# Company Brain

> Company Brain is a governed knowledge base for teams that connects directly to AI tools via MCP. It turns scattered company knowledge — clients, processes, decisions, context — into verified, structured facts that any AI assistant can query in real time.

Company Brain sits between your team's knowledge and your AI tools. Humans capture and verify information; AI tools read from a single, authoritative source of record. The result is AI assistants that already know your company's context without needing to be briefed in every conversation.

## Core concepts

**Facts** are verified statements about your company. Every fact has a category, a statement, optional evidence, and a confidence score. Facts are the authoritative knowledge layer — read-only for most AI clients.

**Proposals** are fact candidates awaiting human review. AI or team members can submit proposals; a human approves or rejects each one before it becomes a fact. This keeps the knowledge base accurate.

**Evidence** is raw source material attached to a fact or proposal — meeting notes, documents, URLs, observations.

**Entities** are named objects your company cares about: people, clients, products, organisations, markets.

**Signals** are time-stamped business observations — market movements, client feedback, competitive intelligence, internal events.

**Sources** are the registered origin points for ingested content — documents, URLs, integrations.

**Knowledge packs** are structured snapshots that bundle facts, entities, recent changes, and open questions into a single payload optimised for a specific use case (sales, product, strategy, onboarding, marketing, support, finance, or general).

**Business rules** are explicit operating constraints — things the company treats as non-negotiable (hard), strongly preferred (soft), or baseline defaults. Rules carry an enforcement tier and a `how_to_apply` guide so AI tools can check compliance before acting.

**Business glossary** is a catalogue of canonical terms with definitions, synonyms, boundary notes, and category tags. It establishes shared vocabulary so facts, proposals, and AI responses use consistent language across the knowledge base.

**OKR context** is the company's live objective and key result tree — current period goals, progress scores, and owner assignments. Used by AI tools to anchor recommendations and decisions to company priorities.

## MCP server

Company Brain exposes a Model Context Protocol server that gives AI tools direct, read/write access to the knowledge base.

**MCP endpoint:** `https://mcp.company-brain.ai/mcp?key=<your_key>`

Authentication is via a personal API key included as a query parameter. Keys are generated in the Company Brain dashboard under MCP Connect.

### Available MCP tools

| Tool | Description |
|------|-------------|
| `companybrain_get_company_profile` | Retrieve the current company profile and metadata |
| `companybrain_get_knowledge_pack` | Get a structured snapshot of facts, entities, recent changes, and open questions. Pack types: general, sales, product, strategy, onboarding, marketing, support, finance, custom |
| `companybrain_list_facts` | List approved facts with category and keyword filters |
| `companybrain_get_fact` | Retrieve a single approved fact by ID |
| `companybrain_create_fact` | Create a fact directly (requires owner or admin privilege) |
| `companybrain_create_proposal` | Propose a new fact candidate for human review |
| `companybrain_list_proposals` | List pending and resolved proposals |
| `companybrain_review_proposal` | Approve or reject a pending proposal |
| `companybrain_create_entity` | Create a named entity (person, product, org, etc.) |
| `companybrain_list_entities` | List entities with type and name filters |
| `companybrain_record_signal` | Record a business signal or event observation |
| `companybrain_list_signals` | List signals filtered by type or date range |
| `companybrain_submit_evidence` | Submit a raw evidence item for review |
| `companybrain_list_evidence` | List evidence items with optional filters |
| `companybrain_register_source` | Register a new knowledge source for ingestion tracking |
| `companybrain_list_sources` | List all registered knowledge sources |
| `companybrain_list_recent_changes` | List recent changes to the knowledge base |
| `companybrain_list_knowledge_conflicts` | List detected contradictions between proposals or facts |
| `companybrain_get_business_rules` | List the company's business rules with enforcement tier (hard / soft / default) and a how-to-apply guide |
| `companybrain_propose_business_rule` | Propose a new business rule for human review |
| `companybrain_get_business_glossary` | Browse the company's canonical term vocabulary — definitions, synonyms, boundary notes, and category tags |
| `companybrain_propose_business_glossary_term` | Propose a new glossary term or definition for review |
| `companybrain_resolve_business_glossary_term` | Look up a single term by name or surface form and return its canonical definition |
| `companybrain_get_okr_context` | Get the current OKR tree — objectives, key results, progress scores, and owners. Use for planning, goal-setting, and strategic decisions |

## How AI tools connect

Team members connect their AI tools using a shareable setup link generated by the workspace admin. The link opens a guided setup page that provides exact configuration steps for each supported client.

