Integration
MCP for SendGrid: Claude, reading your account with a scoped key.
You own deliverability on SendGrid. The transactional stream is healthy enough, but the marketing sends have started brushing spam folders, the bounce rate ticked up last month, and you half-suspect a DKIM record went stale after the last DNS migration. So you open five tabs — Suppressions, Sender Authentication, Stats, the event feed, your DNS console — and start reconciling by eye. Ask Claude for help and it gives you good advice about all of it. Generic advice. It has never seen your suppression groups, your authenticated domains, or last week's bounce curve. An MCP for SendGrid — a way to hand Claude a scoped, read-only key instead of general knowledge — is what closes that gap. Claude stops describing the audit and starts running it against your real account.
What "MCP" actually means here.
MCP stands for Model Context Protocol, an open standard from Anthropic that lets Claude reach outside the chat window. Think of it as a plugin system, except the plugins don't add small features — they give Claude entirely new capabilities: tools that read your data and build real artifacts. Orbit is one such MCP, built for lifecycle and deliverability work. Install it, connect a scoped SendGrid key, and Claude gains the ability to read your account — and, if you grant it, write to it. Orbit is not an official Twilio SendGrid product; it's an independent MCP that calls SendGrid's public v3 API with the key you provide, and nothing leaves your machine except the calls to SendGrid you configured. The full walk-through of what an MCP is and how it changes the work lives on the MCP for marketing page →.
What Orbit does with SendGrid today.
No hand-waving. SendGrid's v3 API is deep and developer-first, so most of this is fully native. Where there's a limit, it's named — and we're careful to separate three different things: what SendGrid's API genuinely allows, what Orbit simply hasn't built yet, and what Orbit holds back on purpose because it's read-only-first and the action is a high-consequence write.
Check the connection and see your scopes
Works todaySendGrid exposes a scopes endpoint that returns exactly what your key is allowed to do. Orbit calls it to confirm the key is live and to show you, in plain terms, which reads and writes you granted — so there's no guessing about what the audit can and can't touch.
Read suppressions
Works todayGlobal suppressions and every suppression group — bounces, blocks, spam reports, invalid emails, and unsubscribes — read natively. This is the raw material for diagnosing why mail isn't landing.
Read sender-authentication status
Works todayWhich domain-authentication records SendGrid considers valid (SPF and DKIM via CNAME), link branding, and the verification state of each. Claude can tell you which records are green and which are still pending in DNS.
Pull stats
Works todayGlobal and category stats from the Stats API — delivered, opens, clicks, bounces, spam reports, unsubscribes — aggregated by day. The numbers a deliverability review is actually made of.
Read event webhook config
Works todayThe event-webhook settings on your account: which events are enabled, the POST endpoint, and whether signed-event verification is on. Orbit reads the configuration so Claude can reason about your event pipeline; it doesn't stand up a receiver for you.
Read Marketing Campaigns
Works todaySingle Sends, automations, contact lists, and segments from the Marketing Campaigns API — read natively, so Claude can see what you've already shipped on the marketing side, not just the transactional stream.
Build and update Dynamic Templates
Works todaySendGrid's Transactional Templates API supports full CRUD on Dynamic Templates. Orbit composes the copy, generates HTML that keeps your Handlebars substitutions intact, and pushes a new active version. A write — so it stays behind a scope you grant on purpose.
Send a test email
Partial — named limitSendGrid's Mail Send API has no separate test mode — a send is a send — but it does have sandbox mode, which runs full validation and returns success without delivering. Orbit validates in sandbox by default and only fires a real message to your own address when you ask and the send scope is on.
Dedicated-IP management
Partial — named limitOrbit reads your dedicated IPs, their assignment to subusers, and their warmup state, so an audit can flag an IP that's mid-warmup or unpooled. Driving the warmup schedule itself stays in SendGrid's UI; Orbit reports on it rather than operating it.
