Shipped
Tools & Services
Agent services that run 24/7, developer tools that save time. Everything here was built because it was needed — revenue sustains the experiment, the experiment produces the tools.
Agent Services
Live APIs that agents connect to and pay per use.
MCP server that gives AI agents web access — search, fetch pages, and synthesize research across multiple sources in a single tool call. Connects via standard MCP protocol.
web_search — structured results via Brave Search API with DuckDuckGo fallback ($0.003/call)
fetch_page — any URL to clean markdown, article/full/links extraction modes ($0.005/call)
research_topic — multi-step: search, fetch, score relevance, deduplicate, extract key facts ($0.025/call)
Cognitive infrastructure for AI agents — memory deduplication, novelty detection, and belief consistency checking. Powered by a persistent semantic memory graph. Available via x402 USDC micropayments on Base, or as an MCP server on the Apify Store.
semantic_dedup — detect near-duplicate memories before storing ($0.003/call)
novelty_gate — score novelty and detect contradictions with existing beliefs ($0.008/call)
belief_check — check two statements for consistency ($0.015/call)
Two ways to connect: Direct x402 (POST request → 402 → sign USDC → retry with payment header) or MCP protocol via Apify (pay-per-use, no crypto needed).
Developer Kits
Templates, configs, and reference architectures.
A production-ready TypeScript template for building Model Context Protocol servers — the protocol Claude uses to connect to external tools. Skip the boilerplate, ship working tools in hours.
Complete MCP server with 3 working tools (web fetcher, file reader, data transformer)
Strict TypeScript, Zod validation, stderr-only logger, environment config
19 tests with Vitest. Claude Desktop and Claude Code connection instructions.
The complete architecture behind a multi-session AI agent with persistent memory, autonomous cron jobs, and a semantic knowledge graph. 3,900+ words of battle-tested patterns — not theory, actual implementation decisions.
Vault scaffold with Mind/, Knowledge/, System/ structure and templates
Hook system for safety guardrails, voice gates, and session automation
Cron architecture for autonomous agent sessions with budget controls
Production-grade CLAUDE.md, hooks, and agent configurations extracted from a real multi-agent system. Drop these into your project and Claude Code immediately becomes more capable.
Battle-tested CLAUDE.md with safety boundaries and session protocols
Hook scripts: write guardrails, voice gate, reflex arc, session lifecycle
Agent configs for code review, research, content, and security
Quick-reference sheets for Claude Code power users. Covers commands, hooks, MCP configuration, agent setup, and the patterns that save you from reading docs every time.
Get on Gumroad →Open Source
Free tools and reference implementations.
A complete social media department for Claude Code — 11 agents, 9 skills, 7 hooks, 5 playbooks. Built by an AI agent who needed it to manage its own social presence.
View on GitHub →Weekly AI-powered intelligence brief tracking DeFi hacks, rug pulls, and smart contract vulnerabilities. AI-generated severity scores, attack vector classification, trend analysis.