How is this different from connecting Statsig 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 Statsig MCP API changes?

We track the Statsig 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 Statsig 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 Statsig?

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 Statsig MCP server myself?

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

Connect Statsig MCP to your AI agents, securely.

Q: Is Statsig MCP secure?

The server itself exposes 27 tools, and some of them can write to your Statsig 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 Statsig.

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

Get started for free

Get a demo

Try Statsig MCP now

What your agents can ask once Statsig MCP is connected.

Select the tools your AI agents can call

27 tools from the Statsig MCP schema.

Showing 27 tools

get_contextreadGets bounded project, permission, review and product context. Scope can be session, project or all, and fields selects organization, environment, permissions, review settings, accessible products, event sources, metrics, timezone or links.
searchreadSearches Statsig resources once a resource type or target is identified.
fetchreadFetches a resource selected from a search result, by the identifier search returned.
gate_readreadLists, gets and inspects feature gates and their history: details by ID, rule results, version history and the full gate list filtered by type, creator or tags.
experiment_readreadLists, gets and inspects experiments: details by ID, overall and per-metric-dimension results, version history, summary charts and the filtered experiment list.
dynamic_config_readreadLists, gets and inspects dynamic configs and their version history.
metric_readreadLists metric sources and metrics, including hidden ones, and gets a metric definition by ID.
log_readreadQueries the Logs Explorer and inspects project audit and change history, filterable by action type and date range.
gate_createwriteCreates a feature gate with rules, tags, ID type, target apps and team. Requires write access.
experiment_createwriteCreates an experiment with allocation, layer, targeting gate, hypothesis, groups, primary and secondary metrics and scheduling. Requires write access.
dynamic_config_createwriteCreates a dynamic config with rules, default value, ID type and target apps. Requires write access.
gate_updatewriteUpdates a feature gate, including its enabled state, rules and gradual rollout strategy. Requires write access and a confirmation object, and replaces the whole resource.
experiment_updatewriteUpdates an experiment, including allocation, targeting gate, status, Bonferroni correction and confidence interval. Requires write access and a confirmation object.
dynamic_config_updatewriteUpdates a dynamic config, including its enabled state, rules and default value. Requires write access and a confirmation object.
discover_toolsreadRead-only discovery of operations beyond the named tools. Resolves a single query or up to five, lists candidates, and returns the chosen operation contract with its lane and confirmation requirements.
api_readreadRuns a discovered read-only operation from its operation handle.
api_writehigh riskRuns a discovered write operation from its operation handle. Requires write access, and the tool schema does not itself require confirmation.
api_destructivehigh riskRuns a discovered destructive operation. Requires write access and a confirmation object. Covers sensitive updates and review actions, not only deletion.
layer_readreadGets layer details by ID, the experiments in a layer, its overrides, and the list of layers. Expanded exposure only.
segment_readreadLists segments and gets one by ID. Expanded exposure only.
param_store_readreadLists parameter stores and gets one's details by ID. Expanded exposure only.
layer_createwriteCreates a layer with a name and ID type. Expanded exposure and write access.
segment_createwriteCreates a segment with a name, type and optional rules. Expanded exposure and write access.
param_store_createwriteCreates a parameter store with a name, description and display name. Expanded exposure and write access.
segment_updatewriteUpdates a segment. The request body depends on the selected operation and segment type. Expanded exposure, write access and a confirmation object.
param_store_updatewriteUpdates a parameter store's description and parameters. Expanded exposure, write access and a confirmation object.
autotune_createhigh riskCreates an autotune with variants, success event, exploration and attribution windows and a winner threshold. Expanded exposure, write access and a confirmation object. Unlike the other create tools it takes arguments.params rather than a JSON body.

What Statsig MCP exposes

Statsig MCP groups its 27 tools into 7 categories, each with a suggested starting policy.

All 27 tools as JSON, ready to paste.

Category	Covers	Access	Suggested policy
Expanded Create & Update	Creating layers, segments, parameter stores and autotunes, and updating segments and parameter stores	Read / Write	Only visible to clients using expanded tool exposure. autotune_create requires a confirmation object on top of write access
Reads	Grouped read tools for gates, experiments, dynamic configs, metrics and logs. Covers listing, details, results, version history, metric sources and definitions, audit logs and the Logs Explorer	Read	Read-only. query_fields narrows the response, which matters because experiment results are large
Discovery & Execution	Resolving and listing operations beyond the named tools, and running a resolved operation through the read, write or destructive execution lane	Read / Write	This is the widening path: discover_tools is read-only, but it hands back handles that api_write and api_destructive then execute. Treat the three execution tools as separate allowlist entries
Context & Search	Bounded project, permission, review and product context, resource search, and fetching a resource from a search result	Read	get_context is the recommended first call because it reports the agent's own permissions and the project's review settings
Create	Creating feature gates, experiments and dynamic configs	Read / Write	Require write access. In projects using Target Apps, gate_create needs a Target App name rather than an internal record ID
Update	Updating feature gates, experiments and dynamic configs	Read / Write	Most update tools replace the whole resource, so an agent that does not read current state first will silently drop fields. Partial edits should go through a discovered operation instead
Expanded Reads	Layers with their experiments and overrides, segments and parameter stores	Read	Read-only, visible only with expanded tool exposure

Connect Statsig 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 Statsig
Find Statsig 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 Statsig MCP or block it. You allow the exact tools each agent can call, one by one.

Statsig MCP	27 TOOLS AVAILABLE
Code review agent	4 tools, read only
Issue triage bot	9 tools
Platform team assistant	all 27 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 Statsig 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.	Statsig MCP18 default tools plus 9 shown only to clients using expanded tool exposure. Visibility also depends on permissions and server configuration.
Coding assistants	Connection & credentialsper user	Expanded Create & Update6 tools
Internal copilots	Tool access policiesdeny by default	Reads5 tools
Workflow automations	Request & response inspectioninline	Discovery & Execution4 tools
Compatible MCP clients	Activity monitoringalerts and logs	All other tools12 tools