Grain
Product ▾
PricingCompareDocsChangelogDownload
Sign inDownload →
YOUR AI AGENT TEAMNEW · 11 SPECIALISTS

Hire an entire
agent team
in 30 seconds.

Grain is a multi-agent command center. Give it a goal in plain English — a team of named specialists researches, plans, executes, and ships you a polished artifact. Not a chatbot. A team.

Download Grain — freeSee pricing
⌘ desktop · web · slack · telegram  ·  no credit card  ·  1,000 free runs / mo
app.grain.sh / threads / q1-pipeline-review
● 4 agents working·$0.42 spent
THREAD · agentic mode
You
Build a Q1 pipeline review for the board. Pull from HubSpot, find the deals at risk, and draft a battlecard for the top competitor we're losing to.
At
Atlas · coordinator · routed to 3 agents
Breaking this into 3 parallel tasks:
✓ Pulse → pipeline analytics from HubSpot
● Closer → deals-at-risk scoring
○ Scout → competitor intel + battlecard draft
Pu
Pulse · analytics · 12s
🔧 hubspot.deals.list({ stage: ['proposal', 'negotiation'] })
🔧 hubspot.deals.aggregate(by: 'owner', period: 'q1-2026')
✓ 47 open · $2.8M ARR · 12 at risk
Cl
Closer · scoring deals…
Cross-referencing engagement signals, competitor mentions, and stage velocity for 12 deals…
DASHBOARDq1-pipeline-review.frame
Share·Export PDF
BOARD REVIEW · Q1 2026

Pipeline at a glance

PIPELINE ARR
$2.8M
+18%
OPEN DEALS
47
+6
AT RISK
12
-3
WEEKLY VELOCITY
DEALS AT RISK · TOP 4
Acme Corp$420KStalled 18d
Globex$310KCompetitor mentioned
Initech$180KChampion left
Stark Ind.$95KBudget freeze
Scout is writing the competitor battlecard…
RUN BY TEAMS AT
Salestools.io
cronis
speexx
Adobe
salesforce
WHY GRAIN

Weeks of work.
Overnight.

One prompt. A team of 10 specialists. Polished artifacts back in your lap before standup. Not autocomplete — actual output you'd ship.

8min
Median time to a finished Frame
4×
Productivity vs solo with one chatbot
90+
Skills your fleet can install
16
Frontier models, swappable in 1 click
TWO EXECUTION MODES

One agent. Or the whole swarm.

Pick the right scope for the job. Grain handles both — and lets you switch mid-thread.

◆
MODE 01

Task Mode

One agent runs one task end-to-end. Best for focused work — write a memo, fix a bug, generate a battlecard. Predictable, fast, cheap.

$ grain run scribe
→ Scribe drafted memo · 11s · $0.04
✓ 1 artifact ready
◈
MODE 02

Agentic Mode

One prompt → automatic breakdown into parallel linked tasks across the swarm. Best for projects: "plan and ship a marketing campaign."

$ grain run "ship q1 review"
→ Atlas split into 3 tasks
→ Closer · Pulse · Scout running…
✓ 1 dashboard + 1 battlecard ready
THE ROSTER

Eleven specialists.
One command center.

Every agent has a role, a model preference, and its own toolkit. Pin the ones you use; ignore the rest. Configure each one's persona, prompt, and approval rules.

Meet the team →
◐
Scout
Research & discovery
● ACTIVEv2.4
◆
Closer
Sales-side work, deal motion
● ACTIVEv2.4
⬢
Shield
Security, compliance, audit
● ACTIVEv2.4
✎
Scribe
Writing, briefs, docs
● ACTIVEv2.4
∿
Pulse
Analytics, monitoring, metrics
● ACTIVEv2.4
◉
Beacon
Outreach, notifications, comms
● ACTIVEv2.4
✦
Spark
Creative, marketing content
● ACTIVEv2.4
◰
Forge
Engineering, code, fixes
● ACTIVEv2.4
◊
Sentinel
QA, testing, gating
● ACTIVEv2.4
✱
Atlas
Coordination, project orchestration
● ACTIVEv2.4
⟳
NEW · SUPERAGENT
BY NOUS RESEARCH

The only agent with a built-in learning loop.

It creates skills from experience, improves them during use, nudges itself to persist knowledge, searches its own past conversations, and builds a deepening model of who you are across sessions.

01
Creates skills
Turns every successful run into a reusable skill
02
Improves mid-run
Refines its own approach while working, not after
03
Persists knowledge
Nudges itself to save what matters across sessions
04
Searches memory
Queries its own past conversations like a knowledge base
05
Models you
Builds a deepening map of your preferences over time
THE ARTIFACT SYSTEM

Agents don't just
write text. They ship Frames.

Block-based, polished, shareable documents. Live chat-to-edit, PDF export, public share links, custom accent colors. Five types — color-coded by purpose.

Explore Frames →
REPORT.frame
Report
Data-rich docs
DASHBOARD.frame
Dashboard
Metrics & KPIs
PRESENTATION.frame
01
02
03
Presentation
Slide-style decks
MEMO.frame
Memo
Long-form writing
BATTLECARD.frame
US
THEM
Battlecard
Competitive intel
LIVE EXECUTION

Watch every thought as it lands.

Split-thread + artifact view. Live diffs streamed every 10s. PR-style review comments. Browser preview with device toggles. You always know what your fleet is doing.

