# AI Agent Supported Actions And Support > Use the current supported-actions matrix, understand Agent Control terms, account for freshness and compliance boundaries, and troubleshoot without sharing secrets or Subscriber content. - Human page: https://mailrith.com/developers/agent-supported-actions - Markdown page: https://mailrith.com/developers/agent-supported-actions.md - Category: AI Agent Platform - Reading time: 10 min read - Last updated: 2026-07-22 - Related keywords: AI Agent Supported Actions And Support, AI Agent Supported Actions And Support developer docs, AI Agent Platform, AI Agent Platform developer docs, Mailrith developer docs, Mailrith public API, Supported Actions, Permission And Approval Terms, Freshness, Compliance, And Provider Boundaries, Support Playbooks, Verified Workflow Evidence, AI Agent Architecture And Security, AI Agent Client Compatibility, AI Agent Quickstarts ## AI Agent Notes - Use this page as implementation guidance, then validate exact endpoint fields against the OpenAPI document. - Keep API keys server-side and workspace-scoped unless a guide explicitly says otherwise. - Do not invent privacy, consent, or lawful-basis evidence. Send only fields that appear in the OpenAPI schema for the endpoint you are using. ## What this guide covers Check read, draft, test, execute, and unsupported states, then use exact permission, approval, provider, and support guidance. ## Supported Actions Mailrith's AI Agent workflows are in beta. The table below uses the same action words as Agent Control: Read does not change workspace data, Draft saves editable work without an external effect, Test uses controlled recipients or simulated sandbox behavior, Execute can affect real Subscribers or external systems, and Unsupported means Mailrith does not offer a safe agent operation for that task. > Execute support means the operation exists; it does not mean the connected credential may use it. Current scopes, workspace policy, rollout mode, beta access, sandbox proof, preview, approval, and provider checks still apply. **Current Support Matrix** ```text Resource Read Draft Test Execute Unsupported Or Restricted Workspace and capability Yes No No No Billing and ownership changes Subscribers Yes Yes No Yes Raw consent evidence and secret fields Tags and custom fields Yes Yes No Yes Unbounded bulk changes Segments Yes Yes Yes Yes Unbounded preview or search Broadcasts Yes Yes Yes Yes Live send without exact approval Sequences Yes Yes Test Yes Activation without exact approval Automations Yes Yes Test Yes Activation without exact approval Forms and landing pages Yes Yes Yes Yes Publication without exact approval Webhooks Yes Yes No Yes Secret display and unapproved administration Analytics and diagnostics Yes No No No Raw-event scans and unbounded reports Recommendations Yes Yes No No Direct execution of a recommendation Experiments Yes Yes No No Per-Subscriber assignment storage Provider credentials Status No No No Secret creation, display, or export by an agent ``` ## Permission And Approval Terms A Permission controls whether a connection may request an operation. Allowed Toolsets narrow which operation groups are available in the workspace. Approval Policy decides which changes require a person. A Preview shows the exact target, affected Subscribers, effects, warnings, and blocking issues. Approve Exact Action authorizes only that unchanged plan. Emergency Revoke stops a connection; it does not undo completed work. Beta Access is separate from a connection permission. High-impact production access requires a verified sandbox action and administrator enrollment. Daily Actions and Daily High-Impact Actions are rolling 24-hour beta limits. ## Freshness, Compliance, And Provider Boundaries Capability, resource, approval, activity, diagnostic, and progress responses describe Mailrith when the request is processed. A saved action plan can become stale when its target changes. Analytics use compact rollups and can lag live activity; always read the response range and completion state. Mailrith enforces its stored consent, suppression, unsubscribe, complaint, frequency, preflight, and provider-readiness rules. You remain responsible for lawful basis, permission to contact Subscribers, the accuracy of model-produced content, sender identity, delivery-provider terms, privacy notices, retention duties, and the laws that apply to your use. Mailrith does not expose provider credentials to agents. Provider acceptance is not proof of inbox delivery. When an outcome is Uncertain, check Broadcast progress and provider evidence before creating another action. ## Support Playbooks Start every support request with the workspace name, client name and version, approximate time, operation, and Mailrith request or action ID. Never send an API key, OAuth token, approval token, Subscriber export, request body, full email body, or provider credential. - Connection failure: check the exact MCP URL, OAuth return address, expiry, resource binding, and requested Permission. Reconnect or rotate only after identifying the cause. - Denied Permission: open Agent Control → Connections, review the connection's Permissions, and either remove the operation or have the owner grant the smallest task preset. - Stale Approval: open Agent Control → Approvals, deny the old action, refresh the target, and ask the same connection to create a new exact plan. - Uncertain Execution: stop retries, open Agent Control → Activity, search by action ID, inspect the target and provider progress, and escalate if the final effect cannot be reconciled. - Compromised Connection: open Agent Control → Connections, choose Revoke or Emergency Revoke, confirm new requests fail, then create a smaller replacement credential only after review. ## Verified Workflow Evidence The quickstarts and repository examples are test recipes, not customer case studies. Mailrith has no verified beta customer workflow to publish yet. A case study may be added only after the customer confirms the workflow, permissions, approval path, measured result, and wording. Until then, Mailrith must not present a hypothetical autonomous workflow as customer evidence. ## Related Guides - [AI Agent Architecture And Security](https://mailrith.com/developers/agent-security.md): Mailrith uses one public resource model, least-privilege credentials, typed operations, exact action approval, existing delivery safeguards, and compact redacted activity across every supported agent client. - [AI Agent Client Compatibility](https://mailrith.com/developers/agent-compatibility.md): Use the current compatibility matrix and read-only starter configurations, then record live conformance against an exact client build or review date before making a support claim. - [AI Agent Quickstarts](https://mailrith.com/developers/agent-quickstarts.md): Follow four task-based paths for a five-minute read, campaign drafting, an exact approval-based live send, and isolated sandbox testing without giving an agent broad workspace control.