Grain
PricingCompareDocsChangelogDownload
Sign inDownload →
PRODUCT · AGENT FLEET

Run a hundred agentslike one team.

Org chart your agents. Set policies once — spend caps, escalation rules, who-talks-to-whom. Fleet tells you at a glance: who's busy, who's stuck, who's about to miss SLA.

Start free
Fleet · mercury-support● 47 ACTIVE· 12 regions · p95 142mslast hour$2.41 spent· 18,402 runs
TEAMS
▼ Front desk
4
Triage
Router
Greeter
Language-detect
▼ Billing
8
Refund-eu
Refund-us
Invoice-explain
Chargeback
▶ Technical
12
▶ Accounts
9
▶ Escalation
3
▶ Analytics
11
ORG CHART · last 60s
TriageTech-routeBilling-routeAccount-routeOutage-agentAPI-debugRefundPassword-resetSSO-config
LIVE
Triage
routed billing
0.1s
Refund-eu
issued €28
0.3s
Outage-agent
waiting on PD
0.8s
Password-reset
✓ done
1.2s
Refund
escalate → human
1.4s
Triage
routed tech
1.6s
API-debug
fetching logs
1.9s
Triage
routed accounts
2.1s
SSO-config
✓ done
2.4s
Triage
routed billing
2.7s
WHY A FLEET, NOT A STACK

Managing 100 agents ≠ 100× work.

⚖︎

Policies, once

Set a spend cap, a retry rule, a PII policy at the fleet level — every agent inherits. Override per-agent only when you must.

Handoffs are typed

Agents pass typed payloads, not loose strings. Route diffs like PRs. The handoff graph is the org chart.

Watch from one screen

Every agent's status, queue depth, error rate, and spend — one dashboard, 60fps live. No more tab hunting.

Pause the whole thing

One toggle pauses the entire fleet. Another drains cleanly. Rollback the whole fleet to yesterday's revision.

FLEET POLICIES

One policy.
Whole fleet.

Define guardrails once. "Never send PII to US regions." "Cap any run at $0.05." "Escalate anything touching a refund over $500." Grain enforces them at runtime.

fleet.policies.yaml
budget: $0.05 # per run, hard cap
retries: 2 # exp. backoff
pii_policy: redact_before_log
escalate_if:
  - refund.amount > 500
  - customer.tier == "enterprise"
  - confidence < 0.7
regions:
  eu: [anthropic-eu, openai-eu]
  us: [anthropic-us, openai-us]
✓ applied to 47 agents · 0 violations
2,400+
agents under management
47ms
policy eval overhead
0
policy violations (prod)
1-click
fleet-wide rollback

Run your fleet
like an org.

Start with 3 agents. Scale to 3,000.

Download GrainSee pricing
Grain

The multi-agent command center. Hire an entire AI team in 30 seconds.

Product
BriefsAgentsFramesAgent FleetRemote Control
Platform
OverviewChannelsSkillsKnowledgePricing
Resources
DownloadDocsChangelogStatusDiscordGitHub
Compare
vs OpenClawvs Claude Codevs Viktorvs Codexvs ChatGPTvs Tasklet
Company
AboutCareersPressSecurityPrivacy
© MMXXVI · GRAIN · SAN FRANCISCOSOC 2 TYPE IIHIPAAGDPR