Work
I build, explain, and share the systems, tools, and lessons that make AI useful in production.
Explore the AgentsKit ecosystem by the problem it solves, the people it is for, and the stage of work it supports.
AgentsKit ecosystem
Open tools and practical systems for teams moving from promising prototypes to work they can understand, operate, and improve.
Learn
Playbook · Active
Learn practical patterns for building reliable production AI agents. A practical guide to adoption, governance, and agent architecture for engineering leaders moving beyond prototypes.
For Engineers and leaders learning how to ship AI responsibly.
Build
AgentsKit · Active
Build production AI agents with a portable JavaScript foundation. Runtime, tools, memory, RAG, UI, evals, and observability behind composable interfaces you can change without rewriting the product.
For Engineers building agents in JavaScript or TypeScript.
Compose
Registry · Active
Compose reusable agent primitives without rebuilding the same patterns. Install agents, tools, and skills directly into your codebase. Adapt them, inspect them, and keep the code.
For Builders who want composable, reusable agent components.
Understand
Doc Bridge · Active
Make human-to-agent documentation handoffs deterministic. Turn repository docs into deterministic starting points, conformance gates, and memory that can move back into the documentation.
For Teams turning documentation into reliable engineering context.
Deliver
AgentsKit Chat · Active
Deliver one consistent agent experience across product surfaces. A typed application layer for web, mobile, and terminal chat — one contract that keeps the conversation consistent as the surface changes.
For Teams shipping agent interactions across multiple interfaces.
Operate
AKOS · Active
Operate agents with the controls and visibility production requires. Orchestration, human approval, access control, and signed audit trails for teams running agents on their own infrastructure.
For Engineering leaders operating AI systems in production.
AI Weekly · Every Tuesday
Research translated into an architecture choice, its trade-off, and one experiment you can run in under 30 minutes.