Name: Systemores AI Agent & MCP Security Auditor
Author: bobcar0

Systemores AI Agent & MCP Security Auditor

AI agent security, MCP 2026 schema & graph audits, <250ms guardrails, and compliance dossiers

Grade A Remote or local · Streamable HTTP · stdio No auth needed 12 tools v1.1.3 Apache-2.0

No reviews yet by bobcar0npm Runs on your machine or at its publisher’s address — not through us

100/100

trust score

tools declared

local

or remote

—

downloads a day

What it does

AI agent security, MCP 2026 schema & graph audits, <250ms guardrails, and compliance dossiers

Quickstart

# 1 — add it to Claude Code
claude mcp add --scope user systemores-ai-agent-mcp-security-auditor '--' npx -y @systemores/systemores-plugin

# 2 — ask your agent for something these tools do
#   systemores_stress_test_agent, systemores_evaluate_drift, systemores_assess_bias and 9 more

Systemores AI Agent & MCP Security Auditor costs nothing on mcprush: there is no plan to choose here, no cap to set and nothing mcprush can bill you for.

Publishernot claimed

bobcar01 server

1 category

All 1 server

Collected from a public index. Nobody has claimed this account, so nothing here was written by its author — claim it if it is yours.

Installation

Where are you running it?

Every route below installs the same thing and ends at the same approval screen. This one can run either way — at an address its publisher runs, or as a process under your own user, with your files and your network — and this marketplace is in the path of neither.

Claude Codedeclared by this listing

This is a public server: you run it yourself and this marketplace is not in the path. You need Node.js — the current LTS (24) is the safe choice; npx comes with it. Claude Code registers it in one command; --scope user makes it available in every project. The '--' is in quotes so that PowerShell passes it on.

claude mcp add --scope user systemores-ai-agent-mcp-security-auditor '--' npx -y @systemores/systemores-plugin

Reconnect, or start a new session, and the tools appear in the model’s tool list.

Or connect to its publisher’s address

Its publisher also runs it at https://audit.systemores.com/api/mcp: nothing to start on your machine, and this marketplace is not in the path.

claude mcp add --scope user --transport http systemores-ai-agent-mcp-security-auditor https://audit.systemores.com/api/mcp

Free on mcprush, and it runs from its publisher’s own package: this marketplace is not in the path, so there is no key of ours to hold and no card to add.

What it asks of you

Nothing is required, but it can take SYSTEMORES_API_KEY (Optional Pro/Team API key (3 static security auditors run free without a key)) if you want to set it.

What it writes

One config entry your client uses to start the process locally. A local server runs with your file system and your network, which is why it is priced without metering.

What you approve

12 tools, with what each one reads, writes and reaches shown before you agree — the same list on every route above. Read the tool surface.

Tool surface

What the model actually sees. Descriptions are diffed on every release — see version history.

Only tools that write or reach the network

Read from its source code. A server that builds its tool list when it starts can list more tools than this.

Exposes the target agent persona to a battery of adversarial prompts (Presses) and evaluates behavioral stability using Murray's Needs-Press diagnostic model.

Parameter	Type	Required	Notes
systemPrompt	string	yes	The behavioral blueprint or internal operational instructions of the target agent. Maximum 8,000 characters.
role	string	yes	The intended operational function of the agent.
pressIntensity	string	no	The intensity level of adversarial testing. 'high' simulates prompt injection batteries.

Simulates multiple conversational turns to evaluate agent output consistency and behavioral drift over extended interactions.

Parameter	Type	Required	Notes
systemPrompt	string	yes	The target agent's system prompt. Maximum 8,000 characters.
turns	integer	no	The number of conversational turns to simulate (1–100).
temperature	number	no	The LLM temperature parameter for the target agent (0.0–1.5). Higher values increase drift probability.

Verifies the agent's compliance with fair-lending, fair-housing, employment non-discrimination (EEOC/SIOP/ECOA), or child-directed safety (COPPA) standards.

Parameter	Type	Required	Notes
systemPrompt	string	yes	The underwriting, hiring, or housing agent system prompt to evaluate. Maximum 8,000 characters.
complianceDomain	string	yes	The legal compliance framework to evaluate against (employment, housing, lending, or COPPA child safety).

