Deploys go out through the release dashboard on Fridays.
One governed brain for
every AI you trust.
GraphStaff is a single permission-enforcing knowledge layer for your company. Claude, ChatGPT, coding agents, and your own AI stack draw on one governed brain instead of keeping their own stale copies of company context. Your team asks it too, and every answer shows its sources.
The direct answer
What is GraphStaff?
GraphStaff is the company brain: a governed knowledge graph of your company knowledge with permission-aware retrieval. People and AI agents ask through search, Ask AI, and MCP. The brain draws on one governed source, applies the asker's permissions, and shows the sources behind every answer.
Everyone asks
All day, your team and their AI are asking.
Different roles, different context. Each one needs the answer that fits.
A new engineer · Ask AI“How do we deploy to production?”
A coding agent · MCP“What is our error-handling convention?”
An account manager · Search“What did we promise this customer?”
An executive · Ask AI“Where are we exposed on compliance?”
A new engineer · Ask AI“How do we deploy to production?”
A coding agent · MCP“What is our error-handling convention?”
An account manager · Search“What did we promise this customer?”
An executive · Ask AI“Where are we exposed on compliance?”
A new hire · Ask AI“How does PTO carry over?”
An AI assistant · MCP“What changed in the March release?”
An on-call engineer · Runbook“How do I open a production incident?”
A support lead · Search“Which customers are on the legacy plan?”
A new hire · Ask AI“How does PTO carry over?”
An AI assistant · MCP“What changed in the March release?”
An on-call engineer · Runbook“How do I open a production incident?”
A support lead · Search“Which customers are on the legacy plan?”
Production deploys run from the CLI, any weekday.
Pretty sure we froze Friday deploys after the incident?
unverifiedwhich one of these is actually current?
The deploy script points at a pipeline that no longer exists.
They all ask the same question. Today they get five conflicting versions.
GraphStaff connects your company knowledge into one governed brain. Every answer comes from it, shaped to fit who is asking.
PEOPLETeams · new hires · leaders
Governed brainOne graph · one policyAI AGENTSClaude · ChatGPT · your stack
THREE SURFACES · ONE POLICY
SearchAsk AIMCP
Answers that fit the asker
One question. The depth each asker needs.
Everyone reaches the same governed brain. What comes back matches the asker: the depth their role calls for, within the permissions they hold, with the sources shown.
“What is our data retention policy?”
An engineer · depth
The full policy Retention windows for each data class, the service that enforces them, and the exceptions that apply.
A new hire · orientation
The plain version A short, plain-language summary and a pointer to the team that owns the policy.
An executive · summary
The one-liner The headline answer and where the company still carries risk, with nothing to wade through.
An AI agent · MCP
Grounded context Permission-filtered sources with citations, scoped to what the requester is allowed to see.
AM
Alex MorganProduct · customers group
GraphStaff policy Allowed contextVISIBLE TO ALEX AND THEIR AI AGENTProduct roadmap · Customer guidance
NOT RETRIEVEDFinance runbook · Executive notes
Questions, answered
Frequently asked questions.
The short version of what teams usually want to know before they see GraphStaff in action.
How is the company brain different from a wiki or knowledge base?
A wiki stores and publishes pages. The company brain connects pages, concepts, and policies into one governed knowledge graph, then serves grounded, permission-aware answers through search, Ask AI, and MCP. The knowledge is maintained once and answers stay traceable to their source.
Does everyone get an identical answer?
No, and that is the point. Everyone asks one governed brain, but the answer is scoped to the asker. An engineer gets depth, a new hire gets orientation, an executive gets the summary, and an AI agent gets grounded, permission-filtered context. Every answer shows its sources, so the differences trace back to permissions and scope.
Does it work with the AI tools we already use?
Yes. One governed MCP gateway serves Claude, ChatGPT, coding agents, and your own AI infrastructure the same permission-aware context. You can swap models without rebuilding your knowledge layer.
How do we see whether it fits?
Request a demonstration and tell us which knowledge set matters most. We can plan a focused walkthrough or a proof of concept built on the questions your team actually gets asked.
Your knowledge is readyGive your company
Give your company
one brain.
Bring us the questions your team gets asked most. We will show you the answer each person and their AI gets back, and where it came from.