beyond-mcp
Provides four example integration patterns for accessing Kalshi prediction market data with AI agents: an MCP server wrapper around a CLI, a standalone CLI with JSON/Human output and local caching, progressive-disclosure file-system scripts, and a Claude Code “skill” that reuses those scripts with team sharing via git.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Uses HTTPS base URL for Kalshi API and states no auth required for read-only public data. Scope granularity and auth controls are effectively minimal because there is no auth layer here. README does not document rate limits, input validation, logging redaction, or secret handling practices; dependency hygiene cannot be confirmed from provided content. Local caching writes to .kalshi_cache, so consider filesystem permissions and data handling in shared environments.
⚡ Reliability
Best When
You need agent-accessible market read tools and want to compare/implement patterns that preserve context (scripts/skills) or standardize integration (MCP) while using local caching for search.
Avoid When
You require robust, documented enterprise-grade interfaces (OpenAPI/SDK, guaranteed error semantics, explicit rate-limit/error recovery guarantees) and audit-ready security controls for production traffic.
Use Cases
- • Building agent tools for reading Kalshi market data (read-only)
- • Choosing between MCP vs CLI vs scripts vs Claude Skills based on context/token constraints
- • Keyword search over markets/series using a local cached index
- • Providing tool access for multiple LLM clients (via MCP) or a single ecosystem (via Claude skills)
Not For
- • Performing authenticated/write operations (repo indicates read-only access without authentication)
- • Production-grade, fully specified, turn-key API/SDK consumption without reviewing code quality and error handling
- • Environments requiring formal SLAs or guaranteed uptime characteristics
Interface
Authentication
README states no authentication required for read-only public data from Kalshi API. This repo itself likely relies on calling Kalshi public endpoints directly from scripts/CLI/MCP wrapper.
Pricing
No pricing information for this repo; cost is mainly engineering + runtime for fetching/building local cache (first run).
Agent Metadata
Known Gotchas
- ⚠ MCP tool calls are stateless per README (each call loses conversational context).
- ⚠ MCP wrapper delegates to CLI via subprocess, which may add latency and complicate debugging.
- ⚠ Search-first-run cache build can take several minutes; agents/users may see progress while cache is generated.
- ⚠ Skill approach is Claude Code specific (not portable to other MCP/agent runtimes).
- ⚠ File scripts require local filesystem access and correct tool/script loading for progressive disclosure.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for beyond-mcp.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.