Agentic Standards

Agentic Standards are rules and rulesets to enforce enterprise policies and guardrails on campaigns and Canvases in Braze. Operator can help you draft rules as you configure a standard. Before a campaign or Canvas launches, you can run an agentic evaluation against your standards to validate brand guidelines, organizational conventions, and technical requirements. For an introduction to Braze Agents, see Braze Agents.

Agentic Standards reduce manual oversight so every message Braze sends is accurate, compliant, and launch-ready.

Plan tiers

Agentic Standards availability depends on your plan tier: free or paid.

Feature	Free tier	Paid tier
Build Campaign Standards and Canvas Standards	Available	Available
Evaluation runs per month per workspace	25	Unlimited
Monthly usage cap	Not applicable (run-count limit only)	High

On the Free tier, Evaluation preview (Simulate response) and launch-time Run evaluation or Re-run evaluation count toward the 25 runs. Each standard counts as a separate run, so if you select multiple standards and run them at the same time, each one counts toward the limit. For example, running three standards against a campaign uses three runs. Contact your Braze account representative to upgrade.

On the Paid tier, there is no monthly run-count limit. More intensive evaluations contribute more toward the monthly usage limit. To see how much of the limit each evaluation used, check the standard’s evaluation logs. Most workspaces don’t reach this cap. If you do, contact your customer success manager.

How it works

Agentic Standards are available in two types: Campaign Standards and Canvas Standards. When you create a standard, you define specific rules grouped into “rulesets” that are evaluated before you launch.

You can choose from pre-built rulesets covering common marketing needs or create custom rules specific to your team. When configured, you can test a standard in the Evaluation preview pane against any existing campaign or Canvas in your workspace. The agentic evaluation provides a detailed report with categories of its findings.

Content Blocks and Liquid

Before evaluation, Braze expands Content Blocks into their content. Each block’s text is labeled with the block’s name, and Agentic Standards also gets a list of the blocks each field uses, including nested blocks and any block that failed to expand. This means rules can reference a block by name (for example, “the footer uses Legal_Disclaimer”).

Liquid isn’t rendered against a specific user. Agentic Standards evaluates the Liquid as written, and it checks for syntax errors, unrecognized variables, and conditionals that have no {% else %} fallback. Connected Content calls aren’t made during evaluation.

Create an Agentic Standard

Campaign Standards enforce rules on one-time and recurring campaigns. Canvas Standards enforce rules on multi-step customer journeys. Both types use the same Details and preview workflow, but Canvas Standards add a second rules step for per-step checks.

• campaign
• canvas

Use Agentic Standards

After you configure a Campaign Standard or Canvas Standard, you can use it to evaluate a campaign or Canvas during the final review process. This confirms your messaging meets all requirements before it is sent to your users or users enter the journey.

Run an evaluation

• campaign
• canvas

Review evaluation results

After the evaluation is complete, a summary shows the findings in these categories: Pass, Fail, Warning, and Ignored.

If an evaluation fails due to a rate limiting error, select Try again.

The Fail tab lists rules that were not met. For each failure, the standard provides the:

• Rule: The specific criteria being checked, such as “Spelling & Grammar Check”.
• Reason: An explanation of why the check failed. For example, the evaluation might identify that “personalized” was used instead of the Australian English spelling “personalised”, or that a delay step is shorter than your team’s minimum wait time.

The Pass tab lists all rules that were successfully followed. This confirms that checks like Offensive Language Detection or Naming Convention Validation have been cleared.

The Warning tab lists non-critical issues that may require attention. For example, if you’re testing an email ruleset against a webhook campaign or a Canvas with only in-app message steps, the evaluation may yield a warning that sender names do not apply.

The Ignored tab lists issues you chose to skip for the current evaluation run.

The Last run indicator shows when a standard was last evaluated against your campaign or Canvas. It stays visible for 90 days after the most recent run, even if the campaign or Canvas is stopped or archived.

Results based on the agent’s judgment

A result in any tab can show Based on the agent’s judgment under its reason. The marker applies only to that rule, not the entire standard. It means the result reflects the agent’s own assessment rather than a directly verifiable fact, such as a setting’s value, a tag, or whether a link is present. For example, deciding whether copy is well-written, correctly spelled, clear, or on-brand is a judgment, even though the agent can read the copy.

