only alongside actions that operate on that kind of resource. The types the service knows are AccessGrant, DataStore, DataSet, DataStream, Source, Route, Processor, Integration, OmniIntelligenceRule, AccessProfile, OmniThread, OmniAgentAsset, AccountConfiguration, OmniDashboard, DataSource, Alert, EvaluationJob, Evaluator, OnlineEvaluation, View, IngestionEndpoint.

• Space and Domain are not valid here — neither of them, for the same reason: a grant is already scoped to a single Space or Domain, so a parent-typed scope is rejected with a 400. To cover the whole Space, omit resources instead.

• Casing matters: DataSet is accepted; Dataset is rejected as an unknown type, and that message lists the accepted spellings. A known type paired with an action that does not operate on it is rejected with Resource type X is not valid for action Y, which does not.

• The Console’s scope picker offers only four of these types; its “Dashboard” entry sends OmniDashboard.

resources inside a scopedActions entry is a plain list of scopes — a JSON array — not an object with a scopes member. scopedActions is the only scope field on a grant; there is no top-level resources or customActions.

Row-level scoping, on DataSet only

rowScopeGroups restricts which records a principal sees, rather than which resources — and only which records. It is not field-level redaction: a record the principal is allowed to retrieve comes back whole, with every field in it. It is available only on a scope whose resourceType is DataSet, and signalTypes and rowScopeGroups must appear together — either both or neither.

DataSet names the plane this scope acts on, and it is the only one. Row scoping filters retrieval through the Dataset. It does not restrict the CloudWatch log groups the Dataset was populated from, which remain readable through the native logs: APIs under separate IAM. Never present row scoping as sufficient to keep a principal away from data — see Grants do not bound the CloudWatch Logs plane under Security considerations.

A DataSet scope may only name the actions GetRecords, GetMetricData, and ListMetrics. Any other action alongside a DataSet scope is rejected.

The structure is a list of groups. A record matches if it satisfies any group, and within a group every condition must hold. Each condition names a field, the operator IN, and up to 50 values. Up to 5 groups, each with up to 5 conditions — the service rejects anything larger with exceeds the maximum of 50 / exceeds the maximum of 5, whatever higher ceilings the API schema advertises.

Anything not named in values is excluded, and excluded data is indistinguishable from absent data — the principal sees a smaller result set with nothing to indicate that a filter removed anything.

Constraints:

• signalTypes and rowScopeGroups MUST both be present or both absent, and MUST appear only on a DataSet scope. Either half alone is rejected, and so is their use on any other resource type.
• IN is the only operator. There is no negation, no comparison, and no pattern match, so a row scope can only ever be an allowlist of values.
• You MUST NOT use row scoping to restrict metrics. The METRICS signal type is not supported and the grant is rejected.
• You MUST explain the any-group / all-conditions structure back to the customer before creating it. A customer who reads the groups as “and” will believe the grant is broader than it is, and one who reads the conditions as “or” will believe it is narrower.
• You SHOULD skip scoping entirely unless the customer has a specific reason. A narrowed grant that excludes the wrong thing is indistinguishable from a broken one.
• When a scopedActions call is rejected, you MUST explain the model the service applies, not just the one bad value: resources is a plain list of scopes, each with a required resourceType and up to five optional resourceArns; resourceType is a plain string validated per (action, resourceType) pair, case-sensitively; Space and Domain are never accepted inside a scope (omit resources to cover the whole Space); and every action must be within the named permission’s tier.

Step 5 — Verify, then report

1. Read the grant back rather than trusting the create response:

   Copy

   aws cloudwatchomni get-access-grant --grant-id <grant-id>

   Confirm the principal, the permission, the Space, and — if the grant was narrowed — the scoped actions and resources are what the customer asked for.

2. Report the outcome in the customer’s own terms as well as the API value: “test-user can now read telemetry in this Space (READ)”. The plain-language half is what the customer can check; the API value alone is not.

