The server itself exposes 18 tools, and some of them can write to your Hex account. Behind TrustGate an agent only sees the tools you allow, requests and responses are inspected, and every call is logged. Anything outside the allowlist is denied before it reaches Hex.

How is this different from connecting Hex MCP directly?

Connected directly, an agent sees every tool the server exposes and holds the credential itself. Through TrustGate you choose which tools each agent gets, and every call is logged against the agent that made it.

What if the Hex MCP API changes?

We track the Hex MCP schema, so the tool catalogue stays current. Tools you have already allowed keep working, and anything new stays denied until you enable it.

Can I give different agents different Hex tools?

Yes. Access is set per agent: one can get a handful of read tools, another the full set. You grant the whole MCP or single tools within it.

Where can I see what my agents did in Hex?

Every call is logged against the agent that made it, with the tool it used. Activity can be sent on to your own monitoring.

Do I need to run the Hex MCP server myself?

No. TrustGate connects to the Hex MCP server and sits in front of it. Your agent only talks to your TrustGate URL, never to Hex directly.

Connect Hex MCP to your AI agents, securely.

One secured endpoint for every Hex call your AI agents make, scoped per user, inspected, and logged.

Get started for free

Get a demo

Try Hex MCP now

What your agents can ask once Hex MCP is connected.

Select the tools your AI agents can call

18 tools from the Hex MCP schema.

Showing 18 tools

search_projectsreadFinds projects in the Hex workspace, returning both published and unpublished ones with their ID, title, description, AI-generated summary, URL and type. Requires the Explorer role or above.
get_threadreadRetrieves the messages and results from a Thread. If it has not finished, it returns the status and a wait time, and is designed to be called repeatedly until the Thread goes idle.
get_mereadReturns the authenticated user ID, name, email, workspace role and workspace ID, confirming which workspace and role the agent is operating under.
create_threadwriteStarts a new Thread conversation so the Hex Agent can answer a data question. It selects among the data connections the user can reach, excluding any marked Sensitive unless an Admin has allowed them, and typically takes several minutes.
continue_threadwriteAdds a follow-up prompt to an existing Thread. A Thread can only be continued once it has finished running.
get_projectreadReturns a project's metadata: creator and owner, created, last edited and last published timestamps, categories, status, reviews, schedules, app view analytics, and optionally its user, group, collection, workspace and public web sharing.
list_cellsreadLists the cells in a project's draft version with their ID, static ID, type, label, attached data connection and source. The list is flat and in notebook order, and does not report which cells sit inside a section.
get_cellreadReturns a single cell by its ID, with the same fields list_cells provides.
create_projectwriteCreates a new, empty project in the workspace and returns its ID. Requires the Editor role or above.
create_cellwriteAdds a SQL, Python or Markdown cell to a project's draft, optionally with a label and a position, including inside a section. SQL cells take either a data connection ID or a flag to query other SQL cells' results.
update_cellhigh riskUpdates a cell's source, and for SQL cells its output dataframe name and data connection. The change lands on the live draft immediately with no staging step. A cell's label cannot be changed after creation.
delete_cellhigh riskRemoves a cell from a project's draft version, directly in the notebook collaborators are working in.
get_runreadChecks a project run's status, URL, trigger, start and end times and elapsed milliseconds. The run is finished at COMPLETED, ERRORED, KILLED or UNABLE_TO_ALLOCATE_KERNEL.
get_cell_outputreadReads a SQL cell's latest output preview or its error, reporting the row limit, rows returned, total row count and whether the result was truncated. It is blocked for projects using sensitive connections unless an Admin has allowed them.
get_cell_imagereadReturns a chart cell rendered as a PNG, optionally at a given width and height between 100 and 2000 pixels. The cell must already have been executed and must not be in an error state.
run_cellhigh riskRuns a single cell and its upstream dependencies in the draft version, executing SQL or Python against real data connections. The run is asynchronous and a dry run validates without executing.
run_notebookhigh riskRuns the entire draft version of a project asynchronously. SQL cells reuse cached results by default, and caching can be disabled to force fresh queries against the warehouse.
list_data_connectionsreadLists the workspace's data connections with their ID, name, type and description, so a connection can be resolved by name before creating or updating a SQL cell.

