Mail Pelican is the open-source Mailchimp alternative built for self-hosters and for AI agents as first-class operators. Safe sends, consent and compliance by default, and deliverability instruments nobody else in the category ships.
Three kinds of operators, one control plane.
The unglamorous parts of email — consent, suppression, idempotency, bounce handling — are the parts Mail Pelican refuses to compromise on.
Every campaign goes lint → preview → audience snapshot → single-use confirmation token → send. Nothing ships by accident, and API keys carry send limits and approval thresholds.
Double opt-in with single-use confirmation tokens, workspace-global suppression, per-dispatch consent rechecks, CAN-SPAM footers, and RFC 8058 one-click unsubscribe.
Send through AWS SES v2, Resend, or any SMTP server behind one RelayProvider contract. Webhook feedback is verified, normalized, and auto-pauses on bounce/complaint spikes.
Pixel and redirect tracking with per-contact opt-out, unique open/click stats per campaign, and no leakage of message existence to scanners.
Templates are typed design documents compiled to client-safe HTML at authoring time — compiler-checked, merge-tag aware, plain-text alternative included.
Every self-hosted mail tool promises inbox placement. The honest truth: SPF/DKIM/DMARC are necessary, not sufficient — IP reputation is the gate. Mail Pelican gives you the instruments instead of the fairy tale.
One check covers SPF, DMARC, MX, forward-confirmed reverse DNS, and the Spamhaus blocklist — the five things that actually decide whether you inbox, verified live before you send.
The same open-source filter ISPs run, shipped in the box. Campaign previews show the composite score and every rule your content trips — no black-box “spam score: bad”.
Gmail, Yahoo, and Microsoft send you free daily authentication telemetry as DMARC rua reports. Mail Pelican ingests and aggregates them: pass rates, failing sources, and spoofing attempts per domain.
A fresh IP sending full volume on day one is how domains burn. Relays can run a daily doubling warmup ramp enforced by the send pipeline, so reputation grows on schedule.
Every operation the console exposes is reachable through the /v1 API — and through
the bundled MCP server, your agents can build audiences, author drafts, and review previews
while sends stay human-confirmed.
One all-in-one container carries the API, the send worker, and the control console. Bring PostgreSQL 16 and Redis; keep your domains, your suppression lists, and your subscriber data on hardware you own.
The software is AGPL-3.0 and free forever. You pay for your own infrastructure — a VPS, PostgreSQL, Redis — and whatever your relay charges for delivery (SES is $0.10 per 1,000 emails). There is no per-subscriber fee and no feature gate.
Two things: deliverability instrumentation (preflight checks, rspamd scoring, DMARC ingestion, warmup ramps — nobody else in the category ships these) and agent-native operation through a first-class MCP server. It is also younger, with a smaller feature set in areas like automation sequences.
No. Most operators send through AWS SES, Resend, or an existing SMTP account. Self-hosting Mail Pelican means owning your data and control plane; the sending IPs can still be rented from a provider with established reputation — which is the setup we recommend.
Not if you do the boring parts — and we make them checkable. Run the preflight, publish the DNS records it verifies, warm up new IPs on the built-in ramp, and watch DMARC reports for failures. What we will not do is pretend a $5 VPS inboxes on day one.
An agent can do everything up to the point of no return: build audiences, author drafts, lint, preview, and prepare. The confirmed send requires a single-use token from prepare plus an explicit confirmation — a human, or an automation you deliberately hand that token to.
Star the repo, spin up the stack, and see what your spam filter sees.