Mentu

The Claude Code Monitor tool idea, taken out of the session and made durable, shareable and accountable. Monitors that keep watching, remember what they saw, and let people and AI agents subscribe.

MetaMCP keeps core MCPs direct and routes the long tail through 3 lazy tools and reviewed declarative Methods.

WebHarness — native macOS APIs for web apps + MCP plugin host for AI agents. Public source mirror.

Meta-MCP server that collapses hundreds of child MCP tools into 12 meta-tools (~1,000 schema tokens). Intent routing, lazy spawning, connection pooling, and V8-sandboxed code execution.

Cross-agent policy harness for Claude Code, Codex, Cursor, and Gemini. One observe/supply/gate core, thin per-agent adapters, fail-open by design.

Deterministic, offline, host-independent calendar scheduling engine (RFC 5545 + IANA tzdata). Byte-identical output; pure Python; Apache-2.0.

Repository navigation & progressive disclosure for AI agents over documentation — a locator whose BM25 default is pre-registered on markdown corpora (doi:10.5281/zenodo.21969901); not yet measured on source code. CLI + MCP server.

Frontmatter validation for evidence-disciplined corpora — the PD-1 schema: derivability-tested fields, a measured summary ban, demotion sets, read-back meter.

Homebrew tap for Mentu. Installs mentu-recipes, the runner for file-based agent workflows.

The Commitment Protocol: an accountability ledger for agent work. Append-only and hash-chained; commitments close against evidence. Specification, agent template, sample ledger and reference verifier.