3. Say when it takes effect: a new grant, like a revoke, applies at the next authorization-cache turnover (minutes), not the instant it is created, so a session the person already has open may need to be refreshed or re-established before they see the Space.

Changing a grant

There is no update operation. Changing a grant’s permission level, its scope, or its name is a revoke plus a create — but the order depends on whether the new grant is at the same named level as the old one, and getting it wrong either fails or opens a gap:

Changing to a different level (an up- or downgrade, e.g. READ → READ_WRITE_DELETE, or SPACE_ADMIN → READ) — create first, then delete. Named-permission grants are one per principal per Space per level, so a grant at a different level coexists with the old one. Create the new grant first, verify it, then revoke the old one. The principal is never without access. Two separate protections matter when the grant being removed is an administrative one:

• Create-first avoids a gap and a lockout for the caller’s own downgrade: the new same-principal grant (say, READ) exists before the SPACE_ADMIN grant goes, so the caller is never without access.
• Last-admin protection is independent of that. The service refuses to delete the Space’s last ADMIN/SPACE_ADMIN grant regardless of what other grants the caller holds — a READ grant does not count. If the caller is the only administrator, another principal must be granted SPACE_ADMIN before the delete will succeed.

1. Read the existing grant with get-access-grant and confirm with the customer which grant is changing and what it conveys.
2. Create the new-level grant with create-access-grant (Step 3), then verify it with get-access-grant.
3. Revoke the old grant with delete-access-grant.

Changing the scope or name at the same level — revoke first, then create. A second grant at the same named level conflicts, so the old one has to go before the replacement can be made. This is the only case with an unavoidable brief gap; do it when the gap is acceptable.

1. Read the existing grant with get-access-grant and confirm.
2. Revoke it with delete-access-grant — see Revoking a grant.
3. Create the replacement with create-access-grant (Step 3), then verify it.

Either way the replacement has a new grant ID, so anything that recorded the old ID (a runbook, an audit note) needs updating — say so.

Downgrading or removing your own or the last admin grant

Removing the caller’s own SPACE_ADMIN, or the Space’s only administrative grant, needs care because the service refuses to revoke the last ADMIN/SPACE_ADMIN grant (ConflictException), and even where it would succeed, doing the delete first can lock the caller out of grant management:

Procedure for a self-downgrade (for example SPACE_ADMIN → READ, a different level):

1. Warn first: revoking your own SPACE_ADMIN removes your ability to manage grants in this Space; only another SPACE_ADMIN (or Domain administration) can restore it.
2. Confirm the grant you are about to remove by listing it, not from memory: list-access-grants --space-id <space-id> --principal-id <caller> --principal-type <type>. Note the SPACE_ADMIN grant’s ID and confirm what it conveys.
3. If you are the only SPACE_ADMIN, the delete will be refused — the Space must keep one administrator — so first grant another principal SPACE_ADMIN and verify it with get-access-grant.
4. Create the READ grant before deleting anything: create-access-grant … --name … --permission READ, then get-access-grant to verify it exists. Different levels coexist, so there is no gap.
5. Only now delete-access-grant --grant-id <space-admin-grant-id>. It does not affect the new READ grant or any other grant you hold.
6. The change takes effect on the timeline described under Revoking a grant; refresh the session before expecting the console to show READ only.

Constraints:

• For a same-level change you MUST NOT create the second grant first — it conflicts; revoke first. For a level change you SHOULD create the new-level grant first and delete the old one after, so the principal is never without access and an admin downgrade does not lock anyone out.
• You MUST NOT revoke a customer’s grant to “change” it without first confirming the replacement’s exact permission and scope; a revoke with no agreed replacement is simply a revocation.
• You MUST describe the change as what it is — a new grant created plus the old one revoked — never as an in-place update, edit, or “downgrade” of the existing grant, because no such operation exists. Say that the replacement carries a new grant ID, so anything that recorded the old one (a runbook, an audit note) needs updating.

