Local MCP capabilities provider for Expo
GrindFree Agent Security is installed from its publisher's own source and answers where it runs, so this marketplace is not in the path of a single call. There is no address here to send one to, and a panel that pretended otherwise would be showing you an answer we made up. Install it and call it from your own client — the Installation tab has the entry for each one.
What it does
Pay-per-call agent security: prompt injection, URLs, tools, shell/SQL, memory, leaks, docs (x402)
Quickstart
# 1 — add it to Claude Code
claude mcp add --scope user --env 'X402_PRIVATE_KEY=${X402_PRIVATE_KEY}' --transport stdio grindfree-agent-security-mcp '--' npx -y @grindfree/agent-security-mcp
# 2 — ask your agent for something these tools do
# promptshield_screen, promptshield_clean, crawlsafe_check and 14 more
It needs X402_PRIVATE_KEY: this reads it from your environment, so export it first.
GrindFree Agent Security costs nothing on mcprush: there is no plan to choose here, no cap to set and nothing mcprush can bill you for.
Collected from a public index. Nobody has claimed this account, so nothing here was written by its author — claim it if it is yours.
Where are you running it?
Every route below installs the same thing and ends at the same approval screen. This one runs on your machine: your client starts GrindFree Agent Security as a process under your own user, with your files and your network, so the tool surface below is what it can do to your computer rather than to a server somewhere else. Its scan result is on the Security tab — read the surface before you approve it.
This is a public server: you run it yourself and this marketplace is not in the path. You need Node.js — the current LTS (24) is the safe choice; npx comes with it. Claude Code registers it in one command; --scope user makes it available in every project. Claude Code reads X402_PRIVATE_KEY (Caller's own Base USDC EVM private key (0x…). Used only by the local MCP host to settle x402 payments. Never a GrindFree key; never commit this value) from your environment each time it starts the server, so the value stays out of its config: export it in your shell profile (macOS, Linux), or run setx X402_PRIVATE_KEY … and open a new terminal (Windows). The '--' is in quotes so that PowerShell passes it on.
claude mcp add --scope user --env 'X402_PRIVATE_KEY=${X402_PRIVATE_KEY}' --transport stdio grindfree-agent-security-mcp '--' npx -y @grindfree/agent-security-mcpReconnect, or start a new session, and the tools appear in the model’s tool list.
It reaches a system of yours, so it needs your own credential rather than the publisher’s: X402_PRIVATE_KEY — Caller's own Base USDC EVM private key (0x…). Used only by the local MCP host to settle x402 payments. Never a GrindFree key; never commit this value.. You set it in your own client’s config, on your machine — this marketplace never holds it.
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.
17 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.
Read from its source code. A server that builds its tool list when it starts can list more tools than this.
Screen untrusted text/HTML for prompt-injection before the model reads it. Paid x402 $0.002 USDC on Base. Returns risk, findings, sanitized. On high risk, prefer sanitized or abort.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| content | string | yes | Untrusted text or HTML to screen |
| format | string | no | Content format; default text |
Convert HTML to clean LLM-ready text (strips scripts/styles/hidden/zero-width). Paid x402 $0.002 USDC on Base.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| content | string | yes | HTML or text to clean |
Check whether a URL may be crawled for your user-agent (robots.txt + X-Robots-Tag/meta). SSRF-guarded. Not a malware scanner. Paid x402 $0.002 USDC on Base.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| url | string | yes | http(s) URL to check |
| user_agent | string | no | Crawler UA; first token matched against robots groups |
| check_page | boolean | no | Also fetch up to 64KB for X-Robots-Tag/meta (default true) |
Pre-tool-call policy gate for HTTP/MCP outbound actions. Returns allow|review|deny with deterministic reasons. Caller MUST enforce. Not malware scanning or an exfil guarantee. Missing policy → review. Paid x402 $0.002 USDC on Base.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| tool | string | yes | Tool name, e.g. http.fetch or mcp.call |
| destination | object | no | — |
| arg_classes | array | no | — |
| taint | array | no | — |
| secret_labels | array | no | — |
| policy | object | no | Owner policy; absent → review |
| policy_hash | string | no | — |
| rule_pack | string | no | — |
Check an agent-generated shell command (bash/sh/zsh/fish/PowerShell/cmd) BEFORE executing it. Returns allow|warn|block with rule codes (rm -rf, curl|sh, force-push to protected branches, credential exfil, cloud/infra deletes, disk wipes). Caller MUST enforce: do not run on block. Paid x402 $0.001…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| command | string | yes | Exact command line the agent intends to run |
| shell | string | no | — |
| policy | object | no | Optional owner policy, e.g. {mode:'strict', allow_rules:[...]} |
Check agent-generated SQL BEFORE executing it: DROP/TRUNCATE, DELETE/UPDATE without WHERE, tautological WHERE, GRANT/role changes, file/command execution, writes in read_only mode, protected tables. Returns allow|warn|block. Caller MUST enforce. Paid x402 $0.001 USDC on Base.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| sql | string | yes | One or more SQL statements |
| dialect | string | no | — |
| policy | object | no | e.g. {read_only:true} or {protected_tables:['users'], allow_write_tables:['events']} |
Screen content BEFORE writing it to an agent's long-term memory: planted instructions, persistence/trigger attacks ('from now on…', 'whenever the user asks…'), exfil setups, secrets and PII. Source-aware (user vs tool/web/email). Returns allow|review|block plus a sanitized copy. Paid x402 $0.002…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| content | string | no | Single memory entry (or use entries) |
| source | string | no | — |
| memory_type | string | no | — |
| entries | array | no | — |
