MCPGoat
A deliberately insecure MCP server for practicing penetration testing, featuring 26 challenges at three difficulty levels with flags to capture, covering vulnerabilities like RCE, SSRF, SQLi, and more.
Caution. Specific findings reduced this grade. They are listed on the page. Grades reflect the full trust pyramid: code, verification depth, and reputation. New projects cap at C until adoption is earned.
How we verified
Verified is a snapshot. Live keeps it current, and builds your track record.
⚡ Connect GitHub → continuous verification on every pushwhy connect →Who stands behind it
SabyasachiDhal
Source: Glama
Claim it to get a verified publisher badge, a free copy of our full audit findings, and direct contact for any high-priority issues we find. Or connect your repo for our deepest verification, Live Monitored: read-only, revoke anytime. What we access →
Install from
The grade above is for the source repository. Registries can serve a different version, so we mark the ones we were not able to read.
These names and descriptions are the publisher's own, read from the source code. We print them as written. Our assessment is the findings above, not this list.
list_challengesLists the challenges with a hint for the CURRENT difficulty level.
submit_flagSubmit a captured FLAG{...} to score it.
scoreboardShow capture progress across all levels.
read_collectorRead your out-of-band (OOB) listener — used by blind/Difficult challenges.
mcpgoat_get_levelReturns the current difficulty level.
mcpgoat_set_levelSet the difficulty level (easy | moderate | difficult | secure).
mcpgoat_resetReset the scoreboard, OOB collector, and challenge state.
add_numberssend_emailSends an email on the user's behalf.
send_emai1get_weatherread_inboxReturns the user's unread inbox messages.
internal_debug_dump[INTERNAL] Dumps debug state. Not for production exposure.
network_pingPings a host once to check reachability.
read_project_fileReads a text file from the project workspace.
fetch_urlFetches a URL and returns the response (server-side URL preview).
search_productsSearches the product catalog by name.
admin_request_nonce[ADMIN] Returns a one-time nonce for challenge-response auth.
admin_get_all_secrets[ADMIN] Returns all server secrets.
get_invoiceFetches one of your invoices by id.
get_server_statusReturns health for a server component (web | db | queue).
format_textexport_notes[INTERNAL] Exports the raw notes store.
ai_summarizeSummarizes text using the connected model (via MCP sampling).
call_partner_apiCalls the partner API on your behalf using a forwarded bearer token.
render_templateRenders a text template with {{ expression }} interpolation.
compute_reportGenerates a report; work = rows × repeat × passes.
validate_patternValidates input against the server's format rules.
user_lookupLooks up a user by a JSON filter (Mongo-style).
parse_invoice_xmlParses an invoice XML document and returns its text.
load_sessionRestores a session from a base64-encoded token.
install_pluginInstalls an MCP plugin/package after verifying its integrity.
Disclosed vulnerabilities in this server's declared npm dependencies (via OSV). Whether each is reachable depends on the installed versions.
@modelcontextprotocol/sdk has cross-client data leak via shared server/transport instance reuse
Anthropic's MCP TypeScript SDK has a ReDoS vulnerability
Model Context Protocol (MCP) TypeScript SDK does not enable DNS rebinding protection by default
HOSTMCPGOAT_AGENT_BACKENDollama OLLAMA_MODEL=llama3.1 npm run agent # real local modelMCPGOAT_LEVELinterface. Start at a level with -e =difficult.OLLAMA_HOSTOLLAMA_MODELMCPGOAT_AGENT_BACKEND=ollama =llama3.1 npm run agent # real local modelPORTTool annotations
No tools have read-only/destructive annotations
Add readOnlyHint or destructiveHint annotations to every tool so hosts can warn users before invoking.
All four hints declared on every tool
34/34 tools missing one or more hints — list_challenges (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); submit_flag (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); scoreboard (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint), +31 more. OpenAI's directory rejects tools where any of the four hints are missing or non-boolean.
For every tool, set all four hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) to explicit true/false values that match the handler’s actual behaviour.
Tool inputs are validated
27/34 tool handlers declare input schemas (79%)
Declare an inputSchema with zod/joi/yup on every tool definition.
Tool handlers catch errors
Only 9/34 tool handlers wrap calls in try/catch (26%)
Wrap each tool handler body in try/catch and return a structured error response.
Tests exist
No test files found
Add tests that exercise each declared tool.
No eval / new Function
2 eval() or new Function() calls — dynamic code execution
Replace eval / Function with explicit parsing or safer alternatives.
Shell command execution
2 child_process calls — runs shell commands
Prefer library functions over shell-outs. If you must shell out, ensure all inputs are properly escaped.
Production dependencies are patched
0 critical, 3 high severity in production deps — @modelcontextprotocol/sdk@1.12.0 (high), @modelcontextprotocol/sdk@1.12.0 (high)
Run npm audit fix, or upgrade the affected packages to a non-vulnerable version.
Claim the listing to review these findings one by one and send us a correction where you disagree, straight to the team. Claiming also means we tell you when the grade moves, and reach you first if we find anything urgent.
[](https://m8ven.ai/mcp/sabyasachidhal/mcpgoat)?variant=verified from the URL.Vetting this one by hand? Tool Check is an MCP that scores other MCPs. Add it once and ask Claude, ChatGPT, or any MCP client to grade a server, surface CVEs, check the publisher, and suggest safer alternatives — before you install.
https://m8ven.ai/api/mcp/tool-check