Bonnard
On this page

With GoodData (gooddata.ai), the agent runs saved visualizations and metrics on your model, and has no SQL tool. With Bonnard (bonnard.ai), the agent reads your schema and your team’s context, then writes one read-only SQL statement.

Five differences

What the agent knows about your data

Bonnard gives its context to the AI client the person already uses. explore_schema searches table names, descriptions, column comments, enum labels, metrics and business terms, and returns up to 10 tables. For each table the agent gets the keys, row count and sample rows from the database, plus what your team adds: descriptions, joins, metrics, business terms and the values of columns you mark as searchable (context). A metric is its calculation in SQL, with instructions. The agent reads it and writes the query. Bonnard does not rewrite the query to apply it.

GoodData gives its context to outside AI clients too. Each workspace has a logical data model (LDM): datasets of facts and attributes, with their keys and relationships. Metrics are expressions in MAQL (GoodData’s query language) on that model. Through GoodData’s MCP server, list_workspace_metrics and list_workspace_attributes list the objects the user can access, and a single metric comes back with its MAQL definition. get_workspace_analytics returns the whole analytics model of the workspace. Generated tools read the other workspace metadata. get_maql_guide and the knowledge tools return GoodData’s guidance on MAQL and the LDM. A workspace manager can mark metrics, visualizations and dashboards as certified.

GoodData enforces the definitions. Its server has no SQL tool, and its docs say AI clients reason with trusted metrics “instead of raw or ad-hoc queries”. An outside AI client gets numbers through the model. execute_visualization runs a saved visualization. create_visualization runs a new definition, and needs the experimental MCP Apps feature. ai_chat sends the question to GoodData’s own AI. The server can also change the model: create_metric adds a metric to the workspace, and deploy_workspace_analytics replaces the whole analytics model.

How wrong answers get found and fixed

Bonnard runs a loop of four steps on how agents use your data. GoodData has a feature for each step, built for its own AI Assistant, and two of them are experimental.

01020304 OBSERVEAgents useyour dataDIAGNOSEBonnard findswhat’s wrongIMPROVEYour datateam answers,into the contextVERIFYEvals check it helped
01020304 OBSERVEDIAGNOSEIMPROVEVERIFY
  1. ObserveAgents use your data
  2. DiagnoseBonnard finds what’s wrong
  3. ImproveYour data team answers, into the context
  4. VerifyEvals check it helped

Observe. Bonnard records each tool call with its SQL, status and error, from every AI client your people use.

GoodData’s AI Observability Workspace is experimental and a paid entitlement. Its dashboard shows usage, conversations, token use, errors and feedback for the organisation, with no drill-down to a single conversation. Ask GoodData whether tool calls from an outside MCP client appear there. The audit log, on Enterprise, records security and change events, and does not record each interaction.

Diagnose. Bonnard reads that record and finds what is wrong: definitions that differ between people, searches that found nothing, sessions that took many calls, repeated errors and tables that stopped updating. Where the data cannot settle a question, Bonnard asks your data team.

Suggested AI Memories, an experimental feature, turns negative feedback on an answer into a disabled memory item. The Semantic Quality Agent checks the titles and descriptions of metrics, attributes, facts and date datasets for duplicates and unknown abbreviations, each time you start a run.

Improve. Your data team’s answer goes into a draft of the context.

On GoodData, an admin decides which suggested memory items to enable. AI Memory holds instructions that the AI Assistant applies always or when relevant. GoodData for VS Code, in beta, keeps analytics as YAML files, with a CLI that validates and deploys them.

Verify. Evals run the same questions against the draft and the live version, and show which answers got better. A person then publishes the draft, and Bonnard keeps each version.

gd-eval is a command-line tool that runs a set of questions against the AI Assistant. It checks visualizations against a definition you expect, and free-text answers with an LLM as the judge. GoodData recommends running the same set again after a change to the model, the memory or the agent.

Two warnings also reach the agent as it works. A result over the row limit tells the agent not to sum or average it. On PostgreSQL, a join that could count rows twice gets a warning.

GoodData’s loop tunes its own AI Assistant, which an outside AI client reaches through ai_chat. Bonnard’s loop starts from the calls that Claude, ChatGPT or Cursor made.

What you need first

Bonnard starts from your database. Add it as a data source through a read-only credential, then give people access. The agent reads the table comments, keys and enum values that the database declares, and the context your team adds in Bonnard.

GoodData starts from a workspace on the Enterprise plan, connected to a database, with a logical data model and metrics on it. The MCP server can do part of that work: scan_datasource reads the database schema, and generate_ldm generates an LDM from the result. Answers are as good as the model, its metrics and their titles and descriptions. Charts in the chat also need Share data with AI on, for the organisation or the workspace. It is off by default.

Who can connect

A Bonnard workspace has unlimited users, your customers’ users included. Each customer’s query runs as that customer’s tenant (the value that marks its data in your database). A customer with no tenant gets no data. Isolation uses the database: row-level security on PostgreSQL, Redshift and SQL Server, or one dataset or database for each tenant on BigQuery and Azure SQL Database (tenancy).

