> ## Documentation Index
> Fetch the complete documentation index at: https://docs.databar.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Configuration

> Authentication, spending guardrails, and task data retention for the Databar MCP server

The hosted server at `https://mcp.databar.ai/mcp` has nothing to install or configure. Databar manages its defaults, and every client gets the same behavior.

## Authentication

* **OAuth** (recommended): add the server URL in your client (Claude Connectors, Cursor, or any remote MCP client) and approve access to your workspace when prompted.
* **API key**: send `Authorization: Bearer your-api-key-here` with every request. Find your key in your [Databar workspace](https://databar.ai) under **Integrations**.

## Spending guardrails

Tools that spend credits (enrichments, waterfalls, table enrichments, flows) check the cost before anything is queued:

* **Balance check**: if your balance cannot cover even the cheapest step, the tool returns an error with `balance_credits` and `required_credits` instead of queueing a run that would fail.
* **Confirmation above 50 credits**: when a run's estimated ceiling is at least 50 credits, the tool returns `status: confirmation_required` with `estimated_credits`. Your agent shows the estimate, and after you agree it re-calls the tool with the `retry_with` object from that response.

## Task data retention

<Warning>
  Task data is stored for **24 hours** after completion. If you need the results later, make sure to retrieve them promptly. After 24 hours, task data is permanently deleted and the status will return `gone`.
</Warning>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.