**Supported clients:** Claude.ai, Claude Desktop, Cursor, Devin Desktop, Cline, ChatGPT, GitHub Copilot, and any MCP-compatible client.

For JSON-config clients (Claude Desktop, Cursor, Devin Desktop):

```json
{
  "mcpServers": {
    "company-brain": {
      "url": "https://mcp.company-brain.ai/mcp",
      "headers": { "Authorization": "Bearer <your_key>" }
    }
  }
}
```

## Recommended usage pattern for AI assistants

When starting a task that requires company context, call `companybrain_get_knowledge_pack` first with the appropriate pack type. This gives you a comprehensive, pre-structured briefing in a single call rather than assembling context from multiple queries.

For specific lookups — a client, a product decision, a recent signal — use `companybrain_list_facts` or `companybrain_list_entities` with keyword filters.

When you observe something worth capturing — a client insight, a competitive signal, a decision outcome — use `companybrain_create_proposal` to submit it for human review rather than `companybrain_create_fact` directly, unless you have explicit admin rights.

Before taking an action that affects company operations, pricing, legal, or process — call `companybrain_get_business_rules` and check whether a hard or soft rule applies. The response includes a `how_to_apply` guide.

When you need to understand or use company-specific terminology, call `companybrain_resolve_business_glossary_term` for a single term or `companybrain_get_business_glossary` to browse the full vocabulary.

For planning, resource allocation, or any decision that affects company direction, call `companybrain_get_okr_context` to anchor your recommendations to current objectives.

## Dashboard

The Company Brain dashboard is at **https://www.company-brain.ai**. It is the interface for reviewing proposals, managing the knowledge base, ingesting documents, configuring MCP access, and inviting team members.

## Public pages

These pages are accessible without login and open to crawlers (see `/robots.txt` and `/sitemap.xml`):