Trigger a Single Send or bulk campaign
Held back by designReading Marketing Campaigns works today. Actually scheduling or firing a Single Send is a high-consequence write Orbit holds back from by default — read-only-first means bulk sends stay in your hands, in SendGrid, on purpose. Not an API dead-end, a deliberate line.
The short version: with a read-only key, Orbit can audit your entire SendGrid deliverability picture — suppressions, sender auth, stats, event config, dedicated IPs — and tell you what's wrong. Widen the scope and it builds Dynamic Templates too. The one line it won't cross without you asking is firing a bulk send: that stays yours.
The deliverability audit it was built for.
This is where a read-only SendGrid key earns its place. Point Orbit at your account and ask Claude to run the checks a deliverability owner does by hand every quarter — except it reads the real state instead of asking you to describe it.
- Domain authentication — SPF, DKIM, DMARC. Orbit reads SendGrid's view of which authenticated domains are valid and which records are still pending, then cross-checks against the DKIM and SPF that SendGrid expects in your DNS. DMARC isn't in SendGrid's API — it lives in your DNS as a
_dmarcTXT record — so Claude reasons about your policy (p=none,quarantine, orreject) from the record itself and tells you whether your SendGrid auth aligns with it. - Suppression health. Orbit sizes your bounce, block, spam-report, and invalid-email lists and reads them against your send volume, so a rising hard-bounce rate or a spam-report spike surfaces as a number, not a hunch — and it separates list hygiene problems from authentication problems, which look the same from the inbox but need different fixes.
- Dedicated-IP warmup.If you run dedicated IPs, Orbit reads their assignment and warmup state and flags an IP that's sending above a sane warmup ramp or sitting unpooled — the kind of thing that tanks a sender reputation weeks before the stats show it.
None of this needs write access. The whole audit runs on a Restricted Access key with read scopes only — which is exactly how you should start.
How to connect it.
One credential, and you decide how much it can do. In SendGrid, go to Settings → API Keys and create a key. Paste it into Orbit's config and you're connected.
- Start with Restricted Access, read scopes only.Grant the read scopes for suppressions, stats, sender authentication, templates, and marketing — and nothing that can send. That key runs every audit on this page and physically cannot mutate your account. It's the read-only-first posture enforced at the source, which is the right default.
- Widen only when you want writes. Add the template-write scope when you want Orbit building Dynamic Templates, or the mail-send scope when you want real test sends. Each is a deliberate step, not a blanket Full Access key handed over on day one.
The key is stored as a secret and never leaves your machine in plain text. That's the whole setup — no OAuth dance, no app to install on SendGrid's side. If the key is missing scopes for something you ask, Orbit tells you which scope to add rather than failing silently.
What's deepest on Braze — and why we say so.
Orbit works with every major ESP, but its deepest single-platform integration is Braze. A handful of Orbit's tools are built around concepts that only exist in Braze — Canvas flows, Content Blocks, its media library — and those stay Braze-named on purpose. Faking a SendGrid equivalent would just be a worse version of a thing SendGrid's model doesn't have.
For SendGrid, that trade barely bites: the portable work — auditing deliverability, reading the program, building templates, running the numbers — is exactly what a SendGrid-based developer or deliverability owner needs. And if you later graduate to a lifecycle-native platform, the tooling and the discipline travel with you.
The half that isn't about the API at all.
Reading and writing SendGrid is the connection. The reason to bother is everything Claude does once it's connected. Orbit ships 79 lifecycle skills — structured protocols for the actual jobs: auditing deliverability, composing an email, laddering a win-back, scoring a subject line. Those are ESP-agnostic; they work the same whether the account underneath is SendGrid or anything else.
Underneath sit 130 tools — an MJML pipeline that compiles to clean, dark-mode-safe HTML, a render-and-QA gate that catches the clipping and inbox breakage before you send, and the calculators a real test needs: sample size, holdout, replenishment timing. SendGrid gives you the send and the deliverability rails. Orbit is the practitioner sitting next to them, reading the same dashboards you do.
