SendFlow, API & MCP
Give your agent a language for the whole journey.
Describe the campaign. Let your agent build the logic. SendFlow makes every trigger, wait, branch, and exit readable, reviewable, and ready for version control.
Connect your AWS account. One flat SendOps fee.
More than a prompt
A concrete definition. A traceable change.
A workflow is a text artifact your team can inspect. Ask for a welcome sequence, review the proposed SendFlow, and understand exactly what will happen before it runs.
workflow "Trial onboarding" v1 {
enter on segment "trial-started"
exit "converted" when attr.plan != "trial"
send "welcome" via topic "onboarding"
wait 2d
if not exists(activity.project_created) {
call webhook "mobile-push" as "first-project" {
failed: send "need-a-hand" via topic "onboarding"
}
}
wait up to 7d until exists(activity.project_created) {
timeout: send "activation-tips" via topic "onboarding"
}
}From intent to a working journey
Draft. Review. Rehearse. Run.
- 01
Give the agent context
Share the customer goal, entry event, available attributes, and the condition that ends the campaign.
- 02
Review the definition
Inspect the SendFlow diff. Keep it in Git when you want the same review process as application code.
- 03
Rehearse the behavior
Validate the workflow, run a shadow cohort, and inspect the timeline and expected sends.
- 04
Operate with evidence
Review enrollment impact, activate with confirmation, and use version history and contact timelines to investigate.
A brief becomes a reviewable change
Let the agent draft. Keep the decision visible.
Give your agent the goal and the available customer signals. SendFlow makes its proposed behavior inspectable; MCP connects the work to your shared definitions.
“Draft a trial workflow that helps people create a project. Stop when they upgrade. Prepare it for review without activating live sending.”
A readable campaign definition, an inspectable rehearsal, and a human decision about when it goes live.
01
Inspect the available context
Have the agent check the trial audience, registered attributes, activities, and deployed templates. Resolve missing signals before it invents a branch around them.
02
Draft a precise definition
Ask for the entry condition, messages, waits, and conversion exit in SendFlow. Review the source directly or keep it in Git for a pull-request diff.
03
Rehearse and explain
Validate the definition and rehearse a bounded cohort. Ask the agent to explain a fast converter’s path and a contact who never activates, using the recorded evidence.
04
Review impact before activation
Inspect the audience or enrollment estimate before confirming live action. Each definition edit remains versioned, whether it came from a person, the API, or MCP.
The whole email toolkit
An agent that can see beyond the sequence.
Connect the native MCP server to work with the same audience, content, and delivery evidence as your team.
Audience & SendQL
Draft segments from contact attributes and activity. Preview membership before using the audience.
Templates & versions
Create messages, inspect deployment state, review changes, and restore earlier content.
Workflows & broadcasts
Prepare campaigns, estimate the audience, rehearse journeys, and review before sending.
Delivery & message search
Explain a missing email from the actual message events and contact timeline.
DMARC & reputation
Inspect domain sources and authentication evidence alongside campaign health.
Website & product activity
Connect Google Tag Manager return visits and product actions to a relevant next message.
How agent-led email work fits your team
Why is SendFlow useful for AI agents?
It represents the workflow as readable structured text. An agent can inspect the trigger, waits, branches, and exits, propose a precise change, and let the team review the diff and history.
Can I keep workflows in version control?
Yes. Keep SendFlow files in your repository and push definitions from CI. SendOps holds the shared editable definitions and keeps in-product version history. Git is your review process, not a required read-only ownership mode.
What can an agent do through MCP?
With appropriate permissions, an agent can work with contacts, audiences, templates, workflows, broadcasts, and reports. Broadcast sending and workflow activation require confirmation after the audience or enrollment impact is shown.
Does an agent’s edit have to open a pull request?
No. Definitions can be edited through the dashboard, API, or MCP. If your team wants pull-request review, keep a copy in Git and push approved definitions from CI. Each edit in SendOps is versioned.
Can an agent test a journey without sending emails?
Yes. Shadow mode rehearses a workflow on real contacts while recording external effects rather than performing them. Inspect the results and review enrollment scope before going live.
Your next workflow starts here
Build the journey. Keep the history.
Connect your AWS account. Build your first journey. Make the next email count.
Free plan available. No credit card required.