FORGE · streaming · refresh every 10ssrc/api/billing.ts
function refundCharge(id: string) {
- return stripe.refunds.create({ charge: id });
+ if (!await canRefund(id)) throw new Forbidden();
+ const r = await stripe.refunds.create({ charge: id });
+ await audit.log('refund', { id, by: ctx.user });
+ return r;
}
💬
Sentinel · review
Looks good — but add a rate-limit check before the refund call.
⟳
Live diffs every 10s
See code stream in real-time, with branch-aware highlighting.
💬
Inline review comments
PR-style comments on diff lines. Approve, request changes, or push back.
🔧
Tool calls in-chat
Every API call, search, and side-effect surfaces in the thread.
⛨
Permission modes
default · acceptEdits · plan · auto · bypassPermissions.
EVERYWHERE YOU WORK

Same fleet.
Six front doors.

Run agents from a desktop app, the web, Telegram, Slack, WhatsApp, Teams — even email. Same threads, same memory, same approvals queue everywhere you go.

All channels →
⌘
Desktop
macOS · Windows · Linux
Grain
◉
◆
✎
∿
⬢
● THREAD · live
Closer
Drafted email to Acme. Awaiting review.
Pulse
Pipeline updated · 47 deals
Native Tauri app with auto-updates, system tray, global hotkeys, and offline thread cache.
#
Slack
Block Kit · team-friendly
At
Atlas APP
Q1 review ready. 2 artifacts attached:
DASHpipeline.frame
BTLcompetitor.frame
ApproveChanges
Mention agents, attach artifacts, approve in-channel.
✈
Telegram
Async · solo-friendly
draft a follow-up to mercury
Beacon · outreach
Drafted. Tap below to review or send.
ReviewSend
Run the swarm from your phone, on the move.
ALSO ON
WhatsApp
Microsoft Teams
Discord
Email
Web app
Obsidian
MODEL-AGNOSTIC

16 providers.
One dropdown.

Latest from every frontier lab. Hosted open weights. Local Ollama. Bring-your-own-key. Per-agent model routing — and evals tell you whether the switch was safe.

Anthropic
Claude · Opus 4.7, Sonnet 4.6, Haiku
OpenAI
GPT-5.5 · o-series · Codex CLI
Google
Gemini 3.0 Pro · Gemini CLI
xAI
Grok-4 · Grok-mini
DeepSeek
DeepSeek V4 · R1
Mistral
Large 3 · Mixtral
Cohere
Command R+ · Embed v4
Groq
Llama 4 · LPU inference
Together
200+ open weights
Fireworks
Hosted open models
OpenRouter
One key, every model
Replicate
Custom + multimodal
Ollama
Local · on your laptop
LM Studio
Local · GUI runner
Perplexity
Sonar with web access
Bring your own
Any OpenAI-compatible endpoint
SKILLS & INTEGRATIONS

90+ skills.
Plus every MCP server.

Installable agent capabilities. Linear tickets, HubSpot deals, GitHub PRs, browser automation, code review — even your own private tools via the Model Context Protocol.

Browse all skills →
/lin
Linear
/hub
HubSpot
/not
Notion
/git
GitHub
/slk
Slack
/apo
Apollo
/brw
Browser
/cr
Code review
/str
Stripe
/sf
Salesforce
/gm
Gmail
/drv
Drive
/cal
Calendar
/fig
Figma
/jra
Jira
/cfl
Confluence
/zd
Zendesk
/inc
Intercom
/seg
Segment
/mix
Mixpanel
/snw
Snowflake
/bq
BigQuery
+
+ MCP
→
+ 70 more
MEMORY & KNOWLEDGE

Your fleet remembers everything.

Persistent agent memory across sessions. Unified knowledge base — search across crawled docs, Google Drive, Notion, Gmail, HubSpot, Linear. Read and write your local Obsidian vault. Memories scoped to projects, teams, or topics.

Per-agent memoryPer-org memoryDomain-scopedUnified searchObsidian read/write
Knowledge & Memory →
knowledge base · 4,827 entries
🔍"who's the champion at acme?"
HubSpotAcme Corp · Maya Lin · VP Eng
GmailRe: Q1 renewal · 14 messages w/ Maya
NotionAcme account plan · last edit 2d ago
MemoryMaya prefers Tuesday demos · 3pm PT
Obsidianvault/customers/acme.md · 12 backlinks
BY THE NUMBERS · Q1 2026
2.4M
agent runs per minute at peak — across every team shipping on Grain. From two-person startups to Fortune 50 support desks.
142ms
median latency
99.995%
uptime / 90d
$0.003
median / task
FIELD REPORT
“We scaled from 3 agents to 100 in a single quarter. Productivity per rep is up 4×, response times down 70%, and our SDR team now spends time closing — not researching. Grain is the only platform that let us run a fleet this big without losing the plot.”
JQ
Jesper Qvist
Founder, Salestools — 100 agents in production
PRICING

Start free. Scale when you ship.

HOBBY
$0for solo devs
  • ✓1,000 runs / mo
  • ✓3 agents pinned
  • ✓Community skills
Start free
TEAM
$49per seat / mo
  • ✓Unlimited runs
  • ✓All 11 agents
  • ✓All 16 model providers
  • ✓Eval suites + CI
  • ✓Approvals & audit log
Start 14-day trial →
ENTERPRISE
TalkSOC2 · HIPAA · SSO
  • ✓BYOC inference
  • ✓Private skill registry
  • ✓On-prem option
  • ✓Custom SLA
Talk to sales

Hire the team.

1,000 free runs a month. No credit card. Your fleet is 30 seconds away.

Download Grain — freeSee 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
Tweaks
APPEARANCE