Comparisons
Bonnard vs Sisense: MCP server features compared
Choose Sisense when agents must answer inside analytics you already embed in your product, under each user's row and column security. Choose Bonnard when you want the answers without the BI platform, at a published price with a free plan.
On this page
With Sisense (sisense.com), the AI client sends a request in plain language, and Sisense’s LLM turns it into a query on your data model. With Bonnard (bonnard.ai), the AI client writes the SQL itself.
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.
Sisense also gives its context to the AI client. getDataSources lists the data models the user can access. getDataSourceFields returns the tables, columns, relationships and formulas of one data source, and can be scoped to a perspective. getDataPreview returns the distinct values of a text column (MCP tools).
The agent does not write the query. It describes what it wants in plain language, and buildQuery has Sisense’s own LLM convert the request to a query on the data model (MCP server). No Sisense tool takes SQL, so the model’s relationships and formulas apply to every query. The tools are read-only.
Two of Sisense’s controls for AI are documented for its assistant and natural language queries, and not for MCP. AI Context (Beta) holds instructions of up to 500 characters on a model, table, column or formula. The hide_from_ai tag keeps a table or column out of AI features, and the 2026.4 release notes state that it does not cover the MCP Server. Use a perspective or data security to keep a field from an MCP agent, and ask Sisense whether buildQuery reads AI Context.
How wrong answers get found and fixed
Bonnard runs a loop of four steps on how agents use your data. Sisense has request logs for the first step, and a data designer does the other three by hand.
- ObserveAgents use your data
- DiagnoseBonnard finds what’s wrong
- ImproveYour data team answers, into the context
- VerifyEvals check it helped
Observe. Bonnard records each tool call with its SQL, status and error, from every AI client your people use.
Sisense records each MCP request against the signed-in user. Its MCP docs state that “Auditing is log-based today”, and that an admin view of agent activity is planned for a later release. AI Usage in Usage Analytics (Beta, 2026.4.1) counts users and requests for the assistant, Narratives, Semantic Enrichment and Search. It does not list MCP, and it does not store the text of questions or responses (2026.4 release notes).
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.
Last 30 days- Inconsistent definitionsQuestion
net revenue3 definitions, 8 people - Searches that found nothingData gap
"gift cards"6 people - Struggling work4 sessions, 9 calls eachContext gap
- ErrorsContext gap
undefined_column14 calls - Stale or misleading dataContext gap
orders_oldstopped in May
Sisense has no feature that turns agent activity into suggested changes. A data designer finds the cause by hand.
Improve. Your data team’s answer goes into a draft of the context.
On Sisense, a data designer changes a name, description or AI Context and publishes the semantics (Optimizing Your Data Model for AI). Git Integration puts data models and dashboards in a Git repository, with review through merge requests on the remote.
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.
- Which products brought in the most revenue in August?Better
- What is our average order value?Better
- What share of orders were refunded?No clear difference
Sisense has no evals for agent answers. The data designer asks the question again by hand.
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.
Sisense prevents one kind of wrong answer by resolving every query on the data model. Bonnard finds wrong answers in the record of agent use and tests each fix before it is published.
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.
Sisense starts from a Sisense deployment on version 2026.3.1 or later, with a data model in it (MCP server). buildQuery and the chart and dashboard tools also need Cloud-Linked Features and an LLM provider: the LLM that Sisense manages, or your own. The two discovery tools need neither. A self-hosted deployment has the server off until an administrator enables it at install or upgrade (self-hosted guide). The answers are as good as the names, descriptions and formulas in the data model.
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 Sisense, each person signs in to the MCP server with their own Sisense credentials or SSO, and there is no shared sign-in (connect an AI agent). Each tool call runs as that user, under their Sisense roles and row- and column-level data security.
Your customers’ users can connect if each has a Sisense user account. Sisense’s OEM guide signs embedded users in through SSO and names the Viewer and Viewer Plus roles for them, and both use a Viewer license. On a self-hosted deployment with multitenancy and more than one tenant, AI clients cannot reach /mcp before 2026.4.1 (self-hosted guide). That release is out for cloud customers only so far (2026.4 release notes).
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).
