Email agents that know your brand
A tokenized generator that takes your brand assets, colours, type, and voice, then builds campaigns from templates our designers already made. Nothing sends without a person approving it.
Connects to
Your ESP, your CRM record, your brand kit
Produces
Flows, campaigns, segments, on-brand HTML
Approval
Human review before every send
Status
Live in engagement
Is this right for you
- ◆You have an ESP and a list, and campaigns still take weeks to ship
- ◆Your brand has real design standards and generic templates undermine them
- ◆You want production speed without handing your brand to a generator
The problem
Email breaks in two places.
Most teams lose email in one of two ways. Either production is the bottleneck, and a flow that should have shipped in a week takes a month because nobody has time to build it. Or production is fast and the output looks nothing like the brand, because whoever built it grabbed a generic template and changed the colours.
Speed and consistency usually trade against each other. The tokenizer is how we stopped making that trade.
What you stop losing
A month of build time is a month of sends that never happened.
A lifecycle flow that takes a month to build is a month of that flow not running. If a welcome sequence produces even a modest share of monthly revenue, the cost of a slow build is not the design hours. It is the revenue from sends that never went out.
The tokenizer moves campaign production from weeks to a day. The templates still take real design time, but that cost is paid once, at the start.
Framing, not a benchmark. Run the arithmetic against your own numbers.
How it works
From brand kit to approved campaign.
Two steps happen once, at the start. Two run every time a campaign goes out.
01 · Once
Tokenize the brand
Colours, type scale, logo lockups, spacing, button treatments, and voice guidelines become structured tokens. Not a style guide PDF, a machine-readable definition every generated asset inherits.
our team02 · Once
Design the templates
Our designers build the campaign and flow templates by hand, in your tokens. The agent composes within them. It does not invent layouts.
our designers03 · Every time
Compose from the record
Subject, preview, body, and segmentation pull from what the contact record actually says: past purchases, quote requests, last touch, lifecycle stage.
agent04 · Every time
Approve, then send
Everything queues for review. Approve records a decision. Edits are logged against the version that was approved.
youThe tokenizer
Your brand, in, campaign out.
The left panel is the brand definition. The email on the right was composed from it. Every value in the message traces to a token or a field on the contact record.
Campaign composer · Harbor Wellness
BRAND TOKENS
PALETTE
1d4f43
b8d4cb
f7f6f3
1a1a18
TYPE SCALE
- h1Canela 28/32
- bodyInter 15/24
- btnInter 13 · 600
VOICE
- ✓Plain, second person, short sentences
- ✓"appointment", "your therapist"
- ✕"unlock", "elevate", "journey"
- ✕Exclamation marks
TEMPLATE
Winback · single CTABuilt by our design team, 12 Jun
HARBOR WELLNESS
It has been a while, Ana
Your last appointment was in March, and the therapist you saw has openings again this month.
We kept your notes, so you would not have to start over.
Book your appointmentHarbor Wellness · 1400 Bay Street · Unsubscribe
◆Illustrative interface with example brand and data. The structure matches how the system works.
See your brand tokenized
Send us your brand kit and we will show you what your tokens look like, and one campaign composed in them, before you commit to anything.
Capability status
What runs today,and what is being built.
We keep this table current rather than describing the roadmap as though it shipped.
Live now
running in production- ◆Brand tokenization from your existing assets
- ◆Designer-built templates in your tokens
- ◆Composition from contact record context (CRM export or connected CRM)
- ◆Segmentation and flow logic in engagement
- ◆Human approval before every send
Shipping next
not available yet- Automatic template expansion from new brand assetsIn active development
- Per-segment layout variantsPlanned
- Subject line testing at flow levelPlanned
- Send-time optimisation per contactUnder research
◆Capability status verified 5 August 2026.See what changed
What stays human
Our designers build every template.
The agent fills them. It does not design them. That is the difference between output that looks like your brand and output that looks like AI made it.
A person on your team approves every campaign before it reaches a list. Approval is enforced in the system, not by policy, and there is no configuration that turns it off.
How we handle your data
- ◆Customer records stay tenant-scoped in your workspace
- ◆Raw contact content is not used to train public foundation models
- ◆De-identified outcome features may improve scoring under contract
- ◆Full export on request, no hold and no fee
- ◆Every field carries provenance: where it came from and when
Connects to
Works with the stack you already run.
Questions
What people ask before they start.
Will the emails actually look like our brand?
They are built from your tokens, and the templates were made by our designers using those tokens. The agent composes within a system a human designed, which means it is choosing words and segments rather than choosing fonts and layouts. If your brand has a defined type scale and palette, the output inherits it exactly rather than approximating it.
Can it match a voice, not just visual brand?
Voice guidelines are part of the token set: sentence length, formality, words you use, words you never use. The templates were written in that voice by people, and composition works within them. It is closer to filling a well-designed form than to free writing, which is why the output stays consistent across dozens of campaigns rather than drifting.
Do we need to leave our current ESP?
No. We work with the ESP you already use. CSV and CRM exports are verified end to end in this environment; Klaviyo and other ESP connectors are available on request and not yet verified on a live account here. The important difference is that the agent composes from your CRM record rather than from whatever data your ESP happens to hold, which is usually a richer picture for personalisation.
Can it send without us approving?
No. Approval is enforced in the system rather than by policy, and there is no setting that turns it off. Approving records a human decision against a specific version of the message; if the content is edited afterwards, the prior approval no longer applies. Consent and suppression are also re-checked at dispatch, not just when the draft was composed.
How long until the first campaign is live?
Tokenization and the template build are the front-loaded part, usually a few weeks depending on how well documented your brand already is. After that, campaigns compose in minutes rather than days. Most of the value shows up after the first month, because the expensive part has already been paid for.
What if we already have templates we like?
We tokenize those. The engine is built around your existing design system, not as a replacement for it. If you have templates that already perform, keeping them and making them composable is faster and lower risk than starting over.
What happens when we rebrand?
Update the tokens once. Every future asset inherits the change, which is the main practical advantage of a token system over a folder of template files. Existing templates are re-rendered in the new tokens rather than rebuilt by hand.
Who owns the templates?
You do. They were built for your brand, they are documented, and they leave with you if the engagement ends. The same applies to your token definitions and your data.
Is this a template library we log into?
No. There is no shared library and no menu of stock layouts. The templates are designed for your brand specifically, and the composition logic is configured for how your business actually segments and sells. That is why setup takes weeks rather than an afternoon.
Proof
Dormant to daily revenue in under six weeks
FUME
The email program was effectively unused. Flows, recovery, post-purchase journeys, and a segmented campaign structure were produced at a volume manual production does not reach, and the automations kept earning without anyone touching them.
Read the benchmark →Why this one
Email was one input among several in that engagement, and the benchmark says so, including where the paid test underperformed. We would rather show the whole picture.
Send us your brand kit.
We will tokenize it and show you one campaign composed in your own system, before you commit to anything.
Not sure which agent fits? Talk to us