Bonnard
On this page

With Omni (omni.co), each query selects fields from your Omni model, and Omni’s AI writes most of them. With Bonnard (bonnard.ai), the AI client writes SQL on your database from context your team adds.

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.

Omni keeps its context in a semantic model. A topic groups the views, joins, dimensions and measures for one use case (topics). Omni sends field descriptions, synonyms and ai_context from the model to the LLM (AI usage). A topic can also hold sample queries (AI optimization), and an Organization Admin can add Markdown files that the Omni Agent reads across all models (global context).

Omni’s MCP tools give the model to outside AI clients too. listModelTopics lists the topics. getModelTopic returns one topic with its views, joins, dimensions, measures and descriptions. getModelViews lists the views and their fields (tools). These tools need a connection for one user, such as OAuth.

Omni’s AI writes most queries. getData runs the question against a topic through the same AI pipeline as the Omni Agent. askOmni gives a multi-step question to the Omni Agent. An AI client can also build a query in Omni’s JSON form and run it with runQuery.

Metric definitions are enforced. Each query selects fields from the model, no MCP tool runs SQL on the database, and runQuery refuses keys that run raw SQL, for example userEditedSQL. Omni’s AI can add CTEs and custom fields to a query, and a model’s ai_settings can restrict both.

How wrong answers get found and fixed

Bonnard runs a loop of four steps on how agents use your data. Omni has the same four steps for the Omni Agent, and Organization Admins find them in the AI Hub.

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.

Omni’s Analytics section logs AI prompts, responses and sessions. The AI Hub classifies each AI session for intent, outcome and sentiment. From the AI client, submitFeedback records a thumbs up or down on an askOmni answer (tools).

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.

Omni’s Suggestions run once a day on a schedule that an admin enables for a model. They read the last 30 days of chat sessions that carry feedback, detected sentiment or an outcome such as stuck or abandoned. Each suggestion proposes a change to ai_context, a description, synonyms or sample values. Learn from conversation proposes model changes from one chat. Ask Omni whether getData calls from an AI client count as sessions for Suggestions.

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

On Omni, an admin implements each suggestion through Omni’s Modeling Agent or ignores it. A Querier or higher accepts, edits or rejects the changes from Learn from conversation. Model changes go through branches, and through pull requests when the model is connected to git (Branch Mode).

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.

Omni’s Evals run a prompt set of up to 25 questions against a model branch or main. An LLM judge that Omni sets gives each answer a pass or a fail, and two runs of the same prompt set show side by side. You run them in the AI Hub, through the API or with omni ai-eval in Omni’s CLI (CLI).

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.

Omni’s evals score the Omni Agent, the engine behind askOmni. Bonnard’s loop improves what Claude, ChatGPT or Cursor read before they write a query.

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.

Omni starts from a connected database and an Omni model with topics (topics). The MCP server queries only topics, unless the model’s query_all_views_and_fields setting and the X-MCP-Query-All-Views header permit all views (Claude guide). The answers are as good as the descriptions, synonyms and ai_context your team has written in the model.

An Organization Admin must set Enable AI and Omni Agent to on (MCP docs). MCP server is on by default, and three more settings control groups of tools (MCP settings).

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).

With OAuth, each person signs in to Omni with an Omni account, and each MCP query uses that person’s Omni permissions. A connection with an API key uses the permissions of the key’s creator (authentication). The AI client lists only the tools that the person’s roles allow (tools).

For embed users, Omni documents a chat agent that you build. Embed users do not sign in to Omni, so your backend mints a token for each one with createEmbedMcpToken, and your agent sends it to the MCP server (embed MCP). Each query then runs with that user’s Omni permissions. Omni’s embed guide recommends its AI Jobs API over MCP for this route.

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).

Omni publishes no licence prices. Its site offers a demo and a free trial through a request form (free trial), and fees are set in an Order Form (terms). Omni’s licences have three tiers: Viewer, Standard and Administrator/Developer (licence types). External users count towards licence usage.

