Agent Postmark

agent-postmark

Use agent-postmark for Postmark delivery incidents, bounce/suppression questions, message status, sender/domain configuration, message streams, and webhooks.

Safety

• Never ask the tool to reveal account or server tokens.
• Never accept pasted Postmark tokens in chat. Ask the user to run agent-postmark profiles add <profile> --form --account-token and/or agent-postmark profiles servers add <profile> <server> --form --server-token --server-id <id> locally so tokens go directly into OS dialogs.
• For initial setup with multiple server tokens, use agent-postmark profiles setup <profile> --form --account-token --server app:<id>:outbound --server billing:<id>:outbound.
• Use agent-postmark profiles update <profile> --form --account-token or agent-postmark profiles servers update <profile> <server> --form --server-token when a stored token needs replacement.
• Use agent-postmark profiles servers remove <profile> <server> to remove a server context and its stored token.
• Prefer read-only commands.
• Remember token scope: account-token commands handle servers, domains, and signatures; server-token commands handle message streams, messages, bounces, stats, suppressions, and webhooks.
• Use agent-postmark suppressions list, not a raw suppression dump, when browsing suppressions. Postmark's dump endpoint can return very large full exports and is intentionally not exposed as an agent-facing command.
• Subjects and addressing fields are visible for triage. Use list/search output first; ask for messages content only when the user needs actual body, header, or attachment details.
• Do not add --yes to mutation commands unless the user explicitly asks for that state change.

Start Here

agent-postmark usage
agent-postmark profiles list
agent-postmark profiles check
agent-postmark config show

For incident-style questions, prefer investigations before low-level resource commands:

agent-postmark investigate delivery --email user@example.com
agent-postmark investigate bounce <bounce-id>
agent-postmark investigate domain-health example.com
agent-postmark investigate stream-health --stream outbound
agent-postmark investigate webhook-health

For local testing, run mockpostmark and set AGENT_POSTMARK_BASE_URL.

Output

Lists and investigations default to NDJSON. Gets default to NDJSON (one line per id). Pass --format json for a single pretty object, or --format yaml for YAML. Errors are JSON on stderr with error, fixable_by, and usually hint.

Get (single + multi). get <id>... takes one or more ids and returns one result per id, in input order. Default output is NDJSON: one line per id — the record, or {"@unresolved":{"id","reason","fixable_by","hint"?}} for an id that couldn't be resolved (e.g. not found / bad id). --format json|yaml collapses to one {"data":[…], "@unresolved":[…]} envelope. A single get <id> is just the one-element case (NDJSON one line by default; was pretty JSON before — pass --format json for the object). Item-level misses stay on stdout and exit 0; only a command-level failure (auth, network) goes to stderr with exit 1 and empty stdout.

Subjects and addressing fields are visible for delivery triage. List output omits bulky bodies, headers, and attachments by default; use agent-postmark messages content <message-id> [message-id...] when the user needs the actual email content. Tokens, secrets, URL credentials, and original raw email blobs are redacted with "[REDACTED]" and top-level @redacted paths when possible.

Investigations emit evidence records:

{"type":"entity","object":"bounce","id":9001,"data":{}}
{"type":"finding","severity":"critical","summary":"...","data":{}}
{"type":"next_command","command":"agent-postmark suppressions check <email>","reason":"..."}

Non-secret profile/config metadata lives in XDG config. Tokens live in Keychain. profiles list and profiles check show token presence but never token values. Use --server <alias> to select a stored server context; use --server-id <id> only when a command needs a numeric Postmark server ID override.

Incremental References

Load only the reference needed for the current task:

• references/scenarios.md: use when the user asks a support-style question and you need a command sequence.
• references/investigations.md: use when choosing or interpreting investigate commands.
• references/commands.md: use when you need exact command syntax or flags.
• references/output.md: use when parsing NDJSON, errors, redaction, or mutation guards.