GlossaryCompany Brain = governed contextShared4d ago
!
Needs review
Weekly pipeline review
Proposed by Sarah · 1h ago
RejectApprove
ChatGPTDrafts launch copy using approved contextSynced
ClaudeWrites GTM brief with company knowledgeSynced
CursorUpdates onboarding docs from latest decisionsSynced
CodexExecutes with trusted contextSynced
y.Your agentsEvery agent, one source of truthSynced
Approved
Becomes company truth
+24
Pricing decision
From GTM review
Proposed
New onboarding step
From CS handoff
Proposed
y.Yempik Company Brain4
Decision
Launch pricing at $59/month
Approved
Process
Client onboarding: 3-step flow
Active
ICP
Focus mid-market B2B SaaS
Approved
Needs review
Weekly pipeline review
RejectApprove
ChatGPT
Synced
Claude
Synced
Cursor
Synced
Codex
Synced
Human approvalAlways in control
Unlimited agentsFree to connect
Private & secureYour data stays yours
Compatibility
In the AI tools your team already uses
Yempik MCP connects to the AI clients your team opens every morning. No extra app to learn: company knowledge arrives right where you already work.
ChatGPT
Claude
Claude Code
Codex
Cursor
And any MCP client · OAuth 2.1 connection
The problem
You pay for four AIs. None of them knows your company.
Some brainstorm on ChatGPT, some build with Codex or Cursor, some prepare reports with Claude. Every chat creates value and the company sees almost none of it. And none of these AIs lifts a finger until someone asks.
Every chat is a silo
Conversations live in personal accounts. What Sara decided with ChatGPT, Marco will never see.
Scattered knowledge
Context re-explained every time
Price list, clients, rules: every new chat starts from zero, and everyone re-explains the company to their own AI.
Time lost every day
Answers that contradict each other
Each AI answers with whatever it knows: old versions of the price list, outdated decisions, invented sources.
Operational risk
The company never learns
Decisions and solutions stay in the chats of whoever wrote them. When a person leaves, that knowledge leaves too.
Continuity at risk
The product
What is the Company Brain
The Company Brain is your company’s operating knowledge in one place: decisions, processes, rules, clients, and glossary, versioned and always with the source shown. People consult it, the team’s AI tools work on top of it via Yempik MCP, and the system uses it to prepare your work: never to execute it on its own.
What makes it different
It’s yours: readable files, exportable whenever you want
Every answer cites the source and the current version
No AI writes on its own: it proposes, and you approve from the Inbox
brain.yempik.ai
The Inbox
The work arrives already done. You decide.
The Brain doesn’t wait for your questions: it crosses what happens (email, chat, calendar) with your company’s rules and puts ready-made proposals in your Inbox. Approve, edit, or reject; it learns from every choice you make.
Routines switch on by outcome, not by configuration
“Don’t let my follow-ups slip
“Prep me before every client call
“Turn chat decisions into Brain
“Watch the renewals coming due
Default: it proposes, you approve. No email goes out on its own, no truth writes itself; autonomy climbs step by step, routine by routine, only with the trust it earns.
09:02
Gmail · Speed-to-lead routine
A lead writes asking for a quote
In your Inbox: the reply draft, anchored to the active price list. Read it, approve it, it goes out.
ApproveReject
11:40
Claude · Sara’s chat
The chat settles the discount for public-sector clients
The AI proposes it to the Brain on its own. Once approved, it holds for every AI on the team.
14:30
Calendar · Call-prep routine
At 3:00 PM you have the Huck May Inc. call
In your Inbox: the brief with history, active decisions, open points. You walk in prepared.
Mon 08:00
Monitor · stalled follow-ups
A deal has been sitting still for over 48 hours
In your Inbox: the reminder with a follow-up draft, written in your voice.
Yempik MCP
What changes inside the team’s chats
Connect Yempik MCP once; from that moment, every AI on the team works from your company’s verified knowledge and contributes to the Brain as it works, without being asked.
Claude · team chatMCP · connected
Ok, closed with Bianchi: for public-sector clients the maximum discount is 12%, not 15.
RA
brain_propose_revision · 2026 Price List
Updating the Brain:
ProposalPublic-sector discount: 12%In Inbox
I proposed the revision to the price-list decision, without being asked. Once you approve it, it holds for every AI on the team.
Proposal in the Inbox · nothing becomes truth without approval
Proposes without being asked
When a decision, rule, or fact that matters emerges in chat, the AI proposes it to the Brain on its own. Capturing is the default behavior, not a command.
brain_propose_item
Searches and cites before answering
On price lists, clients, and procedures the AI queries the Brain and answers with the current decision, never a superseded version. Source cited, always.
brain_search · brain_get_current_decision
Starts routines, never silently
From chat, the AI can also propose routines: monitors, prep, follow-ups. Every proposal goes through the Inbox before it runs.
routine_propose
Governed for every person
OAuth connection, per-user permissions, an audit trail for every call. Everyone sees only what they’re allowed to see.