What Hex MCP exposes

Hex MCP groups its 18 tools into 4 categories, each with a suggested starting policy.

All 18 tools as JSON, ready to paste.

Category	Covers	Access	Suggested policy
Projects & Cells	Creating an empty project, reading a project's metadata with its creator, owner, status, reviews, schedules, analytics and optionally its sharing settings, listing the cells in a project's draft with their type, label and source, reading one cell, and adding, updating or removing SQL, Python and Markdown cells	Read / Write	Edits apply directly to the live draft with no staging step, so a colleague opening the notebook sees them immediately. Read get_project and list_cells before editing, since an update against a stale view silently overwrites someone else's work
Knowledge	Searching published and unpublished projects with their titles, descriptions, AI-generated summaries and links, starting a new Thread conversation against the workspace data, retrieving its messages, charts and tables, continuing it with a follow-up, and confirming the authenticated user, role and workspace	Read / Write	A Thread sets the Hex Agent loose on the data connections the user can reach, choosing among them itself, and takes several minutes. Connections marked Sensitive are excluded unless an Admin has allowed them, which is the control that matters here
Execution	Running one cell and its upstream dependencies or the whole draft notebook asynchronously, polling a run's status and timings, reading a SQL cell's latest output preview with its row limits and truncation flags, and retrieving a chart cell rendered as a PNG	Read / Write	run_cell and run_notebook execute SQL and Python against real data connections, so this is code execution rather than a query. The runId identifies the draft session rather than one execution, so check the reported start time before trusting that a run finished
Data Connections	Listing the workspace's data connections with their ID, name, type and description, and depending on access their connection details, sharing settings and schema filter and refresh settings	Read	Read-only, and the prerequisite for pointing a new SQL cell at the right warehouse. What it returns describes the organization's data estate, so treat the output as internal

Connect Hex MCP to your AI agent

Four steps, under ten minutes.

01
Sign in to NeuralTrust
Open the NeuralTrust platform, sign in and go to Getting started, then Connect your tools.
02
Choose your agent
When asked which agent you use, pick Claude Code, Codex, Cursor, Windsurf, Copilot or your own code.
03
Connect Hex
Find Hex in the tool list and connect it.
04
Connect your agent
Follow the steps on screen to add the config to your agent, then send the test prompt.

Permissions down to the single tool

You do not allow Hex MCP or block it. You allow the exact tools each agent can call, one by one.

Hex MCP	18 TOOLS AVAILABLE
Code review agent	4 tools, read only
Issue triage bot	9 tools
Platform team assistant	all 18 tools

Tool by toolConnecting the server is not the unit of access. Each tool is granted on its own.
Default denyA tool you have not enabled is never exposed to the agent.
Approval on writesTools that change or delete things can require a human to confirm.

Is Hex MCP secure?

One endpoint instead of raw credentials. Every call is checked against policy and logged.

AI agentsConnect through your MCP client configuration.	NeuralTrustApply controls based on your configuration.	Hex MCP18 documented tools at app.hex.tech/mcp over OAuth, with custom URLs for single-tenant, EU and HIPAA deployments. Currently in beta, on Team and Enterprise plans. Knowledge tools need the Explorer role, project editing needs Editor, and per-project permissions are enforced on top. MCP access is not covered by the workspace API access setting; only deactivating the user revokes it.
Coding assistants	Connection & credentialsper user	Projects & Cells7 tools
Internal copilots	Tool access policiesdeny by default	Knowledge5 tools
Workflow automations	Request & response inspectioninline	Execution5 tools
Compatible MCP clients	Activity monitoringalerts and logs	Data Connections1 tool