Revoking a grant

Revoking a grant removes the access it conveyed for future requests. It does not delete anything in the Space, and it does not affect the principal’s other grants. It is not instantaneous, and it does not invalidate credentials. The one model to state, every time: Each Omni call is re-authorized against the Space’s grants, so a revoke takes effect at the next authorization-cache turnover (minutes) — even though the session credentials already issued to the principal remain valid until they expire (up to an hour) or the session is re-established. So a revoke is not a way to cut off access already in flight this second, and a direct IAM caller in the owning account keeps whatever its own IAM policy allows regardless of grants.

Revocation is the only way a grant ends — grants carry no expiry, so “remove their access” always means a delete-access-grant.

Revoking everything a person holds is a destructive action aimed at someone named only by alias, so the order matters:

1. Resolve the exact principal first. search-principals --domain-id <domain-id> --search-query <name>. Use the returned UUID as principalId with principalType IDC_USER (or IDC_GROUP if it resolves to a group).
2. Stop if the name does not resolve to exactly one principal. Zero results, or more than one plausible match, means you report that and stop. You MUST NOT fall back to a principal that happens to appear in the Space’s grant list, and you MUST NOT guess a UUID — revoking the wrong person’s access is the failure this step prevents.
3. List what they hold: list-access-grants --space-id <space-id> --principal-id <uuid> --principal-type IDC_USER, following nextToken. An empty list is a valid finding: nothing to revoke.
4. Present the list and confirm before deleting. Show every grant by its grantId, its permission in plain words alongside the API value, and the resolved name and UUID it belongs to. Get the customer’s confirmation on that list before running any delete — the confirmation is the only thing standing between a typo and a wrong-target revoke.
5. Revoke one grant at a time: delete-access-grant --grant-id <grant-id> for each confirmed grant. If one of them is the Space’s last ADMIN/SPACE_ADMIN grant, the service refuses it (ConflictException) until another principal holds SPACE_ADMIN.
6. Verify and report: run the step-3 list again and confirm it is empty — or, if a last ADMIN/SPACE_ADMIN grant was intentionally retained because the delete was refused, confirm that only that grant remains and say why. Note the revocation timeline above, and that access through an IDC_GROUP grant is unaffected because those grants belong to the group, not the person.

“I revoked the grant and they still have access.” Two things are true at once, and the answer names both:

1. Revocation takes effect at the next authorization-cache turnover, not at the instant of the delete (the timeline above). Five minutes of continued access is inside that window. This is expected behaviour, not a bug — do not delete the grant again and do not call the revoke broken.

2. The person may hold another grant. The revoke removed one grant and nothing else. Check what they still hold, directly and through their groups:

   Copy

   aws cloudwatchomni list-access-grants --space-id <space-id> --principal-id <principal-id> --principal-type <principal-type>
   aws cloudwatchomni list-access-grants --space-id <space-id> --principal-type IDC_GROUP

   An IDC_GROUP grant reaches every member and shows only the group ID, so a person with no direct grant can still reach the Space through a group.

Constraints:

• You MUST confirm which grant is being revoked before revoking it, by grant ID and by what it conveys. A principal often holds several, and “remove their access” rarely means all of them — list them first (What a given principal holds).
• You MUST NOT revoke the caller’s own SPACE_ADMIN grant without warning them that they may lose the ability to manage grants in that Space.
• The service refuses to revoke the last ADMIN/SPACE_ADMIN grant on a Space (ConflictException: at least one permanent ADMIN or SPACE_ADMIN grant must remain) and refuses to delete a service-managed grant (grantType SERVICE_MANAGED). Grant another principal SPACE_ADMIN first if the customer’s goal is to hand the Space over.

Looking up access

list-access-grants is the entry point for every “who has access”, “what grants does X hold”, and “at what permission level” question. get-access-grant answers the “detail of one grant” questions — how it is scoped, and who created it when. Start from the list, and reach for the detail only once you have a grant ID and the customer wants its scope or timestamps. The summaries alone answer “who has access / what does X have”; do not fetch the detail of every listed grant by default. A deliberate full-scope audit (“how is each of these scoped?”) is the one case that warrants details for the whole list.

