Quarri MCP Server

AI data analyst for your business data

What is Quarri?

Quarri connects to your business data sources — APIs, databases, spreadsheets, and scanned PDFs — cleans and transforms the data into a unified analytical model, and lets anyone on your team query it using natural language. No SQL required.

Quarri builds a semantic memory of your business rules, column meanings, and domain context that compounds over time, making every query smarter. Features include automated data pipelines, dimensional modeling, interactive dashboards, reconciliation workflows, and AI-powered statistical analysis — all through a conversational MCP interface.

Getting Started

1. Create an Account

Sign up at quarri.ai or contact theo@quarri.ai for an onboarding session. Each customer receives a dedicated, tenant-isolated database and a unique MCP endpoint.

2. Connect via Claude Desktop

Open Claude Desktop → Settings → Integrations
Search for Quarri in the MCP Directory, or add a custom connector pointing to your MCP URL
Complete the OAuth login flow when prompted
Start asking questions about your data

Authentication

Quarri uses OAuth 2.0 Authorization Code flow with PKCE (S256). After authenticating, your Claude session receives a Bearer token valid for 30 days. No API keys or manual configuration required.

Example Prompts

Example 1: Natural Language Analytics

"What were our total gross dollars by species for January 2026, excluding voided transactions?"

Quarri retrieves the schema and business rules, generates SQL against your unified star schema, executes it, and returns a formatted table. It knows that pay_species is the correct species field and that flag_void = 1 means voided — this context is persisted in business rules.

Example 2: Automated Reports with Visualizations

"Build me a monthly stumpage accrual report broken down by department and counterparty"

Quarri uses saved skills (reusable procedures) to execute a multi-step workflow: query the fact table, apply complex business rules, and render an interactive HTML dashboard with ECharts visualizations — or generate an Excel workbook with pivot tables.

Example 3: Data Pipeline Setup

"I just uploaded a new CSV of species reference data. Can you create a staging view and add it to the data model?"

Quarri introspects the raw table (detecting column types, PII, primary key candidates), proposes a staging transformation (snake_case naming, type casting, deduplication), creates the view in a dev environment, detects foreign key relationships, and regenerates the unified star schema. Changes are promoted to production only after review.

Example 4: Reconciliation

"Compare our vendor invoices against the journal entries and flag any billing discrepancies over $500"

Quarri queries across fact tables, joins on shared keys, computes differences, and surfaces anomalies. In production use cases, this has caught $300K+ in billing errors within the first month.

Tool Categories

Category	Tools	Access
Analytics	`execute_sql`, `get_query_context`, `search_values`, `get_schema`	Read
Metrics	`get_metrics`, `search_metrics`, `create_metric`, `approve_metric`	Read Write
Business Rules	`list_rules`, `create_rule`, `update_rule`, `delete_rule`	Read Write
Data Pipelines	`upload_csv`, `configure_extraction`, `execute_staging_view`, `execute_silver_view`	Write
Environments	`create_environment`, `promote_environment`, `rollback_production`	Write Destructive
Content	`publish_content`, `list_content`, `share_content`	Read Write
Skills	`search_skills`, `create_skill`, `get_skill`	Read Write
Schema Management	`execute_ddl`, `execute_dml`, `generate_quarri_schema`	Destructive

All tools include MCP safety annotations (readOnlyHint, destructiveHint). Destructive operations require a development environment — production can only be modified via promote_environment.

Read/Write Capabilities

Quarri is a read + write MCP server. Read operations include SQL queries, schema inspection, and metric/rule lookups. Write operations include creating views, executing DDL/DML in development environments, uploading CSVs, publishing content artifacts, and configuring data pipelines.

Write operations are environment-gated: development environments are isolated copies; production requires explicit promotion with automatic snapshots for rollback.

Data Handling

The server only accesses data explicitly requested by the user via tool calls
All data transmission is encrypted (HTTPS/TLS)
Each customer has a dedicated, isolated database (tenant-level isolation)
No conversation data is collected or stored
GDPR compliant with US or EU hosting available

Support

Email: theo@quarri.ai

Privacy: Privacy Policy