A Viewer can run only the queries of published content, and gets no getData (authentication). The lowest role that queries topics, Restricted Querier, has a Standard licence.

MCP use also draws AI credits (AI usage). The LLMs that Omni hosts are billed in credits, except Omni AI, which is in early access. On an Omni Enterprise plan, an organisation can connect its own LLM provider key (model providers). After an Organization Admin accepts overage billing, usage over the allotment in your plan costs USD 1.00 for each credit. Ask Omni for the allotment in your plan.

Side by side

Bonnard and Omni MCP servers, verified 16 Nov 2026
FeatureBonnardOmni
Who answersThe AI client the person uses, with Bonnard’s contextOmni’s AI writes the query for getData and askOmni. The AI client can build one for runQuery
Context the agent readsSchema, descriptions, joins, metrics, business terms, searchable valuesTopics, views, joins, dimensions, measures, descriptions, synonyms, ai_context
Context through MCPexplore_schema, for any AI clientlistModelTopics, getModelTopic and getModelViews, for any AI client
Metric definitionsThe agent reads the SQL and writes the queryEnforced. Queries select fields from the model, and no MCP tool runs raw SQL
ObserveRecords each tool call, from any AI clientAI session logs with intent, outcome, sentiment and feedback
DiagnoseFinds what is wrong in the activity recordSuggestions, once a day from chat sessions. Learn from conversation
ImproveThe data team’s answer goes into a draftAn admin implements or ignores each suggestion. Branches, and git with pull requests
VerifyEvals: the draft against the live versionEvals with an LLM judge, on a branch or main
Run a querysql_query, one read-only statement on your databasegetData or askOmni through Omni’s AI, or runQuery with a JSON query
Charts and dashboardsrender_chart and render_dashboard, in the AI clientshowApp shows a saved Omni app in the AI client. askOmni creates dashboards in Omni
Who can connectUnlimited users, your customers’ users includedOmni users. New queries need Restricted Querier or higher. Embed users through a chat agent that you build
Needed firstA data source: your database, or the sample dataA connected database and an Omni model with topics
Databases and warehouses7, plus PostHog15
Tools for the agentUp to 1734
How you payFor each call, no seatsUser licences and AI credits. Licence prices are not published
Free plan3,000 calls a monthA free trial on request

When Omni is stronger

  • You want the semantic layer enforced: through Omni’s MCP server, each query selects fields from the model and no tool runs raw SQL. Bonnard gives the agent the metric’s SQL and does not rewrite the query.
  • You review context changes in git: Omni syncs a model to Azure DevOps, Bitbucket, GitHub or GitLab, and a setting requires a pull request for each model change (git).
  • You want a correction in a chat to become a model change: Learn from conversation drafts the YAML for a Querier or higher to review.
  • You want the Omni Agent in the AI client: askOmni creates dashboards and routines in Omni. A routine delivers a scheduled prompt by email or Slack (tools).
  • You need a dashboard as a file: startDashboardDownload starts a PNG or PDF export, and fetchDashboardDownload returns the file.
  • You build custom apps in Omni: showApp renders a saved app with live data in Claude Desktop, ChatGPT and VS Code with Copilot.
  • Your database is one Bonnard does not connect to: Omni lists 15 (databases).

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

Omni. An Organization Admin sets the AI settings to on (MCP docs). In Claude, add the Omni Analytics connector (Claude guide). In ChatGPT, add the Omni Analytics app (ChatGPT guide). Each person completes the setup. Other clients use this URL, and each person signs in through OAuth.

https://callbacks.omniapp.co/callback/mcp

A connection with an API key uses https://<your-instance>/mcp/https instead (authentication).

Verified 16 Nov 2026 against Omni’s MCP docs, tools, MCP settings, authentication, Claude guide, ChatGPT guide, AI Hub, Analytics, Suggestions, Learn from conversation, Evals, CLI, AI optimization, global context, ai_settings, Branch Mode, git, AI usage, model providers, licence types, embed guide, embed MCP, topics, apps, databases, free trial and terms.

Get started

Start scaling accurate data today

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

Keep reading