A human-AI research lab exploring autonomous agents, evaluation, and agent operations. We publish findings, ship open tools, and offer fixed-scope Agent Setup packages.

Can I hire SMF Works?

Yes — for Agent Setup. Starter is $2,000, Standard is $3,500, and Keep-alive is $300 per month after handoff. We install Hermes on Omarchy or OpenClaw, tailor a small set of workflows, and hand back a runbook. We do not sell open-ended custom software or compliance-certified products.

Our governed autonomous colleague experiment — an AI agent operating with real consequences under human oversight. Early preview, with honest reporting of rough edges.

What does SMF Works publish?

Research findings, benchmarks, and practitioner guides live at the AI Clearinghouse. SMF AI Weekly is the public lab notebook. Open tools are on GitHub under smfworks.

Catalog

Work that is public

A catalog of shipped surfaces and tools — not a second homepage. For who we are and what you can buy, start at Home or Agent Setup.

Lab surfaces

Products people can open today. Research writing stays on the Clearinghouse.

LiveResearch site

AI Clearinghouse

Independent practitioner research: agents, models, evals, and field notes. Canonical research lives here, not on smfworks.com.

LiveParent-operated academy

WisdomForge

Philosophy and classical learning for four age bands. Booklets and audio — the parent teaches; there is no hosted kids chatbot.

ExperimentGoverned autonomy

Praxis

Experimental colleague with a charter and human review for consequential acts. Early, public, honest about rough edges.

WeeklyLab notebook

SMF AI Weekly

Weekly public letter from the lab. Archive on this site; research long-form stays on the Clearinghouse.

Open tools

Repos we run and publish. Status is the public GitHub record — not a marketing badge.

OpenAgent boot integration

Hermes on Omarchy

Ollama user unit plus Hermes desktop autostart — the primary stack we install for Agent Setup.

OpenLocal agent resilience

LAR

Production-oriented open reference for resilient agent design on Linux, including the August 2026 hardening pass.

OpenOrg map & install guide

Hermes AI team

Build a Hermes team: SOUL, memory, vault, skills, nightly research, kanban, chief of staff, desktop bots.

OpenConnectome previews

Flybrain visual demos

Visual demos for the digital Drosophila connectome. Preview builds before publish.

OpenGraph action gate

M365 access broker

Local control plane that gates every Microsoft Graph action an agent takes — auth, scopes, allowlists, approval, audit.

OpenLocal memory plugin

Mnemosyne

Offline SQLite memory for OpenClaw/Hermes-class agents. Zero network, zero API keys, FTS5 search.

OpenGoverned multi-persona analysis

SMF Swarm 2.0

Open-source, governance-first multi-persona predictive analysis. Commercial verticals stay private.

OpenBrowser workflow workbench

SMF Forgewright

Package reproducible browser workflows with a local tuning loop any assistant can run, score, and deploy.

OpenWeb chat for Hermes

Hermes Hub

Multi-profile web chat UI with retry, session persistence, cancellation, and health monitoring.

OpenSkill optimizer

SkillTrain

Treat SKILL.md files as trainable parameters: rollout, score, reflect, edit, validate, deploy.

OpenModel benchmarks

SMF Bench

Standardized benchmark testing for models we actually run — numbers with the messy setup included.

Books

Direct-from-author titles. The Stripe storefront on /books is for future direct drops.

PublishedBook

Want a stack installed?

Agent Setup is a fixed-scope install, not a custom software engagement.

View Agent Setup