Skip to content
Bloqr AI Bloqr AI™

Bloqr Mail (bloqr-mail Worker)

Switch to Zen Mode

Bloqr Mail (bloqr-mail Worker)

Status: placeholder. Every inbound message is currently rejected with a clear “not live yet” bounce. This document describes the DNS/Email Routing setup that’s already live, why the worker exists, and the plan for what it becomes.

Source: worker/mail/ in this repo (mail-worker.ts, wrangler.toml, tests). Deployed as its own Cloudflare Worker script, bloqr-mail — separate from the main bloqr-compiler Worker, but versioned in this monorepo since there’s no dedicated bloqr-mail repo.

Terminal window
deno task wrangler:mail:dev      # local dev
deno task wrangler:mail:deploy   # deploy to Cloudflare

Why this worker exists

Two Bloqr features, still being spec’d, both need a dedicated inbound-mail surface at *@mail.bloqr.dev:

  1. Inbound SMTP-API (the primary driver). Some power users already fold SMTP into their workflows and want Bloqr to be a drop-in target for that — send a request by email, get a result back, no HTTP client involved. Most business is expected to come through the HTTP API; this is about covering that use case for users who need it, not replacing the HTTP API.
  2. Outbound filter-list delivery. Compiled filter lists/artifacts sent as email attachments from *@mail.bloqr.dev addresses (e.g. a scheduled or on-demand delivery of a user’s compiled list straight to their inbox).

Both are premium, not guaranteed for v1 — demand is being validated via a post-signup survey before committing engineering time to the full spec. bloqr-mail exists now so the DNS/Email Routing plumbing, naming, and code location are settled ahead of that decision, without pretending the feature is live.

Why a separate Worker

bloqr-mail is intentionally not a route inside the main bloqr-compiler Worker:

  • Cloudflare Email Routing dispatches inbound mail to a destination Worker, a first-class binding distinct from Custom Domains / HTTP routes. It needs its own script.
  • Keeping it separate means an inbound-mail bug or outage can’t take down the HTTP API Worker, and vice versa.
  • It has no fetch() handler and no public HTTP surface — see Domain / DNS architecture for why that matters and what went wrong when it briefly had one.

The intended design is for bloqr-mail to stay a thin protocol adapter: parse the inbound message, resolve the requesting user, then dispatch into bloqr-compiler’s existing auth/billing/compilation logic — not a second implementation of that logic. See Planned architecture below.

Domain / DNS architecture

bloqr-mail sits behind mail.bloqr.dev, deliberately separate from the bloqr.dev apex:

HostnameOwnerPurpose
bloqr.dev (apex)Google WorkspaceReal human mailboxes (jayson@bloqr.dev, etc.). MX: aspmx.l.google.com + alt*.aspmx.l.google.com. SPF/DKIM/DMARC all Workspace-owned.
mail.bloqr.devCloudflare Email Routingbloqr-mail’s inbound surface. Own MX (route1/2/3.mx.cloudflare.net), own SPF, own DKIM, own bounce handling (cf-bounce.mail.bloqr.dev) — fully independent of the apex.

This split is required, not a preference. MX records are exclusive per hostname — a domain can’t have “some mail to Google, some to Cloudflare” at the same name. Putting Cloudflare Email Routing on its own subdomain is the standard way to get programmatic mail processing without touching Workspace’s ownership of the apex.

If a feels-like-apex address is ever needed (e.g. bounces@bloqr.dev), the pattern is: Workspace stays the single MX owner and forwards/aliases that specific address to its @mail.bloqr.dev equivalent (Workspace Admin → Gmail → Routing, or a per-user forward). Cloudflare never needs apex MX for that to work.

Routing rules (already live)

Configured in Cloudflare Email Routing (dashboard or API), not in wrangler.toml:

  • to: bloqr@mail.bloqr.dev → Worker bloqr-mail
  • catch-all (*@mail.bloqr.dev) → Worker bloqr-mail

No Workers Custom Domain — on purpose

bloqr-mail exports no fetch() handler. Email Workers only need email(). Cloudflare Email Routing dispatches to the destination Worker directly — there is no legitimate reason for HTTP traffic to ever reach this script.