OAuth 2.1 · audit log
Connected in three steps, governed from the first
Each person connects their AI once. One URL, one login, and the chat works on your Company Brain.
Choose your AI
1
Use the Yempik MCP URL
Copy the endpoint and use it in any MCP client.
https://mcp.yempik.ai/mcp
2
Add the connector
In Claude, open Settings · Connectors, choose “Add connector”, and paste the MCP endpoint.
MCP endpoint
https://mcp.yempik.ai/mcp
Add connector
3
Sign in and choose your workspace
At login, pick the workspace. From then on the chat works only on that Brain.
Choose workspace
Acme Inc.
Globex Corporation
Initech
Try these right after connecting
What is the current decision on the public-sector price list?
Propose to the Brain: quotes must be signed within 5 days
Turn on a routine: prep me before every client call
How to start
A simple path, governed from day one
No migration, no new platform to learn: you start from the knowledge you already have and the AI tools you already use.
1
2
3
4
1
Knowledge Transfer
we seed the Brain with you: decisions, price lists, processes, and clients, from the knowledge you already have.
2
Connect the team
each person connects Yempik MCP to their own AI: ChatGPT, Claude, Codex, or Cursor.
3
Work as you always do
chats read the Brain, cite sources, and propose updates as you work.
4
Approve from the Inbox
every proposal passes through you: approve, edit, or reject it. Proposal first, truth second.
AI proposes. You approve. Always.
No chat writes directly to the Brain, and nothing happens on its own: every change starts as a proposal, approved by a person, with a full audit trail. The format is open: the Brain stays in your own files, readable even without us. cowork-os on GitHub ↗
The results
Why it works
One single truth for people and AI. Every answer gets better, every decision sticks.
Company Brain
Decision2026 Price ListActive
ProcessClient onboardingActive
ClientHuck May Inc.Updated
Versioned · sources cited
The team’s AI tools
They understand, answer, and propose.
Only verified sources
Always the current decision
Proposals, never direct writes
Results for your team
Precise answerswith a cited source
Zero contextto re-explain
Consistency across peopleand different AI tools
Decisions that stickand are easy to find
Accurate answers, with sources
The team’s AI tools answer only with verified company knowledge, and every answer shows where it comes from.
Context never needs re-explaining
Price list, clients, and rules are already in the Brain: every chat starts informed, on any tool.
You stay in control
Tracked proposals, clear permissions, human approval on everything. The team adopts it because it doesn’t change their habits.
Guarantees included
The Brain is yours, exportable
AI proposes, you approve
Data hosted in the EU
One brain for every AI on the team.
Start free. Connect agents whenever you want. No credit card required.
Prices in USD. You pay for the brain, not seats. Export on every plan. Design Partner by application.
FAQ
Frequently asked questions
It’s your company’s operating knowledge in one place: decisions, processes, rules, and clients, versioned and with the source shown. People consult it; the team’s AI tools use it via Yempik MCP to answer and propose updates. Unlike a wiki, it stays alive, because it updates from everyday work.
That memory is personal and closed: it lives in one person’s account, inside a single tool, and you don’t govern it. The Company Brain belongs to the company: shared across people and across different AI tools, with versions, sources, and human approval. And if you switch tools down the road, the Brain stays: your own files, exportable.
A wiki is written for humans who read pages. Agents don’t read pages: they call tools. The Brain serves both: people consult it, agents query it via MCP and get only approved, current truth with its source.
RAG retrieves the most similar text; it doesn’t know whether it’s approved, current, or permitted for whoever is asking. The Brain governs truth before retrieval: every change starts as a proposal, gets ratified and versioned. RAG gives you the most similar answer; the Brain gives you the correct one.
Agent 365 tells you which agents are running and whether they are safe. It does not tell you whether what they read is true. We sit on the knowledge side of that line: one approved source of company context, with a human approving what becomes truth, exportable and not tied to one vendor stack.
Those observe the traffic: what users ask and where agents fail. The Brain governs what the agent reads and answers. Complementary layers: one watches the sessions, the other guarantees the truth.
No. Yempik MCP connects to the AI clients you already use: ChatGPT, Claude, Codex, Cursor, and any MCP client. The team keeps working exactly where it works today.
No. Routines switch on by choosing outcomes in plain language: “don’t let my follow-ups slip”, “prep me before every call”. The system translates, watches how you work, and adapts; every correction you make refines it.
It can’t write to the Brain: it can only propose. Every proposal lands in the Inbox and becomes truth only if a person approves it. Answers cite the source, so you can verify it in one click.
You do. The Brain lives in readable files, exportable at any time, hosted in the EU. The format is open: it stays usable even without Yempik.
You can start free as a solo professional (Free: 1 member, no usage limits). Business use only — not for personal or household use. Team is $59/month ($39/month billed annually) per workspace up to 5 members, with automations and a shared workspace. Scale is $179/month up to 20, with internal agents, routines, and advanced automations. On every plan you can export the Brain anytime. You pay for the brain, not seats — connecting agents is free. For co-design, there’s Founding Design Partner by application.