Both are read-only. Neither pivots to IAM — the grants are the authorization records.

Who has access to the Space

Returns grant summaries (items) — for each grant the grantId, grantArn, name, principal (principalType and principalId), permission, grantType, spaceId, and domainId. A summary carries no scope and no timestamps; those are on the detail only.

The result is paginated. If the response carries a nextToken, call again with --next-token <token> and keep going until it is absent. Do not report a partial page as the full set — say how many grants you have read if you stop early.

An empty list is a success: the Space has no grant matching the filters. That is a finding, and it is different from a failed call (see Reporting a lookup).

An access audit has three questions, and a complete answer covers all six points:

• Who can grant access to others? list-access-grants --space-id <space-id> --permission SPACE_ADMIN. SPACE_ADMIN is delegation of grant management, so this is the first question of any audit.
• Who can write or delete? --permission READ_WRITE_DELETE, plus the SPACE_ADMIN holders, because that level includes write and delete.
• Does anyone reach the Space without a grant? For people, console sessions, and Access Profiles, no — access is deny-by-default and the grants are the complete answer. A direct IAM caller in the owning account with no matching grant falls through to its own IAM policy, so a full audit also reviews IAM policies for direct IAM callers.
• Both filtered lists are paginated: follow nextToken to the end, and say what each filter narrowed over (all grants on the Space) and how many grants came back.
• Grants never expire, so unused ones accumulate until someone revokes them.
• Attribution comes from the detail, not the summary: get-access-grant returns createdBy and createdAt — name those two fields when saying who added a grant and when.

Resolving Identity Center UUIDs to names

IDC_USER and IDC_GROUP grants carry only the principal’s Identity Center UUID (occasionally a userName such as an email) — never a display name. There is no by-ID lookup: search-principals has no principal-ID parameter, and identitystore is not the answer. Resolve names the way the console’s Manage permissions page does: pull the Domain’s whole roster once and join it to the grants yourself.

1. List the grants (following nextToken). Each summary carries principal.principalId and the Space’s domainId:

   Copy

   aws cloudwatchomni list-access-grants --space-id <space-id> --principal-type IDC_USER
   aws cloudwatchomni list-access-grants --space-id <space-id> --principal-type IDC_GROUP

2. Pull the roster with the match-all query *, paging until nextToken is absent (the console caps this at 20 pages of 50 — 1,000 principals):

   Copy

   aws cloudwatchomni search-principals --domain-id <domain-id> --search-query '*' --max-results 50
   aws cloudwatchomni search-principals --domain-id <domain-id> --search-query '*' --max-results 50 --next-token <token>

   Each result carries principalId (the UUID), principalType (USER or GROUP), displayName, and — for users — userName. --max-results and --next-token are accepted only with *; a name search returns at most 10 results and does not paginate.

3. Join client-side: a grant’s principalId matches a roster entry’s principalId or its userName (grants are sometimes keyed by the userName). Map USER → IDC_USER and GROUP → IDC_GROUP.

4. Report each principal as displayName (plus userName for a user) alongside the UUID, with its permission in plain words and the API value. For a group, say the grant reaches every member and the grant shows only the group’s ID.

5. Anything the roster does not resolve is reported as its raw UUID with a plain statement that it could not be resolved — the principal may have been removed from the directory, the roster may have been cut off at the page cap, or the Domain may not use Identity Center (in which case search-principals itself returns a 400 saying the Domain must be provisioned with Identity Center). Never guess a name.

Narrow the list server-side with the filters the operation supports, whenever the ask names an exact value:

What a given principal holds

Pass both --principal-type and --principal-id; the same identifier can be valid for more than one type. A principal may legitimately hold several grants on one Space — one per named level, plus up to ten CUSTOM grants — so report all of them, with the permission of each, not just the first.

If the customer wants to know what those grants allow (their scope), follow up with get-access-grant on the grant(s) they ask about.

Finding a person by name or alias

A grant identifies a person by principalId, and what that is depends on the principal type — so how you find “jdoe” depends on how jdoe signs in.

