Knowledge is everywhere and nowhere
Meeting notes in Fireflies, decisions in Slack, specs in Confluence, customer feedback in CRM, architecture in GitHub. Your AI agents can’t find what matters.
Company Brain HarnessBook a callGoverned memory for AI agents and teams
Your company’s knowledge.
Structured. Governed. AI‑ready.
Company Brain Harness turns scattered docs, meetings, Slack threads, CRM records, and tribal knowledge into a source-governed memory that Claude, Codex, and your entire team can trust.
Claude Code & Codex plugin Source-level governance Human-gated review Preview-first writes
The problem
AI agents are only as good as the knowledge they can access
Most companies have knowledge scattered across a dozen tools with no governance, no review process, and no way for AI to find what matters.
AI agents read what they shouldn’t
Without governance, AI tools ingest private conversations, stale documents, and sensitive data. One bad source turns shared context into a liability.
No single source of truth
Every team member has a different version of reality. AI agents hallucinate company facts because there’s no canonical, reviewed knowledge base to draw from.
How it works
From scattered knowledge to governed company brain in five steps
The plugin installs the workflows and deterministic scripts. The brain root stores your knowledge and system state. Source systems feed evidence only after policy, scope, and reviewers are explicit.
1
Install the plugin bundle
One install brings skills, helper scripts, and harness references into Claude Code or Codex.
2
Choose your brain root
A Google Drive folder, git repo, shared volume, or local directory becomes your company brain.
3
Scaffold with preview
The harness shows you the full brain structure before writing a single file. Nothing happens without your approval.
4
Register source instances
Connect specific workspaces, channels, CRM accounts, and meeting recorders. Each source is scoped and approved individually.
5
Review, approve, promote
Raw evidence enters private staging. Curated proposals go through review. Only approved knowledge reaches the shared brain.
Who it’s for
Built for teams at every stage
Whether you’re a 5-person startup or a 500-person enterprise, governed company memory solves the same fundamental problem at different scales.
Startups & small teams
5–25 peopleCritical knowledge lives in founders’ heads and Slack DMs. When someone leaves, weeks of context go with them. Company Brain Harness captures tribal knowledge into a structured brain so every AI tool and every new hire gets real company context from day one.
- Capture what your team knows before it walks out the door
- Give Claude and Codex real product and strategy context
- Onboard new hires with governed institutional knowledge
Growing companies
25–200 peopleAs teams multiply, knowledge silos grow faster than headcount. Engineering knows things sales doesn’t. Customer success hears feedback product never sees. Company Brain Harness connects these sources through governed capture.
- Break down knowledge silos across departments
- Route customer insights to product decisions automatically
- Scale AI adoption without losing control of company data
Enterprise & regulated industries
200+ peopleEnterprises need audit trails, data classification, and compliance controls before AI agents can touch company data. Company Brain Harness provides source-level governance so AI adoption doesn’t become a compliance risk.
- Source-level governance with complete audit trails
- Restricted routing for HR, legal, and financial data
- Approval ledger satisfies compliance requirements
Setup
One install. Full operating system.
The plugin bundle contains everything: skills for routing workflows, deterministic scripts for repeatable operations, and reference docs for the harness contract. Your brain root is separate customer data.
skills/ — routing and workflowsbin/ — deterministic helper scriptsreferences/ — harness contract
Claude CodeCodex
# Install the full plugin bundle
claude plugin install company-brain-harness
# Open your brain root folder and say:
# "I want to set up my company brain."
# The harness scaffolds in preview mode first.
# Nothing writes until you approve.Plain-language roles
Say what you want. The agent does the rest.
No one needs to learn scripts or folder structures. Users state their intent in natural language and the harness routes them to the right workflow.
Brain Owner
Sets policy, approves sources, manages restricted access, decides automation levels, and owns pilot readiness.
“Set up my company brain”“Approve these sources”“Review restricted access”
Brain Operator
Runs daily health checks, reviews staging queues, handles source syncs, and keeps brain notes accurate and fresh.
