Memory
One memory, every agent
Give your whole fleet a shared source of truth. Nothing gets forgotten, duplicated, or contradicted between tools — every agent works from the same up-to-date picture.
GRIFFai Agent OS
AI agents that remember, act, and answer for it.
One shared memory, built-in guardrails, and a clear record for every agent on your team — so you can move fast with AI without losing control of it.
Shared agent memoryGuardrails before risky movesA record for every action
Works with the tools your team already uses
0+
tools your agents can use, out of the box
0
built-in guardrails, active from day one
0
ways to run it — cloud, self-hosted, or local
0%
of agent activity logged and reviewable
The platform
Everything your agents need to remember, act, and answer for it.
GRIFFai Agent OS is the layer underneath your agents: shared memory, task routing, a live control center, and a connector every major AI already speaks — one platform across desktop, self-hosted, and cloud.
Orchestration
The right work, the right agent
Route tasks automatically across your fleet. People step in only where it counts — sensitive moves pause for a named approver instead of running unattended.
Operations
See everything, in one view
A live control center shows what your agents are doing right now, what needs a decision, and what already happened — no digging through logs.
Connectivity
Works with the AI you already use
Claude, ChatGPT, Gemini, Grok, and anything else that speaks the open MCP standard connect in minutes — no custom integration project.
Deployment
Run it your way
Cloud, self-hosted, or fully on your own hardware — same platform, same experience, whatever trust boundary your business needs.
Accountability
A clear record, every time
Know what an agent did, why, and who signed off — in plain language your team, your customers, and your auditors can all read.
The full picture
One platform, every capability your agents need.
Explore what’s built in, grouped by what it does for you — from shared memory and security to billing and delivery.
Memory & Brain
7 capabilities
6 live1 roadmap
- Brain Litelive
- Data Lakehouselive
- VPC Bridgelive
- Brain Ingestlive
- Hybrid VPC Bridgelive
- Brain Lite (Vectorize)live
+1 more in the full catalog below.
Hover or tap any node to see what it unlocks for your team. Browse the full list below.
Live platform pulseDirect browser probe
Brain Lite
Checking…Durable Queue
Checking…Secrets Vault
Checking…Voice Brain
Checking…Browser Reports
Checking…Compliance
Checking…A live signal from the infrastructure behind your workspace, checked right now from your browser. Full detail on system status.
21
live Cloudflare Workers
15
live add-on capabilities
5
platform categories
6
labeled roadmap, not oversold
Billing & Ops
LiveSignup
Self-serve signup with Stripe-backed plan selection.
Cloudflare Worker
Billing & Ops
LiveProvisioner
Automated workspace provisioning — databases, configuration, and connectivity stood up the moment you sign up.
Cloudflare Worker
Billing & Ops
LiveLicensing
Ed25519-signed license issuance for self-host and desktop installs — offline-tolerant.
Cloudflare Worker
Agent Runtime
LiveMCP Gateway
The governed MCP Streamable HTTP surface every connected agent talks to.
Cloudflare Worker
Agent Runtime
LiveSwarm Hub
Durable Object hub coordinating multi-node agent swarms and failover.
Cloudflare Worker
Billing & Ops
LiveAccount
Account identity, JWKS verification, and session issuance.
Cloudflare Worker
Billing & Ops
LiveMetering
Analytics Engine usage capture rolled up to Stripe Meters on a cron.
Cloudflare Worker
Memory & Brain
LiveBrain Lite
Vectorize + Workers AI embeddings — a managed semantic-recall tier that scales to zero.
Cloudflare Worker
Memory & Brain
LiveData Lakehouse
Pipelines to R2 (Apache Iceberg) for queryable interaction-stream history at near-zero egress.
Cloudflare Worker
Agent Runtime
LiveSwarm Node
A dedicated per-workspace Brain node with alarm-driven heartbeat and clean join/leave.
Cloudflare Worker
Memory & Brain
LiveVPC Bridge
Workers VPC + Cloudflare Tunnel bridge from cloud agents to a private on-prem Brain — no public port.
Cloudflare Worker
Billing & Ops
LiveEmail Service
Transactional email for magic-links, invites, and 2FA on Cloudflare Email Service.
Cloudflare Worker
Security & Compliance
LiveSecrets Vault
Account-level Secrets Store binding — central rotation, one audit surface, scoped bindings.
Cloudflare Worker
Security & Compliance
LiveCompliance
20-control NIST 800-53 Rev.5 catalog with two-person freeze/approve enforcement.
Cloudflare Worker
Edge & Delivery
LiveDurable Queue
Cloudflare Queues + DLQ behind Stripe webhooks and Brain ingest bursts — fast-ack, durable retry.
Cloudflare Worker
Edge & Delivery
LiveEdge Security
Durable Object rate limiting engine for per-workspace abuse control at the edge.
Cloudflare Worker
Agent Runtime
LiveCode Sandbox
Worker Loaders isolate per agent-authored run — outbound network blocked by the runtime itself.
Cloudflare Worker
Agent Runtime
LiveBrain Container
Dedicated compute for workspaces that need reserved capacity.
Cloudflare Worker
Edge & Delivery
LiveCustom Domains
Cloudflare for SaaS custom-hostname provisioning against the Workers Custom Domains API.
Cloudflare Worker
Agent Runtime
LiveBrowser Reports
Browser Rendering generates PDF/PNG proof packs — dashboard snapshots and audit receipts on demand.
