Skip to content
Marshal AgentsAgent Orchestration

Agent orchestration, run by Marshal.

Marshal coordinates specialized agents, schedules, tools, permissions, and human review paths into one managed operating layer for founder-led teams. You hold the approvals without carrying the workflow.

Tailored to your processTrained on your businessIntegrated across your systems

Founder-led teams use Marshal

How orchestration works

Marshal routes work to approved action.

Marshal carries the work; you hold the approvals and the rules. Every workflow starts from a business event, routes to the right specialist, checks permissions, and pauses when judgment matters.

01

Trigger

Start from the signal.

Marshal starts from a schedule, form, inbox, CRM event, or manual request, then gathers the context the workflow needs.

02

Route

Send work to specialist agents.

Marshal routes each step to the research, drafting, CRM, finance, or workflow agent built for that job.

03

Check

Check context, permissions, and rules.

Each step uses approved tools and permitted data, so the workflow stays useful without leaving the process you approved.

04

Review

Pause for human judgment.

Sensitive actions wait in Slack, email, or the workflow surface until the owner approves the next move.

workflow:lead_followup4 steps · 1 human gate

Next action

Draft reply is ready. Marshal routes it to the owner before the CRM record changes.
agent path: enrich → draft → approve → sync

Control surface

What you can see, pause, and improve.

01 / Schedules

Marshal runs agents on your clock.

Schedules, business hours, time zones, and manual overrides stay visible. Every workflow shows when Marshal last ran it and when it runs next.

last_run4m 12s ago
next_runtomorrow 04:00 PT
missed0

02 / Approval gates

Judgment stays with the human.

Marshal inserts a gate wherever risk matters. The agent drafts, the owner approves, and sign-off is captured before the action moves forward.

gate_typehuman_approve
requested4m 12s ago
statuspending

03 / Logs and rollback

Every run is replayable.

Each step writes a structured record: inputs, outputs, cost, timing, owner, and side effects.

run_id#0182
duration4m 12s
rollbackavailable

>>> Observability

After-action reports your team can read.

Every agent run keeps the same log shape. Your team can scan the record when things are fine and see where a handoff is waiting for approval.

workflow:lead_followup · run #0182started 14:02:57 PT · duration 4m 12s6 of 7 steps complete · 1 awaiting approval
RUN_LOG · workflow:lead_followup · agent-v314:08:03
14:08:03OK

enrich_contacts · 14 leads matched for review

4m 12s
14:08:01RUN

draft_outbound · 14 personalized drafts ready

38s
14:07:24GATE

approval_gate · awaiting owner review

4m 12s
14:03:11OK

slack.post · #sales-ops notification sent

210ms
14:03:06OK

hubspot.upsert · contacts prepared for sync

480ms
14:02:58RUN

router · intent=qualify_lead · agent=enrichment_v3

240ms

Build a business that runs itself.

Join hundreds of small businesses operating at machine speed with agents on the job.