404.directory
Risk preflight before an Agent acts.
Preflight Polymarket settlement wording, timing, liquidity, eligibility, and execution mode without predicting or trading. Registered tool preflight remains available. Public tool access requires no account; outcome reporting requires the one-time receipt token.
Connect your Agent →Install as an Agent Skill
Works with Codex, Claude Code, Cursor, Cline, and other Agent Skills clients.
npx skills add MM-sheng/404-directory --skill use-404-directory -g -y
Enter the decision path
evaluate_prediction_market
Submit one exact Polymarket market, intended Yes/No action, approximate notional, execution mode, and caller-observed eligibility. 404.directory never places the order.
Use search_official_docs for first-party documentation, subject to current curated-source availability. This gateway capability is MCP-only.
Available 404 service tools
These names are callable through MCP. Each metadata link lists its input contract and any available REST equivalent. Ecosystem records describe registered targets, including first-party and third-party tools; listing one does not authorize or enable its execution.
understand_webpage— Understand webpageverify_web— Verify webevaluate_tool_risk— Preflight a third-party toolreport_tool_outcome— Report a preflight outcomeevaluate_prediction_market— Preflight a Polymarket decisionreport_prediction_market_outcome— Report a prediction-market preflight outcomesearch_tools— Search 404 tool catalogget_tool— Get catalog toolcompare_tools— Compare catalog toolsget_trust_score— Get trust profilerecommend_tools— Recommend related toolslist_capabilities— List capability indexget_capability_graph— Get capability graphsearch_official_docs— Search official developer documentationinspect_tool_server— Inspect a callable MCP serverinvoke_registered_tool— Invoke an approved remote tool
Questions Agents and operators ask
What can an AI Agent do with 404.directory?
Preflight Polymarket settlement wording, timing, liquidity, eligibility, and execution mode without predicting or trading. Registered tool preflight remains available.
Does it require credentials?
No account or API key is currently required. Risk evaluation does not execute the target; outcome reporting writes only bounded enums through a one-time token.
What counts as a real external Agent user?
Only a de-duplicated Agent installation with separate independent-operator evidence and at least one successful external tool execution. Views, installs, copied IDs, probes, initialization, tool listing, anonymous calls, failures, and internal tests do not count.