An IAM user or role. The ARN embeds the name, so list the Space’s IAM grants and match on the ARN client-side:

Repeat with --principal-type IAM_USER if the person may be an IAM user. JMESPath contains is case-sensitive, so match the alias as it appears in the ARN. Follow nextToken across pages before concluding there is no match.

An Identity Center user or group. The grant holds only the Identity Center UUID — the summary carries no display name, so no substring of an alias will ever match a grant directly. Resolve the name to an ID first, in the Space’s Domain (for the reverse direction — many UUIDs to names — use the roster join in Resolving Identity Center UUIDs to names):

Then filter the grants by that ID:

Also check the person’s groups: a grant to an IDC_GROUP reaches every member, and the grant shows the group ID, not its members. If the direct lookup is empty, say that the person holds no grant of their own and may still reach the Space through a group grant — list the Space’s IDC_GROUP grants and, if the customer knows the person’s groups, match on those IDs.

When the name cannot be resolved — the Domain does not use Identity Center, search-principals returns nothing, or the caller cannot run it:

• Say so plainly. Identity Center principals appear as UUIDs, and this lookup could not resolve the alias to one.
• Offer to list the Space’s Identity Center grants (--principal-type IDC_USER or IDC_GROUP) so the customer can identify the principal from the UUIDs and names.
• Never pretend an alias was resolved, and never invent or guess a UUID.

How a specific grant is scoped

Use this after the list when the customer wants a single grant’s scope — the actions and resources it allows — or who created it and when. Neither is on the summary. Read the detail and report from it:

• A grant’s scope is scopedActions — each entry a set of actions over resources, with optional contextConditions — and it comes back exactly as the service holds it. That is the only scope field on a grant.
• A grant with no scope fields grants its permission level unscoped across the whole Space — say that, rather than reading the absence as an error or as “no access”.
• For a workload principal (ALERT, AGENT), the detail is how you learn which Access Profile it may assume: a trust grant is a CUSTOM grant whose single action is AssumeAccessProfile, and the profile’s ARN sits in that entry’s resources. The summary shows only that a CUSTOM grant exists.
• The detail carries createdBy, createdAt, and updatedAt — use these to answer “who granted this / when”. There is still no expiry.
• A grant ID that matches no grant comes back from get-access-grant as ResourceNotFoundException. That is a successful lookup whose finding is “nothing found” — the ID is wrong, or the grant was revoked — not a failure of the lookup. Say so and offer to list the Space’s grants; do not retry with a guessed ID. (delete-access-grant is idempotent: deleting a grant that no longer exists succeeds.)

Reporting a lookup

• Answer in the customer’s terms, from the operation’s own output — “jdoe has READ_WRITE_DELETE on this Space, granted as an IAM role: they can read, write, and delete within it” — not a dump of the raw JSON. Give every permission value its plain-language meaning, not just the unusual one:

  API value	Say
  READ	can read the Space’s telemetry
  READ_WRITE_DELETE	can read, write, and delete within the Space
  SPACE_ADMIN	can manage grants — give or revoke others’ access — and also read, write, and delete
  CUSTOM	can perform exactly the listed actions and nothing else (read the detail for the list)

• Only report what you read. A summary gives you the principal and permission; you do not know a grant’s scope until get-access-grant has returned it. Do not describe a scope you have not fetched, and do not describe an expiry that does not exist.

• Say what you narrowed over. If you filtered server-side, or matched client-side, or stopped before the last page, tell the customer how many grants you looked at — a filtered answer presented as complete is worse than an honest partial one. For an audit, name the population each filter ran over (all grants on the Space) and how many came back at each level.

• Attribute grants from the detail. When the question is who added a grant, or when an audit turns up grants nobody recognizes, get-access-grant carries createdBy and createdAt; summaries do not. Grants never expire, so stale ones stay until revoked.

• A lookup that fails, fails. AccessDeniedException, expired credentials, a missing --space-id, or a service error is a broken call, never a conclusion about the Space. Never report “nobody has access” or “X has no access” from a failed call; say the lookup failed, quote the error, and — when it points at expired credentials or a missing permission — relay that rather than retrying the same call.