Cloudflare Worker
Agent Runtime
LiveVoice Brain
Whisper transcription in, LLM answer, natural TTS out — one gated Workers AI round trip.
Cloudflare Worker
Memory & Brain
LiveBrain Ingest
Queue new memory sources for ingestion without upgrading your plan.
Add-on
Security & Compliance
LiveGovernance Manifest
Enterprise governance manifest tooling available on a non-enterprise plan.
Add-on
Billing & Ops
LiveExtra Seats
Additional named seats beyond the plan’s included count, in packs of five.
Add-on
Billing & Ops
LivePriority Support
Priority queue on support requests with a 4-hour response SLA.
Add-on
Agent Runtime
LiveDedicated Swarm Node
A cloud Brain node reserved for your workspace instead of the shared pool.
Add-on
Security & Compliance
LiveFederal Compliance Pack
FedRAMP-aligned controls, ISSM role, and freeze/approve audit workflow.
Add-on
Billing & Ops
LiveAI Gateway Budget Control
Every model call routed through one AI Gateway with per-workspace attribution, caching, and metered billing.
Add-on
Security & Compliance
LiveManaged Secrets Vault
Account-level Secrets Store for every per-worker credential — central rotation, one audit surface.
Add-on
Edge & Delivery
LiveDurable Ingest Queue
Durable retry behind the Stripe webhook and Brain ingest bursts — no webhook-timeout risk.
Add-on
Memory & Brain
LiveHybrid VPC Bridge
The supported way for a cloud agent to reach your private on-prem Brain over a Cloudflare Tunnel.
Add-on
Memory & Brain
LiveBrain Lite (Vectorize)
Managed semantic recall on Vectorize + Workers AI embeddings — scales to zero, no graph to run.
Add-on
Billing & Ops
LiveMetered Recall Overage
Pay-as-you-go recall, ingest, and agent calls beyond your plan’s included pool.
Add-on
Agent Runtime
LiveCode Mode Agent Sandbox
Agent-authored JS in a fresh per-request Worker Loaders isolate — outbound network blocked.
Add-on
Agent Runtime
LiveVoice Brain
Talk to your Brain: whisper transcription, LLM answer, natural TTS — one gated round trip.
Add-on
Agent Runtime
LiveBrowser Rendering Reports
On-demand PDF/PNG proof packs — dashboard snapshots, audit receipts, SBOM reports.
Add-on
Security & Compliance
RoadmapSecure Onboarding Pack
Turnstile bot/abuse defense on signup and reset, plus email-based magic-links and 2FA.
Add-on
Agent Runtime
RoadmapManaged OAuth + MCP Portal
Your workspace becomes a full sign-in provider for AI tools — bring any agent, on your terms.
Add-on
Memory & Brain
RoadmapR2 Data Lakehouse
Opt-in activity history streamed to low-cost storage, queryable on demand at near-zero cost.
Add-on
Edge & Delivery
RoadmapEdge Security Pack
Managed WAF rules, per-route rate limiting, and Bot Management across the whole zone.
Add-on
Edge & Delivery
RoadmapCustom Brain Domain
Front your Brain on your own domain with Cloudflare for SaaS auto-provisioned certs.
Add-on
Agent Runtime
RoadmapDedicated Brain Container
Reserved compute for your workspace’s Brain Container instead of the shared pool.
Add-on
How it works
From a request to a result you can trust.
1
Connect your agents
Point Claude, ChatGPT, or any compatible AI at your workspace. Desktop, self-hosted, and cloud all speak the same language.
2
They share real memory
Every agent reads from and writes to the same trusted memory — checked for accuracy before anything acts on it.
3
Risk pauses for a person
Everyday work keeps moving. Spend, external messages, secrets, and production changes wait for a named approver.
4
Every action is logged
Each task closes with a clear, exportable record: what happened, who approved it, and how to verify it later.
Run it your way
Cloud, self-hosted, or local. You choose the boundary.
Same platform, same experience — pick the deployment that matches how your business already works.
GRIFFai Cloud
Fastest startSign up and connect an agent in minutes — isolated infrastructure for your workspace, simple self-serve billing, no setup required.
Start with CloudSelf-hosted
Your infrastructureOne hardened image running on your own servers — full control, full ownership, the same experience your cloud users get.
Talk to us about self-hostingDesktop
Local-firstNative apps for Windows, macOS, and Linux. Your agents, your memory, your machine — works offline, no data leaves your device.
Join the desktop waitlistQuestions
Straight answers.
What is GRIFFai Agent OS?
A shared memory, an activity control center, and a safety layer for every AI agent your business runs — deployable in the cloud, self-hosted, or on the desktop.
Does it replace my agents or models?
No — it sits underneath them. Claude, ChatGPT, Gemini, Grok, and any compatible AI connect to the same shared memory and toolset you already control.
What makes the memory trustworthy?
Every read is checked for accuracy and freshness before an agent acts on it, and leaves a record behind. Your agents act on memory you can verify, not on whatever surfaced first.
Can I run it entirely on my own hardware?
Yes. The self-hosted and desktop options run the full platform locally with offline-tolerant licensing — the cloud option is there when you want it, not required.
Give your agents an operating system.
Self-serve signup and a shared memory your whole team can build on. Start on Personal or Starter and scale to Enterprise when you’re ready.