AdversaryGate runs on your own machine — the client starts it, so there is no endpoint to ping. 1 085 installs a week from pypi. Last commit 28 Oct 2026.
Fail-closed merge gate an agent calls before saying done: MERGE, BLOCK or INCONCLUSIVE.
We read the source, 5 d ago · rules 995398eba69f
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
proc = subprocess.run(
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
We found places where it runs commands, builds paths or queries from values it is given. None of that is a flaw by itself — it becomes one when the code changes, and code changes quietly between releases. We re-read it on every one.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add adversary-gate -- uvx adversary-gate-mcp
{
"mcpServers": {
"adversary-gate": {
"command": "uvx",
"args": [
"adversary-gate-mcp"
]
}
}
}
[mcp_servers.adversary-gate]
command = "uvx"
args = ["adversary-gate-mcp"]
{
"mcpServers": {
"adversary-gate": {
"command": "uvx",
"args": [
"adversary-gate-mcp"
]
}
}
}
{
"servers": {
"adversary-gate": {
"type": "stdio",
"command": "uvx",
"args": [
"adversary-gate-mcp"
]
}
}
}
Optional settings from the registry entry, left out of the block above:
ADVERSARY_GATE_BASE_REF (Baseline git ref (e.g. origin/main). Unset, the agent may pick it, and whoever picks the baseline picks the oracle), ADVERSARY_GATE_REQUIRE_PINNED_BASE (Set to 1 to refuse any run without ADVERSARY_GATE_BASE_REF, so the agent can never choose the baseline (the oracle)), ADVERSARY_GATE_PYTHON (Interpreter that runs the tests; must be one the agent cannot write to), ADVERSARY_GATE_POLICY (Extra gate flags set by the operator; evidence flags are refused).
Fail-closed verify-before-you-act gate for AI agents. Signed receipts. Pay-per-call via x402.
Fail-closed Cedar policy gate + Ed25519 signed receipts for agent tool calls. Denies on any error.
Security gateway for MCP agents: blocks prompt-injection-driven tool calls before they execute.
Decision-theoretic triage and fail-closed agent gating: 32 tools over stdio.
Score a planned spend before an agent pays. Verdicts: approve, ask_human, or block.
Verify the facts an agent's action depends on before it acts — ALLOW, REVIEW, or BLOCK.
Verify the facts an agent's action depends on before it acts — ALLOW, REVIEW, or BLOCK.
An AI agent's trading mandate it can't break: checked before every order, reasoning sealed in IRL.
Answers built from our own checks of this server.