Give your AI assistant a backend

Connect Claude, Cursor, or any MCP-compatible client to your Centrali workspace with a single URL. Browser-based OAuth, workspace-scoped access, 100+ callable tools.

AI agents need data, but connecting them is painful

Today's AI assistants are powerful — until they need to touch your actual data. Then you're writing custom tool definitions, managing auth tokens, and building one-off integrations.

What teams actually do with it

Natural-language access to your workspace

Three patterns we see from teams pointing Claude and Cursor at their Centrali workspace.

Data exploration

Ask your AI to query records, analyze schemas, and explain data relationships in natural language.

You ask

Show me all orders from last week with total > $500

Automated workflows

Let agents create compute functions, wire up triggers, and build orchestrations from descriptions.

You ask

Create a function that sends a Slack alert when inventory drops below 10

Permission audits

AI assistants can scan service account permissions, simulate access checks, and suggest remediation.

You ask

Audit the billing-service SA and fix any missing permissions

One URL. 100+ tools. Zero configuration.

Centrali's hosted MCP server exposes your entire workspace to AI assistants through a single URL. Authentication, workspace scoping, and tool registration are handled automatically.

Single URL setup

Add https://mcp.centrali.io to your AI client. No API keys, no environment variables, no local installation required.

Browser-based OAuth

Authenticate by logging in through your browser. Tokens carry your identity — audit logs show who did what.

Workspace scoping

Select which workspace the agent operates in. All tool calls are automatically scoped to that workspace's data.

100+ built-in tools

Collections, records, compute functions, orchestrations, search, pages, IAM — all exposed as callable MCP tools.

Discovery tools

Agents call describe_centrali to understand your workspace, then explore schemas, queries, and capabilities on their own.

Service account mode

For CI/automation, use service accounts with RBAC permissions. Fine-grained access control via roles, groups, and policies.

Connect in seconds

One command in Claude Code, or drop the URL into your client's MCP config.

claude mcp add centrali --transport http https://mcp.centrali.io

Connect your AI assistant in 30 seconds

Add the MCP URL to your AI client, log in, and start building with your data.