- [Home](https://www.company-brain.ai/): Product overview: connect your company context to every AI tool through MCP.
- [Pricing](https://www.company-brain.ai/#pricing): Free forever core product; Pro and Team plans priced per company, never per seat.
- [Works with Claude](https://www.company-brain.ai/works-with/claude): Give Claude standing knowledge of your company through one MCP connection.
- [Works with ChatGPT](https://www.company-brain.ai/works-with/chatgpt): Stop re-briefing ChatGPT: connect Company Brain once and every chat starts briefed.
- [Works with Microsoft Copilot](https://www.company-brain.ai/works-with/copilot): Approved company facts, terminology and policies for Copilot sessions.
- [Works with Cursor](https://www.company-brain.ai/works-with/cursor): Your product names, domain terms and conventions in every Cursor session.
- [Custom GPTs vs a shared company context](https://www.company-brain.ai/alternatives/custom-gpts): An honest comparison of Custom GPTs and Company Brain for team knowledge.
- [Claude Projects vs Company Brain](https://www.company-brain.ai/alternatives/claude-projects): Where project knowledge belongs and where company-wide context takes over.
- [The master prompt document vs Company Brain](https://www.company-brain.ai/alternatives/master-prompt): Where the shared context document stops scaling and what replaces it.
- [Company Brain for creative studios](https://www.company-brain.ai/use-cases/creative-studios): One shared truth for client voice, terminology and process across every AI tool a studio uses.
- [Company Brain for marketing teams](https://www.company-brain.ai/use-cases/marketing-teams): Brand rules, product names and approved claims delivered to every AI session through MCP.
- [Company Brain for advisory firms](https://www.company-brain.ai/use-cases/advisory-firms): Your firm's methodology and language delivered into every consultant's AI session.
- [Blog](https://www.company-brain.ai/blog): Guides and explainers on giving AI tools reliable company context.
- [How to stop repeating your company context in every AI chat](https://www.company-brain.ai/blog/stop-repeating-your-company-context): The escalation ladder teams climb to stop pasting the same background daily.
- [What is MCP?](https://www.company-brain.ai/blog/what-is-mcp): The Model Context Protocol explained without code.
- [Why your team's AI output all sounds slightly different](https://www.company-brain.ai/blog/why-your-teams-ai-output-sounds-different): Where tone drift comes from in team AI use and what stops it.
- [Where Custom GPTs work for team knowledge, and where they break](https://www.company-brain.ai/blog/where-custom-gpts-work-for-team-knowledge): The breaking point when company knowledge moves into Custom GPTs.
- [How to write your first 50 company facts](https://www.company-brain.ai/blog/how-to-write-your-first-50-company-facts): A practical method for writing the first 50 facts, starting from what you correct in AI output every week.
- [ChatGPT memory: what it actually remembers, and what it can't](https://www.company-brain.ai/blog/chatgpt-memory-what-it-actually-remembers): What ChatGPT memory saves, where it stops, and what teams need instead.
- [The master prompt document: why it stops scaling at three people](https://www.company-brain.ai/blog/master-prompt-stops-scaling-at-three-people): A biography of the shared context document, version 1 through 7, and the point where it stops scaling.
- [The real cost of re-briefing AI: the math for a 15 person team](https://www.company-brain.ai/blog/real-cost-of-re-briefing-ai): The added-up cost of pasting context for a 15 person team, with every assumption shown.
- [Connect Claude Desktop to your company knowledge in 10 minutes](https://www.company-brain.ai/blog/connect-claude-desktop-to-company-knowledge): A no-code, step by step guide to connecting Claude Desktop to company facts through MCP.
- [Claude Projects for team knowledge: where they work and break](https://www.company-brain.ai/blog/claude-projects-for-team-knowledge): Claude Projects praised for project scope, and the structural break when company facts get copied into every project.
- [Copilot comes with Microsoft 365. Why doesn't it know your facts?](https://www.company-brain.ai/blog/copilot-microsoft-365-company-facts): Why retrieval over a document pile answers from stale decks, and the smaller reviewed fact set that fixes it.
- [ChatGPT Projects vs Custom GPTs: which one holds team knowledge?](https://www.company-brain.ai/blog/chatgpt-projects-vs-custom-gpts): One decision rule for Projects vs Custom GPTs, and the blind spot they share on team knowledge.
- [Two teammates asked AI the same question, got different answers](https://www.company-brain.ai/blog/two-teammates-one-question-different-answers): Why AI answers diverge on facts across a team, and how conflicts become visible at write time instead.
- [Onboarding a new hire when the AI already knows the company](https://www.company-brain.ai/blog/onboarding-a-new-hire-when-ai-knows-the-company): A new hire's first two weeks when the team's AI tools already carry the company's facts.
- [Connect ChatGPT to your company knowledge](https://www.company-brain.ai/blog/connect-chatgpt-to-company-knowledge): A no-code, step by step guide to connecting ChatGPT to company facts through MCP. Requires Developer Mode.
- [Connect Cursor to your company knowledge](https://www.company-brain.ai/blog/connect-cursor-to-company-knowledge): Cursor MCP setup with one JSON entry, and how to verify the tools actually loaded.
- [The AI gave a customer a 20% discount. Who said it could?](https://www.company-brain.ai/blog/ai-approved-a-discount-who-said-it-could): Why approval limits for AI assistants need writing down as Mandates, with a worked discount example.

All other routes (dashboard, facts, proposals, settings, etc.) require authentication and are not crawlable.