Measures operational latency, response length, and execution drag parameters to analyze cognitive throughput stability within the Behavioral OS ecosystem.

Parameter	Type	Required	Notes
promptLength	integer	yes	The character length of the prompt sent to the agent.
responseLength	integer	yes	The character length of the agent's response.
latencyMs	number	yes	The end-to-end response latency in milliseconds. Must be >= 1.

Evaluates whether the agent's system prompt establishes the behavioral conditions for safe, reliable tool use — auditing tool selection discipline, argument construction fidelity, and failure resilience when tool returns are ambiguous or erroneous.

Parameter	Type	Required	Notes
systemPrompt	string	yes	The agent's system prompt to evaluate for tool use safety. Maximum 8,000 characters.
toolContext	string	no	The domain of tool usage being audited. Affects risk weighting.

Audits the structural integrity, naming conventions, schema completeness, safety annotations, and injection surface of an AI agent's exposed tool definitions. Identifies malformed schemas, privilege overreach, deceptive safety hints, duplicate names, and description quality gaps. Operates entirely…

Parameter	Type	Required	Notes
toolDefinitionsJson	string	yes	A JSON-serialized array of tool definition objects to audit. Each object should contain at minimum 'name' and 'description' fields. Accepts standard MCP ListToolsResult format.
strictMode	boolean	no	When true, applies stricter scoring thresholds — missing optional fields such as parameter descriptions, inputSchema, and MCP tool annotations are penalized. Defaults to false.

Statically scans multi-agent execution topologies (DAGs/meshes) using DFS algorithms to detect cyclic delegation loops and privilege escalation paths from public untrusted sub-agents to backend write/admin tools.

Parameter	Type	Required	Notes
graphTopologyJson	string	yes	JSON-serialized graph topology object containing 'nodes' (id, privilegeLevel) and 'edges' (source, target).
maxDepth	integer	no	Maximum depth to traverse during cycle and privilege path search.
strictPrivilegeCheck	boolean	no	When true, flags any path from 'public'/'user' nodes to 'admin'/'system' nodes as a breach.

Generates statistical cohorts of synthetic user personas across enterprise operational domains with IPIP-NEO-120 30-facet psychometric profiles, demographic diversity, and prompt injection payloads.

Parameter	Type	Required	Notes
domain	string	yes	The operational domain for persona generation.
cohortSize	integer	no	Number of synthetic personas to generate (1–100).
includeAdversarial	boolean	no	Includes prompt-injection and noisy data submitter personas in the cohort.

Configures dynamic chaos injection policies for tools, mock APIs, and databases during simulations (HTTP 429 rate limits, 500 errors, timeouts, 401 unauthorized, malformed JSON).

Parameter	Type	Required	Notes
faultType	string	yes	The specific chaos fault to inject.
targetToolName	string	yes	The target tool name to attach fault injection policy to (or 'all_tools').
failureProbability	number	no	Probability of fault triggering per invocation (0.01 to 1.0).

Ingests standard OpenTelemetry (OTel) trace JSON payloads to detect financial token cost burn traps (> $2.50 threshold), runtime loop deadlocks between sub-agents, and cognitive latency drag hotspots (> 3000ms).

Parameter	Type	Required	Notes
otelTraceJson	string	yes	JSON string containing OpenTelemetry trace spans or MCP telemetry log data.
maxCostThresholdUsd	number	no	Maximum allowable cost in USD per session before triggering burn trap alert.
latencyThresholdMs	integer	no	Latency hotspot threshold in milliseconds.

Aggregates findings from graph audits, synthetic cohort simulations, chaos tests, and trace analytics into formal compliance report packages for EU AI Act (EU 2024/1689), NIST AI RMF 1.0, EEOC/ECOA, and SOC 2 AI Addendums.

Parameter	Type	Required	Notes
auditSessionId	string	yes	Unique identifier for the completed audit session.
targetFrameworks	string	no	Comma-separated list of regulatory frameworks to generate formal dossier sections for: 'eu_ai_act', 'nist_ai_rmf', 'eeoc_ecoa', 'soc2_ai'. Defaults to all if omitted.
companyName	string	no	Legal entity or client name for dossier attribution.