This was learned the hard way: under the previous name (bloqr-email), a Workers Custom Domain accidentally pointed email.bloqr.dev at this script’s HTTP surface. Since it had no fetch() handler, every stray HTTP request (bots, browsers, monitors hitting the bare hostname) errored in Observability with Handler does not export a fetch() function. That Custom Domain/DNS record has been deleted along with the email.bloqr.dev hostname entirely — the worker now only exists behind mail.bloqr.dev via Email Routing. Do not re-add a Custom Domain for this script.

Current behavior (placeholder)

worker/mail/mail-worker.ts exports only email():

export default {
    async email(message, env, ctx) {
        const recipient = parseRecipient(message.to);
        const route = ROUTES[recipient.local]; // ROUTES is empty today
        if (!route) {
            message.setReject(NOT_LIVE_MESSAGE);
            return;
        }
        await route({ message, env, ctx, recipient });
    },
};

Every inbound message — bloqr@mail.bloqr.dev, the catch-all, anything — is rejected with a clear bounce ("Bloqr Mail does not accept inbound requests yet...") rather than silently accepted and dropped. A paying customer’s request disappearing into a black hole is worse than a bounce they can act on.

Planned architecture (once the spec lands)

Routing convention: Cloudflare subaddressing

Subaddressing is already enabled on the bloqr.dev zone (local+detail@domain). The intended convention:

<command>+<userToken>@mail.bloqr.dev
  • local-part (compile, list, …) selects the command, via the ROUTES registry in mail-worker.ts — adding a real command is a one-line addition once the spec lands.
  • +detail subaddress carries an opaque per-user token, used to identify the requester for the entitlement/billing check.

parseRecipient() in mail-worker.ts already implements this split and is unit-tested.

Business logic stays in bloqr-compiler

bloqr-mail should not grow a second implementation of auth, billing-tier checks, or compilation logic. The recommended path: a Cloudflare Service Binding from bloqr-mail to bloqr-compiler (Worker-to-Worker call, no public network hop — see the commented-out [[services]] stanza in worker/mail/wrangler.toml), reusing the same worker/utils/route-permissions.ts / worker/utils/user-access.ts patterns the HTTP API already uses for entitlement checks. Since this is a premium, per-user-gated feature, the entitlement check must happen before any compilation work starts, not after.

Outbound delivery (attachments)

Sending compiled lists as email attachments is a bloqr-compiler-side concern (triggered via createEmailService(), not something bloqr-mail does), and it’s currently blocked on a real gap: EmailPayloadSchema (worker/services/email-service.ts) has no attachment field. Adding one — and the per-provider encoding work it implies (Resend supports attachments natively; the CF REST/binding paths need MIME multipart built by hand) — is follow-up work, not yet started.

Unrelated to bloqr-mail directly, but discovered during this work: CfEmailWorkerService (the lowest-priority outbound fallback in worker/services/email-service.ts, behind Resend and CF Email REST) sends FROM: notifications@bloqr.dev / noreply@bloqr.dev — bare apex addresses. Now that the apex is exclusively Google Workspace, that fallback tier’s FROM domain has no Email Routing behind it and will likely fail delivery. Low-stakes today (it’s the last-resort tier), but the FROM addresses should move to @mail.bloqr.dev when someone picks this up. See worker/mail/wrangler.toml and Hybrid Email Architecture for more detail.

Testing

worker/mail/mail-worker.test.ts covers parseRecipient() (plain vs. subaddressed local-parts) and the placeholder email() reject behavior. Run with the standard worker test task (deno task test:worker); type-checked via deno task check:worker.

Further reading

© 2026 Bloqr AI™, a trademark of Bloqr Systems. Created by Jayson Knight.