CORPILUS

Connect your AI agents directly to Claude.ai and ChatGPT

Your knowledge base, database, and workflows — natively available in the AI tools you already use. No copy-paste, no exports, no custom integration.

Works with Claude.ai · ChatGPT · Cursor · Claude Desktop

What is MCP?

MCP (Model Context Protocol) is an open standard from Anthropic, released in 2025. It lets AI assistants like Claude and ChatGPT discover and call external tools through one universal protocol.

Practically: instead of copy-pasting data into AI chats or building custom one-off integrations, you connect your AI assistant once to Corpilus and it sees your real-time data — knowledge base, databases, workflows.

The key shift from traditional API integrations: with MCP, the AI itself decides which tool to call based on the user's question. You don't program the orchestration — the AI does.

AI client

Claude.ai, ChatGPT, Cursor

Corpilus MCP

Auth, gating, BYOK, RLS

Your data

KB, DB, Workflows

Real-world scenarios

Concrete examples of what your team can do once Corpilus MCP is connected.

Sales rep

Summarize everything we know about client X

Claude searches the KB, pulls the last 5 deals from your DB, and lists relevant emails — all in one response.

Consultant

Generate a Q3 cash-flow report

ChatGPT runs a workflow that queries the accounting system, summarizes, and posts the result to Slack.

Support

What's the status of ticket #1234?

The AI fetches the ticket, latest comments, and suggests a response in your tone of voice.

Marketing

What's the customer sentiment from the last month?

Claude calls the analysis tool, runs sentiment scoring across recent feedback, and returns structured insights.

Supported MCP clients

Connect any of the following — all work with the same Corpilus MCP server URL.

Claude Desktop

Anthropic's desktop app for macOS and Windows. Static API key, JSON config.

Static API keySetup guide →

Claude.ai (web)

Browser-based Claude. OAuth-driven Custom Connector — no install needed.

Plug-and-play (OAuth)Setup guide →

ChatGPT

OpenAI ChatGPT (Plus/Team/Enterprise). OAuth-driven MCP Connector.

Plug-and-play (OAuth)Setup guide →

Cursor

AI-first code editor. Static API key, mcp.json config — for developer use cases.

Static API keySetup guide →

REST API

Any HTTP client. Same auth, same scopes — for backend automation, scripts, integrations.

Static API keySetup guide →

The setup guide is provided as part of the installation, walking you through every step.

Quick start

What you'll paste into your AI client

Your API key is issued as part of the installation — you'll paste it into the snippet below.

  1. 1Open Claude Desktop → Settings → Developer → Edit config.
  2. 2Paste the snippet below into claude_desktop_config.json.
  3. 3Replace YOUR_CORPILUS_API_KEY with the key from your Corpilus dashboard.
  4. 4Restart Claude Desktop — Corpilus tools appear in the tool picker.
{
  "mcpServers": {
    "corpilus": {
      "url": "https://api.corpilus.com/api/v1/mcp/v1",
      "headers": {
        "Authorization": "Bearer YOUR_CORPILUS_API_KEY"
      }
    }
  }
}

YOUR_CORPILUS_API_KEY is issued as part of your installation, available in Settings → MCP → API Keys.

Request a quote

Access starts once your quote is agreed.

Available tools

25 production-ready MCP tools, organized by category — every tool is available in your installation, none held back.

Knowledge

4 tools
  • Search KB by keyword
  • Ask KB in natural language (RAG)
  • List collections in KB
  • Add documents to KB

Database

4 tools
  • Query DB in natural language
  • List available tables
  • Describe table schema
  • Run a SQL query directly

Workflows

3 tools
  • List workflows in your tenant
  • Trigger a workflow
  • Check workflow run status

Analysis

3 tools
  • Smart Analysis (multi-step reasoning)
  • Classify text into your taxonomy
  • Extract named entities

Protection

3 tools
  • Scan text for PII / sensitive data
  • Scan text for malicious patterns
  • Anonymize text (replace PII with placeholders)

Profile

2 tools
  • Get tenant profile / settings
  • Update tenant profile / settings

Shield protection

6 tools
  • Scan uploaded files (PII / malware)
  • Shield protection statistics
  • Active threats and attacks
  • Security events history
  • Add detection rule
  • Verify bot/human token

Security model

Built on Phase 1+2+3a controls: tenant isolation at the database level, BYOK for AI tokens, OAuth for browser-based clients.

BYOK

Bring your own OpenAI / Anthropic key. We proxy through it — never store it elsewhere.

Tenant isolation (RLS)

PostgreSQL row-level security enforces that no tenant can see another tenant's data, ever.

Killswitch

Disable MCP for your tenant from Settings with one click. Takes effect immediately.

OAuth + revocable tokens

Web clients authenticate via OAuth — see exactly which scopes they request, revoke individual connections without affecting others.

Pricing

Included with your installation

MCP server access is part of your Corpilus installation — no separate license fee. Usage is covered by your agreed scope, with surcharges only for cloud storage and, if you use our AI models, for tokens.

Scope is agreed with you upfront as part of the managed installation — never a surprise paywall mid-task.

With BYOK, you pay your AI provider directly for tokens — we don't mark up your provider's pricing.

See full pricing →

Frequently asked questions

Which AI tokens are used when MCP calls OpenAI?
If you've configured BYOK in Settings → AI, your own OpenAI/Anthropic key is used. Otherwise, we use the platform key — usage is covered by your agreed scope, with a token surcharge since you're using our model rather than your own key.
Does Anthropic / OpenAI see my data?+
It depends on your data protection mode (Settings → Data Protection). 'Strict' mode anonymizes PII before any third-party AI call. 'Permissive' sends raw data. Default is anonymize.
What happens if Claude.ai goes down?+
Nothing on our side — the Corpilus MCP server runs in our cloud (or on-prem if you self-host). Claude.ai is just a consumer. Your data stays available; only that specific AI consumer is affected.
Can I revoke one client (e.g. Claude.ai) without affecting Cursor?+
Yes. Each connection is tracked separately in Settings → MCP Server → Active Connections. Revoke them individually.
How does Claude.ai authenticate vs Claude Desktop?+
Claude.ai uses OAuth (browser-based, revocable tokens). Claude Desktop uses a static API key you generate in your dashboard. Both are equally secure for their respective deployment.
Does this work behind a corporate firewall?+
Cloud Corpilus needs outbound HTTPS from the AI client. For strict firewalls, we offer on-premise installation — your MCP server runs inside your network, AI clients connect to it on your terms.
What are the usage limits?+
MCP usage is part of the scope agreed with you in the proposal, sized to your expected usage. AI tokens are a surcharge only if you use our models — bring your own key and you don't pay us for tokens. Request a quote and we'll size the scope with you.
How do we get started?+
Request a quote — once it's agreed, we set up your access and issue your API key as part of the installation. From there, follow the in-dashboard Setup Guide for your AI tool of choice.

Connect your AI tools in 2 minutes

Managed installation, priced to your scope. Request a quote to get started.