Your clients' knowledge, finally in one brain
Prices live in spreadsheets, policies in PDFs, promos in someone's head, customer history in chat threads. GraphOS turns all of it into a knowledge graph — and every agent you sell answers from it, correctly, on every conversation.
No credit card required · Your trial runs on our AI keys — nothing to configure
The Explorer — a real client graph. Every dot is an entity, every line a fact with validity in time. The slider shows the graph as it was known on any date.
How it works
1 · Connect the sources
Upload PDFs, Word, Excel. Connect REST APIs and SQL databases. Pull from 800+ apps via Composio. Declare facts manually — with start and end dates for promos.
2 · The graph builds itself
AI extracts the entities (people, products, prices) and the facts connecting them. When information changes, old facts are superseded — never confused with current ones, never lost.
3 · Agents read — and write
Every agent with the Knowledge Graph enabled queries it automatically mid-conversation. Enable agent memory and they also record what they learn: scores, outcomes, preferences.
Why a knowledge graph beats plain document search
Answers across documents
"Which services changed price this year?" spans your price list, your announcements and your chat history. A graph connects them; a document search can't.
Time-aware by design
A promo that expired last month is history, not an answer. Facts carry validity windows, and agents always use the current truth.
Shared memory for the whole squad
Every agent on the account reads the same graph. What the sales agent learns, the support agent knows.
Built for the way you sell
- One isolated graph per client — their data never mixes
- White-labeled like everything else on the platform
- Agents that remember their clients' business = retainers that renew themselves
- Full dev access: REST API, TypeScript SDK, CLI and MCP
Costs your clients can see and cap
Your trial includes 50 knowledge episodes on our AI keys — no setup needed. After that, extraction runs on each account's own OpenAI key (BYOK). The Usage panel shows episodes, queries and the estimated monthly LLM spend — and a monthly cap stops ingestion before it overspends. No surprises, no black box.
Everything the UI does, the dev tools do too:
m2f knowledge sources add "Client docs" --type file m2f knowledge upload src_abc ./price-list.pdf m2f knowledge query "What does the cleaning cost?" m2f knowledge usage --set-limit 500
Simple pricing, no surprises
Pay for the platform, resell at your own prices.
Starter
2,400 credits per month
- Full white-label
- Unlimited sub-accounts
- Your prices, your margin
- $0.03 per extra credit
Agency
12,000 credits per month
- Full white-label
- Unlimited sub-accounts
- Your prices, your margin
- $0.02 per extra credit
Business
50,000 credits per month
- Everything in Agency
- The cheapest credit on the ladder
- $0.015 per extra credit
- Priority support
Calculate your margin
How many clients you need and what you keep on each plan.
Calculate your margin
How many clients you need and what you keep on each plan.
Who you bill. Each one has their own agent and their own users.
The distinct people who write to that client's agent in a day.
Each reply costs 1 credit; what the users write costs nothing. A typical request is resolved in 5 to 10 replies.
Set for you to leave a 80% margin. Move it if you already know what you will charge.
Advanced settings
Credits per month
12,000
One agent reply = 1 credit. What the users write costs nothing.
Recommended plan
Agency
$240.00/mo
12,000 / 12,000 included
- Your monthly cost
- $240.00
- Your monthly revenue
- $1,250.00
- Cost per client
- $48.00
Tokens from your LLM provider (OpenAI, Anthropic or another) are billed separately, straight to your key. This calculator estimates platform cost only.
Give your agents a memory
Create your account, upload a document, and ask your first agent something it could never have guessed.
Your free trial includes — no credit card required
We build your agent from your use case
Tell us what you want to automate. We read it back, you confirm it, and your agent is running with its own knowledge brain.