On GoodData, each tool call runs as a GoodData user, with the GoodData permissions of that user (GoodData’s MCP docs). Each MCP URL names one workspace, and GoodData’s multitenancy gives each tenant its own workspace. Interactive SSO, in beta, signs the person in through your identity provider. A per-user API token acts as its owner, and a shared token gives all callers one identity. For your customers’ users, your backend signs a JWT and GoodData acts as the mapped user.

How you pay

Bonnard bills for each call, with no seats or minimums. Free includes 3,000 calls a month, then USD 0.95 for each 100 calls once an admin turns on extra usage (pricing).

GoodData’s MCP server is on the Enterprise plan only, at a custom price (GoodData pricing). AI Memory and AI observability are on Enterprise too. Professional has no MCP server, and its price is a platform fee plus the number of workspaces. Professional has unlimited users, and Enterprise includes everything in Professional. Neither plan shows a price. Plans are annual. By default, each user can run 30 AI queries a day. Ask GoodData whether MCP calls count towards that limit.

Side by side

Bonnard and GoodData MCP servers, verified 16 Nov 2026
FeatureBonnardGoodData
Who answersThe AI client the person uses, with Bonnard’s contextThe AI client, with GoodData’s model, or GoodData’s AI through ai_chat
Context the agent readsSchema, descriptions, joins, metrics, business terms, searchable valuesLogical data model, MAQL metrics, visualizations, dashboards, MAQL guide
Context through MCPexplore_schema, for any AI clientlist_workspace_metrics, list_workspace_attributes and get_workspace_analytics, for any MCP client
Metric definitionsThe agent reads the SQL and writes the queryEnforced. MAQL on the model, with no SQL tool
ObserveRecords each tool call, from any AI clientAI Observability Workspace, experimental, for GoodData’s AI Assistant
DiagnoseFinds what is wrong in the activity recordSuggested AI Memories from negative feedback, experimental
ImproveThe data team’s answer goes into a draftAn admin enables a suggested memory item
VerifyEvals: the draft against the live versiongd-eval from the command line, for GoodData’s AI Assistant
Run a querysql_query, one read-only statement on your databaseexecute_visualization on a saved visualization, or ai_chat
Charts and dashboardsrender_chart and render_dashboard, in the AI clientcreate_visualization in an MCP Apps host, experimental
Who can connectUnlimited users, your customers’ users includedGoodData users, through SSO, a JWT or an API token
Needed firstA data source: your database, or the sample dataA workspace with a logical data model and metrics
Databases and warehouses7, plus PostHog21 connectors, 4 of them in beta
Tools for the agentUp to 1735 named tools, plus generated tools, found through search_tools
How you payFor each call, no seatsEnterprise plan, custom price, annual
Free plan3,000 calls a monthNone on the pricing page. Ask GoodData about a trial

When GoodData is stronger

  • You want the semantic layer enforced: GoodData’s server has no SQL tool, so numbers come from the model. Bonnard gives the agent the metric’s SQL and does not rewrite the query.
  • You already model in GoodData: the agent reads your metrics with their MAQL, and adds new ones with create_metric.
  • You want agents to create metric alerts and scheduled exports, or to export a dashboard as a PDF or a PPTX slide deck with run_export.
  • You need a database Bonnard does not connect to: GoodData lists 21 connectors, 4 of them in beta.
  • You need the server on your own Kubernetes cluster: GoodData.CN installs it with its Helm chart, where deployMCPServer is true by default.

Add each server to your AI client

Bonnard. Create a workspace, add Bonnard to your AI client, then add a data source (quickstart). Use the install links for Claude, Cursor and ChatGPT in the console at app.bonnard.ai, or add this URL.

https://app.bonnard.ai/mcp

GoodData. The MCP server is on the Enterprise plan. Each workspace has its own MCP URL. In Claude Code, run claude mcp add --transport http GoodData "<url>", with an Authorization header that carries an API token. Claude Desktop connects through mcp-remote, run with npx, and an API token. Cursor and VS Code with GitHub Copilot take the URL in mcp.json. With Interactive SSO, leave out the header and sign in through the browser. It needs Interactive SSO on in the organisation’s Early Access settings. tools/list returns only search_tools and call_tool, and the agent finds the other tools with search_tools.

https://<your-gooddata-host>/api/v1/actions/workspaces/<workspaceId>/ai/mcp

Verified 16 Nov 2026 against GoodData’s MCP docs, MCP Apps docs, pricing, data model, object certification, AI Memory, AI Observability Workspace, AI Assistant Evaluation, Semantic Quality Agent, audit log, GoodData for VS Code, AI Assistant settings, data sharing, multitenancy, databases, release notes, GoodData.CN install, Helm chart options and GoodData.CN MCP docs.

Get started

Start scaling accurate data today

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

Keep reading