Sisense publishes no prices. Its plans page names two plans: Self-serve, with a free trial link, and Enterprise, which links to a demo request. The MCP docs tag the server for the Launch, Grow and Enterprise tiers. Sisense licenses users by role category: Administrator, Designer or Viewer (user roles).
buildQuery and the chart and dashboard tools use an LLM on Sisense’s side. The LLM that Sisense manages draws on AI credits, which refresh each month by subscription (2026.2 release notes). A group that uses up its allocation cannot use credit-based AI features until the next cycle or a new allocation (credit usage). Your own LLM provider bills you directly and uses no Sisense credits.
Side by side
| Feature | Bonnard | Sisense |
|---|---|---|
| Who answers | The AI client the person uses, with Bonnard’s context | The AI client the person uses. Sisense’s LLM writes each query |
| Context the agent reads | Schema, descriptions, joins, metrics, business terms, searchable values | Tables, columns, relationships and formulas of a data model |
| Context through MCP | explore_schema, for any AI client | getDataSources, getDataSourceFields and getDataPreview, for any AI client |
| Metric definitions | The agent reads the SQL and writes the query | Enforced. buildQuery resolves each request on the data model, and no tool takes SQL |
| Observe | Records each tool call, from any AI client | Request logs. No admin view of agent activity yet |
| Diagnose | Finds what is wrong in the activity record | By hand |
| Improve | The data team’s answer goes into a draft | By hand. Git Integration for data models and dashboards |
| Verify | Evals: the draft against the live version | By hand. No evals |
| Run a query | sql_query, one read-only statement on your database | buildQuery, plain language that Sisense’s LLM converts to a query |
| Charts and dashboards | render_chart and render_dashboard, in the AI client | createWidget, createDashboard and 16 edit tools, shown in Claude and ChatGPT as drafts |
| Who can connect | Unlimited users, your customers’ users included | People with a Sisense user account |
| Needed first | A data source: your database, or the sample data | A data model in Sisense 2026.3.1 or later, and an LLM provider for queries |
| Databases and warehouses | 7, plus PostHog | 12 live connectors, and more for imported data |
| Tools for the agent | Up to 17 | 29, of which 7 are for Pulses |
| How you pay | For each call, no seats | No published prices. Users licensed by role, plus AI credits on Sisense’s LLM |
| Free plan | 3,000 calls a month | None listed. Self-serve has a free trial link |
When Sisense is stronger
- You want metric definitions enforced: every MCP query goes through
buildQueryon the data model, and the agent cannot write SQL. Bonnard gives the agent the metric’s SQL and does not rewrite the query. - Your data models are in Sisense: agents use their relationships and formulas under each user’s row- and column-level data security (MCP server).
- You keep models in git: Git Integration tracks data models and dashboards, and changes go through merge requests on your remote repository (Git Integration).
- You have more kinds of source: Sisense has 12 live connectors, among them Snowflake, Databricks, ClickHouse and Oracle, and imports from files and applications (data sources).
- You self-host: the MCP server runs on your own Sisense deployment (self-hosted guide).
- You want alerts on your data: since 2026.4.1, 7 MCP tools list, inspect and run Pulses, the scheduled checks that a user creates in Sisense’s assistant (MCP tools).
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/mcpSisense. On Managed Cloud, the MCP server is on by default under System Configuration > MCP. Claude’s and ChatGPT’s connector directories list Sisense, and ask for your Sisense URL before sign-in. To skip that prompt, add a custom connector with your Sisense URL as the target. On Claude Team and Enterprise, an owner adds the custom connector. Cursor takes the same URL.
https://mcp.sisense.com/mcp?target=<your-deployment-url>Charts and dashboards show in Claude and ChatGPT after an administrator adds each client’s origin under Admin > Security Settings > CORS Allowed Origins. Without the entry, users get text answers. A self-hosted deployment serves the MCP server at its own URL, added as a custom connector.
https://<your-deployment>/mcpVerified 16 Nov 2026 against Sisense’s MCP docs, connect guide, tools page, self-hosted guide, Optimizing Your Data Model for AI, AI Context, credit usage, Git Integration, plans, user roles, OEM guide, data sources and release notes for 2026.2 and 2026.4.
Get started
Start scaling accurate data today
Free for your first 3,000 calls. No seat limits.
