under construction
The MAP to replace all MAPs.
Built by operators who've spent years inside the big platforms — cloning programs, chasing tokens, navigating database scale limits, herding approvals... The list goes on. The MAP we need won't be built by an enterprise org. It's up to us to make this ship happen..
see it workBuilt by Operators.
We've run them all. We know the forty-minute program clone, the token nobody dares touch, the approval that gets lost in a slack thread, the sync queue you check before your coffee. veloquent is the platform we kept waiting for someone else to build — so we're building it, in public.
It isn't finished. That's the point of this page: you're early, and everything below is real, running software.
the assistant
Ask veloquent to build like it's your teammate.
“Draft the launch email for enterprise trial accounts, run QA, and submit it for review.” The assistant queries your warehouse, sizes the audience, drafts the email, wires the send, and submits it — narrating each step as a tool call you can open.
go deeper →
campaigns
Everything lands in one tree.
What it built is a real campaign, not a pile of loose assets: every workflow with its audience, email, and delivery in view, tokens inherited from the folders above it, and review status rolled up where you can see it.
go deeper →
review
Nothing sends until a human says so.
Every workflow the assistant (or anyone else) builds waits in one review queue. Approve it in the app — or right from Slack, where the notification carries the same buttons.
go deeper →
veloquent app 2:47 PM
Fall Product Launch is ready for review
Basic Batch · Enterprise Trial Accounts · 5 recipients
Decisions land back in veloquent with a full audit trail.
workflows
Flows that do real work.
Sends, waits, event branches, a step that writes a line of copy per person, a 50/50 test. Each step runs as its own job, which is what makes the boring guarantees hold: a one-week wait is still a week later, a step that errors retries instead of quietly dropping someone, and a flow can check your warehouse mid-journey — not just when they entered.
go deeper →
what's taking shape
The pieces, and where they stand
One tree. Everything belongs to a campaign.
Batch sends, evergreen journeys, emails, audiences — all of it lives inside a campaign, and every campaign lives in one tree. No more guessing which corner of the app something is hiding in.
Tokens that show their work.
Set a token once on a folder and everything underneath inherits it. Override it on one campaign when you need to. Either way, the UI shows exactly where each value comes from — no more sends going out with last quarter's date.
Approvals that don't live in email.
Mark a send or a journey ready for review and it lands in one queue, with status rolled up on the campaign. Approvers decide there or straight from Slack — no screenshots in threads.
An agent that speaks fluent MOPs.
Ask for what you want in plain English, or click it together yourself — same result, interchangeable. The agent drafts and explains its work, and nothing sends until a human hits approve.
Drafts from your own data.
Emails, workflows, and audiences generated from the sources you connect — routed through Vercel's AI Gateway, so the right model does the job and you're never locked to one vendor.
A Slack agent riding along.
Approvals, status checks, quick changes — handled in Slack, where your team already is. The platform comes to you, not the other way around.
Plugged into your stack.
Query your warehouse for audiences and sync Salesforce both directions — so the same platform carries your PLG motion and your traditional demand gen.
QA you define.
Checks run before anything reaches review — links resolve, UTMs are right — and we're going past the technical: brand-voice and guideline checks you can tune yourself, in the app.
the stack
Built on Vercel primitives. All of it.
veloquent runs on the same infrastructure that serves some of the busiest sites on the web, which in practice means three things: your queries don't time out when the database gets big, there's no maintenance window the week of your launch, and we spend our time on the product instead of the plumbing.
- compute that follows demand
- Your 9am send doesn't wait in line behind anyone else's.
- tracked jobs and queues
- A failed step retries itself instead of dropping the person.
- scheduled runs
- Recurring sends fire on time without a server we babysit.
- managed storage
- Images and files are someone else's uptime problem.
- one AI gateway
- The right model per job, and no vendor lock-in.
- agents
- The assistant and the Slack agent run on the same rails.
build log
[done] campaign tree — folders, nesting, campaign homes
[done] cascading tokens with a visible inheritance map
[done] unified review queue + ready-for-review states
[done] slack approvals — decide without leaving the thread
[now] tree drag-and-drop, campaign duplication
[next] a slack agent you can talk to, not just approve from
[next] configurable qa — brand voice and guidelines, judged in review
[next] richer a/b testing via branched flows
[next] the rest — this list stays honest
The longer version lives on the blog.
Follow the build.
Sign up if you'd like to keep apprised of the latest ships.