| policy | object | no | Optional owner policy, e.g. {mode:'strict', allow_rules:[...]} |
Scan agent OUTPUT before it is rendered or sent: zero-click markdown-image exfil, data smuggled in URLs/DNS labels, deceptive links, leaked canary tokens, system-prompt leakage, hidden Unicode, active HTML, secrets and PII. Returns allow|review|block plus a sanitized copy. Paid x402 $0.002 USDC on…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| content | string | yes | Agent output about to be shown or sent |
| format | string | no | — |
| allowed_domains | array | no | Hosts allowed to auto-load images |
| canaries | array | no | Canary tokens planted in your system prompt |
| system_prompt | string | no | Optional; used only for verbatim-overlap detection, never stored |
| policy | object | no | Optional owner policy, e.g. {mode:'strict', allow_rules:[...]} |
Scan a document BEFORE the agent reads it for hidden instructions: PDF (invisible/white/microscopic/off-page text, annotations, metadata, JavaScript), DOCX/XLSX/PPTX (hidden runs/sheets/rows/slides, speaker notes, comments, remote templates, DDE), HTML, Markdown. Returns allow|review|block and…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| path | string | no | Local file path (read by this MCP server on your machine) |
| content_base64 | string | no | File bytes, base64 (alternative to path) |
| content | string | no | HTML/Markdown/text (alternative to path) |
| format | string | no | — |
| filename | string | no | — |
| include_text | boolean | no | Return visible-only safe_text |
| policy | object | no | Optional owner policy, e.g. {mode:'strict', allow_rules:[...]} |
Return machine-readable discovery URLs and unpaid 402 challenge shapes for all GrindFree agent-security APIs (no payment).
No parameter schema on file.
No parameter schema on file.
No parameter schema on file.
No parameter schema on file.
No parameter schema on file.
No parameter schema on file.
No parameter schema on file.
No parameter schema on file.
- Every tool, no call limit
- No card, and mcprush charges nothing for it
- Source published under MIT
- Runs on your machine — nothing of it reaches our gateway
- Nothing to cap, because mcprush bills nothing
What counts against your monthly calls
| Tool | Unit | Calls used | Out of the allowance |
|---|
Nothing here is billable: mcprush charges nothing to install GrindFree Agent Security or to call it, at any volume.
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.
| Term | Level | What it prices | Points |
|---|---|---|---|
| capability-exposure | critical | capability blast radius (critical) — client exposure if the model is manipulated | −10 |
| verification-discount | repo | publisher verification (public source) — no provenance, but the source is public and inspectable | −1 |
| coverage-honesty | source | inspection depth (source) — how much of the target the scan could see | −0 |
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.
Tools were statically extracted from the published source (17 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. critical is the level these add up to.
| Tool | Capability tags | Why the tag was assigned |
|---|---|---|
| promptshield_screen | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| promptshield_clean | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| crawlsafe_check | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| toolguard_evaluate | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| commandguard_command | code-execexternal-sink | code-exec: parameter "command"; external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| commandguard_sql | sensitive-sourceexternal-sink | sensitive-source: parameter "sql"; external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| memoryguard_screen | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| wireshield_scan | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| docshield_scan | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| saferead_read | untrusted-inputexternal-sink | untrusted-input: keyword "fetch", parameter "url" with "fetch" in the description; external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| saferead_read_batch | untrusted-inputexternal-sink | untrusted-input: keyword "fetch", parameter "url" with "fetch" in the description; external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| docread_convert | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| docread_tables | external-sink | external-sink: handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| trifectagate_check | external-sink | external-sink: keyword "http_request", handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| askhuman_request | external-sink | external-sink: keyword "webhook", handler sends an HTTP POST (fetch(url) at src/index.js:50 |
| askhuman_status | no tags | |
| grindfree_discovery | external-sink | external-sink: handler sends an HTTP POST (fetch(ep.url) at src/index.js:438 |
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.
Supply chain and provenance
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.
Release history
Always latest — the install command names no version, so it installs the newest release when it runs; the current release on file is 0.6.0. Pin a release below and the command asks for exactly that one.
No release note was published with this version.
No release note was published with this version.
No release note was published with this version.
- The version this listing was on when mcprush began following its releases. The registry was not asked when it shipped, so this row carries no date.
One review per account, from a signed-in account that does not publish this listing. Nothing else is asked — you can say what you think before you install it. Publishers can reply once per review, and a review can be edited or deleted by whoever wrote it.
Any signed-in account that does not publish GrindFree Agent Security can review it, once — before installing it or after. A review from an account that has installed it is marked as one.
Nobody has reviewed this listing. The rating on the card is the mean of the reviews written here and nothing else, so there is no rating until somebody writes the first — which takes a signed-in account that does not publish it, and nothing else.