“Run today’s brain check”“What needs review?”“Show health score”
Team Member
Contributes domain knowledge, suggests sources, reports stale facts, and delegates scoped access in plain language.
“Add what I know about our API”“This fact is outdated”“Connect my work calendar”
Source governance
Every source is scoped, approved, and auditable
Slack, Confluence, Gmail, CRM, Fireflies, GitHub, and calendars can all be valuable. But each workspace, channel, inbox, and account must be individually registered and approved before capture begins.
Allowed after approval
Company workspaces, approved channels, shared CRM accounts, company GitHub orgs, meeting recorders for team calls.
Excluded by default
Personal inboxes, private meeting notes, private CRM data, user spaces, broad transcript dumps, secrets and credentials.
Each workspace, channel, inbox, calendar, repo org, or API key is a separate source instance. Connector type alone is not enough.
Personal accounts are excluded by default unless explicitly delegated, scoped, approved, and registered.
Raw transcripts, emails, and exports are private evidence, never shared knowledge.
HR, legal, finance, and sensitive customer material routes to restricted review or stays excluded entirely.
Repeated syncs use cursors and hashes to find changes without duplicating notes.
Brain structure
Clean taxonomy, ready for any company
Domain-first, lowercase, readable. No numbered folders or company-specific assumptions. Another organization can adopt the brain structure without inheriting someone else’s naming conventions.
brain/companyCompany context, operating model, principles, and history brain/productProduct roadmap, decisions, launch readiness, and active work brain/engineeringArchitecture, repo maps, stack guardrails, and implementation routes brain/go-to-marketPositioning, pricing, sales assets, and growth experiments brain/customersOnboarding, support, success, feedback, and account context brain/operationsTeam profiles, workflows, daily loops, and department context brain/intelligenceMeeting notes, market signals, customer voice, and decision history brain/sourcesSource-derived notes and reference maps system/Configuration, source registry, sync state, staging, evidence, and policyOperating rhythm
Start human-gated. Automate what earns trust.
The first two weeks establish trust: run checks, stage proposals, review them, promote only what has a source and reviewer. Once the queue is trusted, approved sources can stage unattended while final promotion stays governed.
Daily
Connection health, source registry check, staged queue review, brain lint, and next actions.
Weekly
Stale note audit, source ownership review, routing cleanup, and team adoption check.
Monthly
Permission audit, retention policies, automation rule review, and source approval cycle.
Ready to build your company brain?
Book a setup call to map your company’s operating context, choose a brain root, and get the harness installed with governed source capture running in your first session.
Book your setup callFree setup consultation. No commitment required.
FAQ
Common questions
Is this a personal second brain app?
No. Company Brain Harness uses second-brain principles but the product boundary is team-owned company memory with source governance, review queues, and role-based access.
How is this different from a RAG chatbot?
RAG tools answer questions by searching documents. Company Brain Harness builds the governed knowledge layer upstream. It handles source registration, evidence capture, staging, review, and promotion so the knowledge AI agents read from is accurate, approved, and current.
What do customers actually install?
A plugin bundle that includes skills (routing and workflows), bin (deterministic helper scripts), and references (harness contract). The brain root is separate customer data.
Does it populate company knowledge automatically?
Setup creates the structure. Knowledge comes from team contribution and approved source syncs, always through staging and human review before it reaches the shared brain.
Can source syncs run unattended?
Health checks and approved-source staging can be scheduled. Final promotion to the shared brain stays human-approved unless a narrow, low-risk automation rule is explicitly approved by the Brain Owner.
Where does raw source data live?
Under system/staging/raw as private evidence. It never appears in the shared brain/ folders. AI agents are instructed to treat raw evidence as private, not shared knowledge.
Which AI tools does it work with?
Claude Code and Codex through their plugin systems. The brain root is also readable by any AI agent that can access the filesystem or shared drive.