Performs real-time (~250ms) operational and behavioral turn auditing using Systemores System One. Computes policy compliance score, sycophancy probability, unauthorized tool scope, and returns an inline circuit-breaker action (DISPATCH, INTERCEPT, QUARANTINE, STRIP_TOOL).

Parameter	Type	Required	Notes
systemPrompt	string	yes	The behavioral blueprint, constraints, or security policies of the agent under audit.
userMessage	string	yes	The prompt or message sent by the user in this interaction turn.
agentResponse	string	no	The proposed text response or output from the agent.
proposedToolCall	string	no	Optional tool invocation or function call proposed by the agent (name and arguments).

Free

$0 forever

Every tool, no call limit
No card, and mcprush charges nothing for it
Source published under Apache-2.0
Runs on your machine — nothing of it reaches our gateway
Nothing to cap, because mcprush bills nothing

Nothing here is billable: mcprush charges nothing to install Systemores AI Agent & MCP Security Auditor or to call it, at any volume.

ATrust score 100 / 100

Release 1.1.3 · scanned 13 Nov 2026

Two independent axes, because powerful and malicious are different questions. The grade is threat only. The capability level is blast radius, and it is never a penalty on the grade — it is priced as one subtract-only term in the score, where you can see it.

Threat score100

Client score100

Findings0 threat · 0 capability

Capabilityminimal

Coveragesource

Verificationsource

Surface digest7d9318a8a0f334527066cb05d34d37a4a16f53529a31d52b419cab857837b8bc

Score vectorthreat 100 → client 100

Term	Level	What it prices	Points
capability-exposure	minimal	capability blast radius (minimal) — client exposure if the model is manipulated	−0
verification-discount	source	publisher verification (provenance) — cryptographic build provenance ties the artifact to its source	−0
coverage-honesty	source	inspection depth (source) — how much of the target the scan could see	−0

score = clamp(round(threatScore − E_cap − E_ver − E_cov), 0, 100). No model, no opinion, no network call at scoring time: the same methodology version against the same bytes always produces this number, and every line above is a field of the report rather than a summary of it.

What the scan could actually read

A grade is only as meaningful as its coverage, so the scanner publishes its own depth before it publishes its result.

Tool surfaceread

Implementation sourceread

Package metadataread

Live transportnot available

Tools were statically extracted from the published source (12 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command "npx -y <package>".

Capability — what it could do if the model were manipulated

Tags derived from each tool’s schema and the implementation, not from what the tool calls itself. minimal is the level these add up to.

Tool	Capability tags	Why the tag was assigned
systemores_stress_test_agent	untrusted-input	untrusted-input: declares openWorldHint: true (interacts with entities outside the server)
systemores_evaluate_drift	untrusted-input	untrusted-input: declares openWorldHint: true (interacts with entities outside the server)
systemores_assess_bias	untrusted-input	untrusted-input: declares openWorldHint: true (interacts with entities outside the server)
systemores_measure_drag	no tags
systemores_tool_integrity	untrusted-input	untrusted-input: declares openWorldHint: true (interacts with entities outside the server)
systemores_audit_tool_integrity	no tags
systemores_audit_agent_graph	no tags
systemores_generate_cohort	untrusted-input	untrusted-input: declares openWorldHint: true (interacts with entities outside the server)
systemores_inject_environment_fault	no tags
systemores_analyze_trace_telemetry	no tags
systemores_export_dossier	no tags
systemores_audit_turn_live	untrusted-input	untrusted-input: declares openWorldHint: true (interacts with entities outside the server)

Toxic-flow graph

The lethal trifecta, checked as a graph rather than as a checklist: untrusted input, a sensitive source and an external sink have to meet before there is a path worth worrying about.

The public result for this release does not print the flow graph, so there is nothing to show here. That is not the same as "no paths were found": what the scan did read is above, under coverage.

Supply chain and provenance

Integrityfirst-seen

Surface digest7d9318a8a0f33452

Scanned2026-11-13

This is the first scan of this surface here, so there is nothing yet to compare it against.

Every result on this tab comes from one deterministic pass over the published package — offline, rule by rule, and auditable line by line above. Same methodology version, same bytes, same score.