Is this for you?
This is for the developer or deliverability owner who runs SendGrid and wants a straight read on it — the person maintaining the sender auth, watching the bounce curve, and keeping the suppression lists honest. If that's your week, Orbit hands you a Claude that reads the real state through a key you scoped yourself. It's a weaker fit if you want a hands-off tool that runs the program while you watch a dashboard. Orbit doesn't run your program, and it won't fire your bulk sends. You do. It just gives you a Claude that works like a deliverability practitioner instead of a search engine — and a straight answer about what your ESP can and can't do, which is rarer than it should be.
Try it.
Orbit is free — every skill and tool included, no seats and no subscription. Create a read-only SendGrid key, paste it in, and Claude goes from advisor to deliverability audit the same afternoon.
Run a different ESP?
The same honest treatment, per platform:
Frequently asked.
What is an MCP for SendGrid?
An MCP — Model Context Protocol, an open standard from Anthropic — gives Claude real tools instead of general knowledge. Connected to SendGrid through Orbit with your own scoped API key, Claude can read your suppression lists, sender-authentication status, and stats, and build Dynamic Templates in your account — rather than describing what you should do and leaving the doing to you. It is not an official Twilio SendGrid product; it is an independent MCP that calls SendGrid's public v3 API with the key you provide.
Can I connect SendGrid with a read-only key?
Yes, and it's the recommended way to start. SendGrid API keys are scoped, so you can create a Restricted Access key scoped to reads only — suppressions, stats, sender authentication, templates, marketing — and nothing that can send or mutate. Paste that into Orbit and every audit still runs; the write features simply stay dark until you decide to widen the scope. Orbit's posture is read-only-first, and SendGrid's key model lets you enforce it at the source.
What can Orbit read for a deliverability audit?
The pieces a real audit is made of. Orbit reads your suppression groups and global suppressions — bounces, blocks, spam reports, invalid emails, and unsubscribes — so Claude can size and reason about why mail isn't landing. Sender-authentication status comes back too — which domain-authentication (SPF and DKIM) records SendGrid considers valid, link branding, and the verification state of each. And it pulls your Stats API numbers — delivered, opens, clicks, bounces, spam reports, unsubscribes, aggregated by day. DMARC lives in your DNS rather than SendGrid's API, so Orbit reasons about your policy from the record itself.
Can Orbit build and update SendGrid templates?
Yes, if your key carries the template-write scope. SendGrid's Transactional Templates API supports full CRUD on Dynamic Templates, so Orbit can create a versioned template and update it. Orbit composes the copy, generates HTML that leaves your Handlebars substitutions intact, and pushes a new active version into your account. This is a write, so it stays behind a scope you grant on purpose — read-only-first means it's off until you turn it on.
Can Orbit send a test email through SendGrid?
It can send a real one, and that's the honest framing. SendGrid's v3 Mail Send API doesn't have a separate test-send mode; a send is a send. What it does have is sandbox mode — a flag that runs the full validation path and returns success without delivering anything — which is the safe way to prove a payload is well-formed. So Orbit will validate in sandbox mode by default, and only fire a genuine message to your own address when you ask for it and the send scope is granted. Alongside that, Orbit runs its own local QA gate on the HTML — dark mode, clip length, accessibility — which needs no send at all.
Is SendGrid the right home for a serious lifecycle program?
SendGrid is strongest as a deliverability and transactional-email engine — the layer that gets mail to the inbox reliably, with scoped keys, dedicated IPs, and event data that developers trust. Its Marketing Campaigns side runs Single Sends and automations, which cover a lot of SMB lifecycle work. Where it strains is deep, branching, behaviour-driven journeys and true incrementality testing across a large audience; if that's the week you're living in, a lifecycle-native platform like Braze is the graduation. Orbit works with both, deepest on Braze — so the tooling and the discipline travel with you if you move.