• AccessDeniedException on a grant lookup means the caller holds no grant on this Space. list-access-grants and get-access-grant are read actions: every named permission (READ, READ_WRITE_DELETE, SPACE_ADMIN) includes them, and a CUSTOM grant includes them only if it names ListAccessGrants/GetAccessGrant. So the denial says the caller has no grant here at all (or a CUSTOM one that omits those actions). The answer is “someone who holds SPACE_ADMIN here either runs the lookup or grants you access — READ is enough to look up grants” — not an IAM policy edit, not sts get-caller-identity, and not “retry once you have permission”. Contrast it explicitly with an empty list, which is a successful answer meaning no grant matched.

Troubleshooting

Not every authorization error is a missing grant. If create-space succeeded and the Space then fails in use, suspect the space access role’s trust policy, not a grant — and answer from spaces-and-domains.md (“If the Space was created but cannot be used”), whose checklist is the complete answer. In short: create-space validates only that dataAccessRoleArn is non-blank and in the caller’s account — it never attempts to assume the role, so a wrong trust policy is accepted at create time and fails in use, and a successful create is not proof the role is right. Check, on that role: the service principal is cloudwatch.amazonaws.com; all three of sts:AssumeRole, sts:TagSession, and sts:SetContext are allowed (assume-role alone is accepted at create and fails in use); and both confused-deputy conditions are present — aws:SourceAccount equal to the account, and aws:SourceArn matched against the cloudwatch space/ resource with Region and space ID wildcarded, since the space ID did not exist when the role was created. Then read the Space back with get-space to verify. Do not recreate the Space as the first remedy — the same role fails the same way — and do not create a grant to fix a trust-policy problem. A missing grant instead means nobody could reach the Space from the start, for everyone.

Rule: When a call returns an error, surface the error code and message verbatim, then map to the mitigation below. Do NOT invent error text, do NOT paraphrase what the service returned, and do NOT synthesize a mitigation for an error that is not listed.

If the error does not match a row above, quote it verbatim, say it is unmapped, and ask the customer how to proceed.

Two error classes have more to them than one row can hold. When one of them is the question, the complete answer covers every point below:

A rejected scopedActions scope (ValidationException on resourceType):

• Space and Domain are both invalid as a scope’s resourceType — name both. The grant is already bound to one Space or Domain; to cover the whole Space, omit resources.
• resourceType is a plain string matched case-sensitively: DataSet, not Dataset. An unknown spelling is rejected with Unknown resourceType. Allowed values: …, which lists the accepted spellings.
• Validation is per (action, resourceType) pair: a known type next to an action that does not operate on it is rejected with Resource type X is not valid for action Y, which lists nothing.
• resources is a plain list of scopes — not an object with a scopes member — and each scope has a required resourceType and up to five optional resourceArns.
• Every action in an entry must be within the named permission’s tier.

ResourceNotFoundException from get-access-grant:

• It is a successful lookup whose finding is “no such grant”: the ID is wrong or the grant was already revoked. It is not an authorization problem and not a failed call.
• Find the right ID with list-access-grants, optionally filtered by principal; never retry with a guessed ID.
• delete-access-grant is idempotent — deleting a grant that no longer exists succeeds rather than raising this error, so a revoke never needs a pre-check for existence.