When the agent assumes something that it didn’t find stated in your campaign or Canvas, the marker is followed by a short note that explains what it assumed and why. For example: Based on the agent’s judgment: The campaign’s name and copy suggest it is a Home campaign.

If you want a rule to be checked against concrete facts instead, revise it to reference specific settings, tags, or content. For example, use “The email footer includes an unsubscribe link” instead of “The email makes it easy to opt out.”

Resolve or ignore issues

For every identified failure or warning, you can decide how to proceed before launching. Select Resolve next to an issue, then select from the following options:

• Mark as fixed: Select this after you have updated your campaign or Canvas configuration, journey structure, or copy based on the evaluation suggestion.
• Ignore this issue: Select this to skip the issue for this run only. This is useful for intentional deviations or edge cases where the evaluation suggestion may not apply.
• Ask BrazeAI Operator: Select this to fix the issue using Operator.

After all critical issues are resolved or ignored, you can proceed to launch your campaign or Canvas.

Review evaluation logs

To review past evaluations of a standard, go to Agent Console > Agentic Standards, select the standard, then select the Evaluation logs tab. Each row is one evaluation run. Use the Details column to expand a run and see its individual rule evaluations, or to open the details for an evaluation. To export the logs, select Export CSV.

See who ran an evaluation

The Run by column shows the name of the Braze user who ran each evaluation, or their email address if no name is on file. Rule evaluations inside a run were run by the same user, so their rows leave this column blank. Evaluations that weren’t run by a user show N/A. The CSV export also includes a Run by column.

To show only evaluations run by a specific user, select Filters, then choose the user under Run by. Users who ran the standard in the selected date range are listed first. Other users in your company are listed but can’t be selected.

Usage per evaluation

On the Free tier, evaluation logs don’t show usage because your limit is the number of evaluation runs per month rather than usage.

On the Paid tier, the Limit used column shows the percentage of your workspace’s monthly usage limit that each run and rule evaluation used, such as “0.2%”. A share too small to show at two decimal places appears as “< 0.01%”. Use this column to spot evaluations that use more of your limit than you expect, such as runs of standards with many rulesets.

• Cached means the result was reused from an earlier, identical evaluation and didn’t count toward your limit.
• The same percentage appears as Limit used in the details panel and as Limit used (%) in the CSV export.

For details, see Plan tiers.

Limitations

Agentic Standards evaluates the configuration and content of your campaign or Canvas as they’re saved in Braze. It doesn’t send messages, render them for individual users, or visit external pages.

Keep the following limitations in mind when you write rules, and keep your existing manual checks for these areas.

• No test sends or per-user previews: Agentic Standards can’t send test messages, preview a message as a specific user, or render previews for many users at once. Rules that depend on what a particular user receives aren’t supported. Use test sends and previews for those checks.
• Liquid and Connected Content aren’t rendered: Agentic Standards evaluates Liquid as it’s written, not as it renders for a user, and it doesn’t make Connected Content calls. For details, see Content Blocks and Liquid.
• Links aren’t followed: Agentic Standards can check that links are present and correctly formatted, such as URLs, deep links, and UTM parameters. It doesn’t open them. It can’t detect broken links, such as links that return a 404 error, or confirm that a link’s destination matches what the message says.
• No user or segment lookups: Agentic Standards can’t look up individual user profiles or check whether a specific user is in a segment. It can only evaluate the targeting settings on the campaign or Canvas itself.
• Canvas step coverage: Step rulesets run only on Message, Content Optimizer, and Audience Paths steps. Other step types can’t be evaluated.
• Very large messages: If a campaign or Canvas has more content than Agentic Standards can evaluate at once, such as a multivariate email with many large variants, the content is shortened before evaluation. The results then show a warning, and rules may have been checked against incomplete content.

Best practices

• Start with templates: Use the pre-built rulesets for setup and links and tracking first, as these cover the most common manual errors and give the agentic evaluation the right context.
• Be specific: When writing custom rules, provide clear examples of what “correct” looks like. For example, instead of writing “Check the naming convention,” try “The campaign or Canvas name starts with the current year (for example, 2026_).”
• Test with known outcomes: In Evaluation preview, test against a campaign or Canvas that should pass all rules and one that should fail specific rules. This confirms your rules are evaluated accurately before you rely on them at launch.
• Iterate often: As your brand guidelines or internal processes change, update your Campaign Standard and Canvas Standard rulesets to keep your automated checks relevant.