Comparisons
Bonnard vs Hex: MCP server features compared
Choose Hex when analysts want an agent that works inside their notebooks and builds the analysis with them. Choose Bonnard when everyone else, customers included, wants an answer in the chat they already use, with no seat to buy for each of them.
On this page
With Hex (hex.tech), Claude or ChatGPT passes the question to Hex’s own agent, which answers from Hex’s context. With Bonnard (bonnard.ai), Claude or ChatGPT reads the context and writes the query.
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.
Hex keeps its context for Hex’s own agent. Context Studio holds guides, endorsed data and semantic models, on the Team and Enterprise plans (Context Studio). You write semantic models in Hex or sync them from dbt MetricFlow, Cube or Snowflake Semantic Views (Semantic Model Sync). A Hex Thread puts semantic models and endorsed data first (Threads). In Governed mode, an admin can keep the agent to semantic models only, so it does not query warehouse tables directly. Hex describes Governed mode as a strong steer and not a security boundary (Governed mode).
None of Hex’s 18 MCP tools reads a guide, an endorsement or a semantic model (Hex’s MCP docs). Claude or ChatGPT gets Hex’s context by asking Hex’s agent with create_thread. Hex’s docs state that Threads “typically take several minutes to complete”.
How wrong answers get found and fixed
Bonnard runs a loop of four steps on how agents use your data. Hex has the same four steps for its own agent, on Team and Enterprise.
- 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.
Hex’s Observability page marks conversations with missing context, user doubt or a data limit, and filters them by source, MCP included.
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
Hex’s Suggestions come from patterns across agent conversations. They propose changes to guides, rules, descriptions and endorsements.
Improve. Your data team’s answer goes into a draft of the context.
On Hex, an admin accepts or rejects each suggestion. Guides and semantic models keep a version history.
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
Hex’s Evals grade the agent’s answers against rubrics you write, with an LLM as the judge, and can run on a preview of a context change. You define and run them from Hex’s 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.
Hex’s loop improves the answers of Hex’s agent. 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.
Hex starts from a Hex workspace on Team or Enterprise, with a data connection in it. MCP Threads select among the data connections that the user can access, and do not query connections marked Sensitive unless an admin allows it. The context is as good as the semantic models, endorsements and guides your team has set up in Hex.
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 Hex, each person signs in through OAuth as their own Hex user, at the Explorer role or higher. The MCP server enforces the user’s project permissions. Hex’s MCP server has no route for your customers’ users.
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).
Hex’s MCP server is on the Team and Enterprise plans only. Team is USD 75 for each Editor a month, with a 14-day free trial, and Enterprise prices are custom (Hex pricing). Explorers, the lowest role that can use MCP, are paid seats too (roles). Agent runs from MCP use Hex credits. Team gives 40 credits a month for each Editor and 10 for each Explorer (credits). Without credits, users cannot send more agent prompts, and an Admin can set automatic top-ups of 50 credits for USD 25.
Side by side
| Feature | Bonnard | Hex |
|---|---|---|
| Who answers | The AI client the person uses, with Bonnard’s context | Hex’s agent, through create_thread |
| Context the agent reads | Schema, descriptions, joins, metrics, business terms, searchable values | Guides, endorsed data, semantic models |
| Context through MCP | explore_schema, for any AI client | No tool reads it |
| Metric definitions | The agent reads the SQL and writes the query | Put first by Threads. Governed mode can keep the agent to semantic models only |
| Observe | Records each tool call, from any AI client | Observability, with MCP as a source |
| Diagnose | Finds what is wrong in the activity record | Suggestions from agent conversations |
| Improve | The data team’s answer goes into a draft | An admin accepts or rejects each suggestion |
| Verify | Evals: the draft against the live version | Evals with an LLM judge, from the CLI |
| Run a query | sql_query, one read-only statement on your database | create_thread through Hex’s agent, or run_cell for Editors |
| Charts and dashboards | render_chart and render_dashboard, in the AI client | Thread charts in a widget in Claude |
| Who can connect | Unlimited users, your customers’ users included | People with a Hex account, Explorer role or higher |
| Needed first | A data source: your database, or the sample data | A Hex workspace with a data connection |
| Databases and warehouses | 7, plus PostHog | 16 |
| Tools for the agent | Up to 17 | 18, of which 13 need the Editor role or higher |
| How you pay | For each call, no seats | For each Editor and Explorer seat, plus credits for agent runs |
| Free plan | 3,000 calls a month | Community is free and has no MCP. Team has a 14-day trial |
When Hex is stronger
- You want the semantic layer enforced: in Governed mode, Hex’s agent answers from semantic models only. Bonnard gives the agent the metric’s SQL and does not rewrite the query.
- You already have a semantic layer: Hex syncs models from dbt MetricFlow, Cube and Snowflake Semantic Views.
- You keep context in git: Context Sync brings guides, semantic models and evals from a repository through Hex’s CLI or a GitHub Action (Context Sync).
- You want agents to build notebooks: Editors write SQL, Python and Markdown cells into Hex projects and run them with
run_cellorrun_notebook. - You want agents to build and publish Hex apps: in Hex 4.0 (public beta), agents create, edit and publish projects through MCP (Hex 4.0).
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/mcpHex. The MCP server is on the Team and Enterprise plans. In Claude, select Settings > Connectors > Browse Connectors and search for Hex. On a Claude Team or Enterprise plan, an owner adds the connector. On an individual Claude plan, the user adds it. Cursor, ChatGPT and Codex have a Hex plugin or app. Other clients, and workspaces on another Hex host, use this URL.
https://app.hex.tech/mcpSingle-tenant, EU and HIPAA workspaces replace app.hex.tech with their own Hex URL, for example eu.hex.tech.
Verified 16 Nov 2026 against Hex’s MCP docs, pricing, roles, credits, Threads, Context Studio, Semantic Model Sync, Governed mode, observability, Suggestions, Evals, Context Sync, Hex 4.0 and data connections.
Get started
Start scaling accurate data today
Free for your first 3,000 calls. No seat limits.
