# MCP and tools

> One MCP for any AI client: sign-in, schema discovery, result limits, errors an agent can fix and logging for every call.

Powerful MCP Capabilities

## Connect Bonnard

- Claude
- ChatGPT
- Cursor
- Claude Code
- Any other client: https://app.bonnard.ai/mcp

[Talk to us](/get-in-touch)

## Authentication

- OAuth 2.1: PKCE mandatory
- Client registration: Automatic (CIMD, DCR)
- Tokens: Audience-bound per server
- Your own IdP: SSO supported

## Security

- Access: A grant per data source
- Tenancy: From the sign-in
- Read-only: From the first query
- Encryption: AES-256-GCM at rest
- Logs: Every call, logged

## Never leave where you work best

- `get_versions` (Context editors): Lists the context versions and what changed in each. Only for editors and admins, in their own AI client.
- `edit_draft` (Context editors): Edits the context draft: descriptions, keys, time columns, which tables AI clients see, joins and metrics. Only for editors and admins, in their own AI client.
- `publish_draft` (Context editors): Publishes the draft as a new version for every AI client. Works only once an admin allows publishing from AI clients. Only for editors and admins, in their own AI client.
- `explore_schema` (Discover): Finds the tables, columns, keys, joins and metrics for what the agent is asking. The context comes first.
- `explore_docs` (Discover): How-to pages for choosing a chart, each chart type, KPIs, dashboard params, styling and annotations.
- `sql_query` (Ask): Runs one read-only SQL SELECT and returns the rows.
- `render_chart` (Show): Draws a chart, or a carousel of up to ten, right in the chat.
- `render_dashboard` (Show): Shows a dashboard in the chat as an interactive grid of tiles.
- `save_chart` (Charts): Saves a chart the person just saw, to find again later.
- `explore_charts` (Charts): Lists the charts this person saved.
- `delete_chart` (Charts): Deletes one of the person’s own saved charts, once they confirm.
- `save_dashboard` (Dashboards): Saves a dashboard for this person, or edits one they saved.
- `explore_dashboards` (Dashboards): Lists the dashboards the agent can show: verified, shared and saved.
- `delete_dashboard` (Dashboards): Deletes one of the person’s own dashboards. Verified ones stay.
- `create_export_file` (Export): Turns a result into a CSV, or a chart into a PNG, without running the query again.
- `fetch_view_data` (In the chart): Loads a rendered chart or dashboard’s data into the chart, or re-renders a dashboard for a filter pick. Called by the chart, not by the agent.
- `search_param_values` (In the chart): Finds the values for a dashboard’s filter menu. Called by the chart, not by the agent.

## What the MCP handles

- **Sign-in.** Each person signs in with OAuth 2.1, and the agent only sees what they’re allowed to. Automatic client registration (CIMD, DCR).
- **Schema discovery.** The agent gets an overview, then opens only the tables it needs. Each listing suggests the next step.
- **Result limits.** Big results don’t flood the agent, and it knows when it only has part of the data. 250 rows shown, 5,000 fetched.
- **Reply size.** Replies stay small, which leaves the agent room to work.
- **Errors an agent can fix.** A bad call tells the agent how to fix it, so it can retry without you. Nearest real names suggested.
- **Logging.** Every call is logged against the person and client, including the ones that failed. Why the agent made the call.

## Start scaling accurate data today

Free for your first 3,000 calls. No seat limits.

[Talk to us](/get-in-touch)