SendRaven checks agent-email limits on every send; LaunchRepo says its pause-before-paying default is not software-enforced
Two products launched the same week take opposite approaches to a core AI-agent trust question: whether a stated limit on an agent's actions is checked by software before it executes, or left to the agent to follow.
SendRaven, an email-sending API and Model Context Protocol server built for AI agents, enforces a daily send limit, a recipient allowlist and an optional human-approval requirement on every request, according to the company's website, captured Sept. 27. "A daily send limit, a recipient allowlist, human approval. Checked on every send before anything goes out, through REST or MCP alike, and the key cannot change its own limits," the site states. An API key can be configured to draft email without permission to release it; held messages wait in an approval queue, and the key "cannot approve its own" sends, according to the company. SendRaven also publishes signed webhook delivery events in real time, which it describes on its site as a record of what went out.
LaunchRepo, a tool that has an AI agent autonomously create accounts and submit listings to online directories, takes a different approach to a similar kind of control. Its default workflow pauses the agent before paid actions, but the company's FAQ states plainly that the pause is not enforced in software: "These permissions are instructions for your agent, rather than a software-enforced approval gate," according to the site. The product otherwise runs with minimal human involvement: "Automated, end to end. Your agent creates the account, confirms the verification email, fills in the forms and submits the listing on its own," the site states, with a person needed only to handle logins and CAPTCHA challenges before later sessions resume the remaining work unattended.
The distinction carries the same weight in both cases: an agent acting with financial or account-creation authority carries different risk depending on whether a stated safeguard is checked by the software before an action executes, or relies on the agent choosing to follow an instruction it has been given.
SendRaven, a candidate entry in AI Dispatch's index, prices its service at $19 a month starting at 50,000 emails sent, billed by usage rather than by seat, with volume pricing published up to $379 a month at 1 million emails. LaunchRepo, already indexed, sells as a one-time purchase at €104.98 with no subscription.
Entries in this piece 1
Published index entries backed by the same source documents this piece cites.
Sources 10
“A daily send limit, a recipient allowlist, human approval. Checked on every send before anything goes out, through REST or MCP alike, and the key cannot change its own limits.”
sendraven.ai · checked Sep 27, 2026“Let the agent draft. Let a human decide. A key can create email without permission to release it. Held sends wait in an approval queue instead of going out, and the key cannot approve its own.”
sendraven.ai · checked Sep 27, 2026“Signed delivery events, in real time, when your application needs them.”
sendraven.ai · checked Sep 27, 2026“$19 a month, from 50,000 emails to start, then automatic volume pricing — $109 at 250K, $379 at 1M.”
sendraven.ai · checked Sep 27, 2026“Pay for emails sent. Never for contacts stored.”
sendraven.ai · checked Sep 27, 2026“the default workflow pauses before paid actions.”
launchrepo.dev · checked Sep 27, 2026“These permissions are instructions for your agent, rather than a software-enforced approval gate.”
launchrepo.dev · checked Sep 27, 2026“Automated, end to end. Your agent creates the account, confirms the verification email, fills in the forms and submits the listing on its own.”
launchrepo.dev · checked Sep 27, 2026“You handle logins and human challenges; later sessions resume the remaining work.”
launchrepo.dev · checked Sep 27, 2026“LaunchRepo one-time €104.98”
launchrepo.dev · checked Sep 27, 2026