Gotchas — expectations that look like faults

• A principal with a grant still cannot see anything. Check the Space, not the grant. A grant names one Space, and a principal working in a different Region is working against a different Space. Nothing is wrong with the existing grant. Confirm which Space the person is in, check that Space’s grants for them first (list-access-grants --space-id <other-space-id> --principal-id … --principal-type …), then create a separate READ grant on that Space. Do not broaden the original grant to avoid a second one, and do not reach for an IAM policy change — grants do not span Regions or Spaces at any permission level.
• A second grant at the same level is rejected. Named-permission grants are one per principal per Space per level. Revoke and re-create to change a level.
• There is no write-only permission. READ_WRITE_DELETE is the only mutating named level. A customer expecting to grant writes without deletes needs CUSTOM.
• There is no update operation. Renaming, re-leveling, or re-scoping a grant is a revoke plus a create, and the grant ID changes.
• A narrowed grant appears to be missing data. Row scoping is an allowlist with IN only. Anything not named in values is excluded, and excluded data is indistinguishable from absent data.
• A revoked grant seems to still work. Each Omni call is re-authorized against the Space’s grants, so a revoke takes effect at the next authorization-cache turnover (minutes) — even though the session credentials already issued to the principal remain valid until they expire (up to an hour) or the session is re-established. If access persists beyond that, the principal holds another grant (directly or through an IDC_GROUP), or is a direct IAM caller whose own IAM policy allows the call.
• A grant “should have expired” by now. Grants never expire. If access was meant to be temporary, someone has to revoke it; nothing does so automatically.
• The list shows no scope for a CUSTOM grant. Summaries never carry scope. Read the grant with get-access-grant.
• A person’s alias matches nothing. Identity Center principals are UUIDs in the grant; resolve the name with search-principals first, and remember access may arrive through an IDC_GROUP grant that shows only the group ID.
• An empty list looks like an error. It is not. list-access-grants returning no grants is a successful answer: nothing matched. Only an actual error (access denied, not found, throttled) is a failed lookup — and a failed lookup says nothing about who has access.
• The caller has valid credentials and is still denied. Credentials say who the caller is; grants say what they may do in the Space. Check for a grant before suspecting IAM — see aws-observability → references/cloudwatch-omni/programmatic-access.md.

Security considerations

• Grant to Identity Center groups rather than individual users where possible. Offboarding a person then requires no change to grants.
• Prefer the narrowest named permission over CUSTOM with a long action list. A named level is auditable at a glance; a list of 50 actions is not.
• SPACE_ADMIN conveys the ability to grant access to others. Treat it as delegation of administration, not as a higher read level. Reviewing who holds it (--permission SPACE_ADMIN) is the first question of any access audit — and because it includes write and delete, “who can write or delete” is the READ_WRITE_DELETE holders plus the SPACE_ADMIN holders.
• IAM_ROOT has no session identity to attribute actions to. Avoid it unless the customer explicitly asks.
• Grants do not bound the CloudWatch Logs plane. Row scoping is an access control, not a redaction, and it binds one plane only. It restricts which records a principal can retrieve through the Space’s DataSet; it does not remove sensitive fields from records they can retrieve, and it does not touch the CloudWatch log groups the Dataset was populated from — those stay readable via logs:GetLogEvents, logs:FilterLogEvents and logs:StartQuery under IAM no grant constrains. Account-wide log-group:* forwarding is the onboarding default and every record carries @logGroupName, so a row-scoped grant is not sufficient on its own: restrict the principal’s logs:* on the source log groups too. And because a row scope is an IN allowlist, excluded data is indistinguishable from absent data to the principal — they cannot tell a service with no records from one the scope hides — so it is not a way to signal that something is being withheld either.
• Review grants periodically with list-access-grants. Grants never expire and the per-Space limit of 500 is high enough that unused grants accumulate unnoticed. Use createdBy / createdAt on the detail to find who added a grant nobody remembers.
• Grants are per Space by design. Resist the temptation to give a principal a broader permission in one Space to avoid creating grants in others.

Additional resources

• spaces-and-domains.md — creating the Domain and Space a grant applies to
• org-domains.md — Domain access grants, for administration across an organization, and the cross-account credential path
• access-profiles.md — bounding what an agent, alert, or integration can do, and the permission and trust grants that make a profile work
• aws-observability → references/cloudwatch-omni/concepts.md — how Domains, Spaces, grants, and profiles fit together, and the setup order
• aws-observability → references/cloudwatch-omni/programmatic-access.md — calling Omni from the CLI, SDKs, or code, and why an IAM caller still needs a grant