Skip to content
Mariete

Autonomous agents that finish the work.

Describe a task in plain English — research these prospects, send this outreach, handle this inbox, follow up on these deals. Builder plans the work, does it end-to-end, and hands back a finished result. Think of it as a new hire who starts on their hundredth day, not their first.

Already running across 1,400+ teams on the Mariete platform.
builder · overview
Lead Researchlive
Outreachlive
Supportlive
Battlecardlive
CRM Synclive
Billinglive
50+agent types
What it does

Built around one idea: finish the work.

Configure once, run continuously. Every capability below is a production-grade surface, not a demo.

01

Configured, not prompted.

Every agent is a role: a Lead Research Agent, an Outreach Agent, a Support Agent. You set scope, tools and guardrails once — Builder runs them on a schedule or on demand.

02

Full workflows, end to end.

Agents chain steps — research a company, draft outreach, send it, follow up, escalate warm replies — so the result arrives done, not halfway.

03

Plugged into your tools.

Native connectors for CRMs, ticketing systems, data warehouses, inboxes and calendars. Builder reads from them and writes back to them.

04

Every run is reviewable.

Plan, tool calls, sources and output are captured. Approve, edit or override any step — then promote a run to a recurring job.

In the wild

Five ways teams actually ship with Builder.

Scenario · 01

A morning of research, done before stand-up.

Drop in 500 prospect companies. Builder visits each website, finds the right person on LinkedIn, checks if they match the kind of buyer you want, and writes a personalised opener for your sales team.

3 hours of work → 8 minutes
Try this scenario
builder · scenario · lead research
PLANOutreach · 12 steps
TOOLhubspot.contacts.search
TOOLgmail.drafts.create
OUTPUT48 drafts · awaiting approval
Intelligence suite

Four surfaces. One runtime. Zero duct tape.

Every Builder module shares the same orchestration layer, the same audit trail, and the same integration fabric — so a win in one surface lands everywhere.

builder · surface
Lead Researchlive
Outreachlive
Supportlive
Battlecardlive
CRM Synclive
Billinglive
Module · 01

Configured, not prompted.

Every agent is a role: a Lead Research Agent, an Outreach Agent, a Support Agent. You set scope, tools and guardrails once — Builder runs them on a schedule or on demand.

builder · runtime
PLANOutreach · 12 steps
TOOLhubspot.contacts.search
TOOLgmail.drafts.create
OUTPUT48 drafts · awaiting approval
Module · 02

Full workflows, end to end.

Agents chain steps — research a company, draft outreach, send it, follow up, escalate warm replies — so the result arrives done, not halfway.

builder · integrations
Lead Researchlive
Outreachlive
Supportlive
Battlecardlive
CRM Synclive
Billinglive
Module · 03

Plugged into your tools.

Native connectors for CRMs, ticketing systems, data warehouses, inboxes and calendars. Builder reads from them and writes back to them.

builder · audit
PLANOutreach · 12 steps
TOOLhubspot.contacts.search
TOOLgmail.drafts.create
OUTPUT48 drafts · awaiting approval
Module · 04

Every run is reviewable.

Plan, tool calls, sources and output are captured. Approve, edit or override any step — then promote a run to a recurring job.

Under the hood

The surfaces the product is made of.

Agent catalogue

Fifty-plus agents. One runtime.

Browse a library of specialists — from SDR to Controller — or clone one into your own. Every agent shares the same planning engine, tool catalogue and audit trail.

  • Sales · Marketing · Support · Ops · Finance
  • Clone, customise, version
  • Shared memory across agents in a workspace
builder · agent catalogue
Lead Researchlive
Outreachlive
Supportlive
Battlecardlive
CRM Synclive
Billinglive
Run surface

Watch the plan, review the output.

Every run opens a live trace: the plan, each tool call, the artefacts produced. Step in at any point to correct course — or approve the whole thing with a click.

  • Step-by-step reasoning
  • Artefact diffing
  • One-click rerun
builder · run surface
PLANOutreach · 12 steps
TOOLhubspot.contacts.search
TOOLgmail.drafts.create
OUTPUT48 drafts · awaiting approval
Schedules

Runs that run themselves.

Promote a successful run to a recurring job. Trigger on time, on webhook, or on a signal from another product in the Mariete suite.

  • Cron · webhook · signal triggers
  • Retry and alerting built in
  • Owned by a person, not a prompt
builder · schedules
Lead Researchlive
Outreachlive
Supportlive
Battlecardlive
CRM Synclive
Billinglive
For developers

Call it from anything.

CLI, SDKs and a single HTTP API. Every product on the platform exposes the same surface — same auth, same idempotency, same observability.

Node
Python
Go
Ruby
Java
cURL
NodePythonGoRubyJavacURL
# Kick off a Builder agent from the CLI
mariete run lead-research \
  --icp "Series-A SaaS, NA, 50–200 FTE" \
  --targets targets.csv \
  --output ./briefs

# Promote this run to a weekly job
mariete schedule lead-research \
  --cron "0 7 * * MON" --owner marc@acme.co
SDK
Typed clients
Webhooks
HMAC signed
SSO
SAML · SCIM
Specifications

The shape of the product.

What comes in, what goes out, what it runs on. Nothing hidden in a sales deck.

Inputs
Natural-language briefs, schedules, webhooks, CSV/CRM sync
Outputs
Drafts, emails, tickets, CRM updates, reports, API calls
Integrations
Salesforce · HubSpot · Zendesk · Gmail · Slack · Notion · 140 more
Runtime
Planner-executor loop with tool use and long-running jobs
Deployment
Hosted SaaS · SSO/SCIM · SOC 2 Type II in progress
Ready when you are

Start with Builder. Ship before the week is out.

No credit card. No sales call. Run one workflow end-to-end and decide.