New

Use clariBI from inside Claude, Cursor, and any LLM client.

The clariBI MCP Server exposes your workspace through the Model Context Protocol. Sign up, list dashboards, run analyses, and generate reports without leaving the chat surface.

Start free trial Read the docs

Spec-compliant MCP

JSON-RPC over streamable HTTP and SSE. OAuth 2.1 with Dynamic Client Registration. Both transports and protocol versions 2024-11-05 and 2025-03-26 supported.

Sign up from chat

New users can create a Trial workspace without leaving their LLM client. Two unauthenticated tools handle the email verification dance.

Typed SDKs

Python (pip install claribi-mcp) and TypeScript (npm install @claribi/mcp). Method surface is codegen-driven from the live tool catalog.

60-second setup

1

Mint an MCP API key

Open Settings > Developer > MCP Server in your clariBI workspace and click Mint MCP API key. The key starts with claribi_mcp_ and is shown once.

2

Paste it into your LLM client

Claude Desktop config snippet:

{
  "mcpServers": {
    "claribi": {
      "url": "https://claribi.com/mcp/v1/",
      "headers": {
        "Authorization": "Bearer claribi_mcp_..."
      }
    }
  }
}
3

Ask Claude about your data

Restart the client. Try "List my clariBI dashboards" or "What was revenue last quarter by region?". Claude calls the right MCP tool, returns the answer in chat.

Detailed Claude Desktop setup → | Cursor setup →

What the LLM can do

15 tools, 3 resource URI schemes, 2 prompt templates. The full catalog is generated from running code, so it stays in sync with what the server actually exposes.

Read your data

  • list_dashboards, get_dashboard
  • list_reports, get_report
  • list_data_sources
  • get_data_source_schema

Run AI analyses

  • run_analysis (consumes AI credits)
  • get_analysis_status
  • generate_report

Manage the account

  • get_usage
  • get_billing_status
  • create_checkout_session

Sign up from chat

  • register_account (no auth)
  • verify_email (no auth)
  • check_pricing (no auth)

Embed live artifacts

  • dashboard://<id>
  • report://<id>
  • data-source://<id>/schema

Prompt templates

  • analyze_question
  • weekly_review
See the full tool reference →

Built for shared workspaces

Tokens hashed at rest

API keys, OAuth access tokens, and refresh tokens are stored as SHA-256 hashes. The raw value is shown once and never again.

PKCE mandatory

Public clients must use PKCE with SHA-256 challenges. Authorization codes are single-use and expire in 60 seconds.

Per-IP and per-org rate limits

Open registration is rate-limited at 20 requests per hour per IP. Authenticated traffic is throttled per organization to bound damage from runaway agents.

Full audit trail

Every tool call writes an audit row with the calling user, organization, scope, duration, and credits consumed. Write actions feed the organization-level audit log.

Build your own client

Python

pip install claribi-mcp

from claribi_mcp import Client

with Client(api_key="claribi_mcp_...") as client:
    client.initialize()
    result = client.run_analysis(
        question="What was MRR last month?",
    )
    print(result.text)

TypeScript

npm install @claribi/mcp

import { Client } from '@claribi/mcp';

const client = new Client({
  apiKey: 'claribi_mcp_...',
});
await client.initialize();
const r = await client.callTool('run_analysis', {
  question: 'What was MRR last month?',
});
console.log(r.text());
Full SDK guide →

Ready to wire it up?

Trial accounts include MCP server access from day one. Sign up via the website or directly from your LLM client.

Start free trial Read the docs