Run AI evaluations, record agent traces and gate releases on evidence, with scoped org access. Connect the official EvalGate MCP server to Claude, ChatGPT, or Cursor through gate — one gateway URL, no config files, security-checked.
Free to start · No credit card · No hosting needed
Pick your client. Copy, paste, done — or let gate handle sign-ins and policies for you.
claude mcp add --transport http evalgate https://www.evalgate.com/api/mcpRun in your terminal, then restart Claude Code.
EvalGate · https://www.evalgate.com/api/mcp
Real prompts you can type the moment it's connected — no setup, no docs to read.
“Use “project_inspect”: Inspect the evaluations and project keys available in the authenticated organization.”
“Use “project_plan”: Return a non-mutating next-step plan for connecting a repository, evaluation, or trace source.”
“Use “evaluations_list”: List evaluations with a stable cursor for the organization.”
“Use “runs_list”: List evaluation runs with stable cursors and regression signals.”
Every server in the gate directory has to pass the same four checks before your AI can touch it. Here's how EvalGate holds up.
Operated by EvalGate (evalgate.com) at its documented MCP endpoint — not a third-party mirror or community re-host.
This server only exposes public data — no account and no credentials are involved when you connect.
gate checks every tool for prompt injection, hidden instructions, data-exfiltration hints, and over-broad permissions before the server is available through your gateway.
If EvalGate adds or changes tools later, gate re-scans and alerts you — before your AI acts on the change.
Add one gateway URL to Claude, ChatGPT, or any MCP client. One-time setup, about two minutes.
Choose EvalGate in the gate directory. No sign-in needed.
EvalGate's tools are live in every AI client you've connected — with rules and logging built in.
Free to start · No credit card · No hosting needed
Fetched live from the official endpoint and re-checked daily by gate — not marketing copy.
project_inspect — Inspect the evaluations and project keys available in the authenticated organization.
project_plan — Return a non-mutating next-step plan for connecting a repository, evaluation, or trace source.
evaluations_list — List evaluations with a stable cursor for the organization.
runs_list — List evaluation runs with stable cursors and regression signals.
run_explain — Explain a run using its status, failed-case metadata, and evidence-lineage summary.
baseline_lineage — Show the published baseline anchor and its evidence lineage.
docs_search — Search the published OpenAPI operation catalog for relevant help.
connector_health — Show the latest connector-readiness reports for the organization.
+ 4 more tools once connected
https://www.evalgate.com/api/mcpYes — EvalGate operates its own remote MCP server at www.evalgate.com. gate connects you to that official endpoint and adds security scanning, per-tool access rules, and an activity log on top.
Add gate's gateway URL to your AI client once, then pick EvalGate in the directory. No account or sign-in is needed. From then on, EvalGate's tools are available in every client you've connected to gate.
No. EvalGate runs the server, and gate is fully managed. You just connect one URL — there's nothing to install, deploy, or maintain.
It's the vendor's official endpoint, and gate scans its tools for prompt injection and hidden instructions before it goes live, re-checks it whenever it changes, and lets you allow, block, or require approval for every single tool.
Yes — you can get started with gate for free and connect EvalGate in minutes. No credit card required.
Every server below works through the same gateway URL — connect once, add tools anytime.
Set up EvalGate in: Claude · Claude Code · ChatGPT · Cursor · VS Code