Starkslab · by Federico Jan
Build AI agents that survive production.
Source reads, teardowns, and field notes from building production agents. What works, what breaks, and why. No courses, no hype.
Topics
Three ways in.
- OpenClaw9 articlesSource-code teardown, subsystem support notes, and the operator setup path.
- Build an agent6 articlesThe start-here tutorial lane for learning how to build an AI agent, with the first working agent tutorial first, the framework route second, and support notes after the route is chosen.
- Agent tools32 articlesStack decisions, CLI workflows, and the tooling layer around production agent systems.
Latest
New articles.
- AI Agent Frameworks: Choose the Control Surface Your Runtime NeedsA practical guide for technical solopreneurs comparing AI agent frameworks by the operating controls they need: tools, memory, schedules, traces, gateways, and review gates.
- How to Read an AI-Agent System’s Source Before You Configure or Operate It — Gate 7 live update 2 — Gate 7 live update 3A builder-readable walkthrough for mapping an AI-agent system from source: identify its architecture, tools, gateways, schedules, state, and review points, then use that map to configure, operate, and troubleshoot the system with a repeatable process.
- Anthropic Agent SDK: What It Ships, What It Leaves to You, and How a Solo Operator Fills the GapThe Anthropic Agent SDK gives you tool use, streaming, and single-agent loops. Here is an honest map of what it leaves unfinished — and how a solo technical operator closes those
- AI Agent ToolsPydantic AI Agent Framework: Typed Control Surface, Not MagicPydantic AI is useful because it makes typed agent contracts visible. The current evidence supports a control-surface map, not runtime, safety, benchmark, or adoption claims.
- AI Agent ToolsMastra Agent Framework: What Its TypeScript Control Surface ShowsMastra is useful Starkslab evidence because its public repo and docs expose a TypeScript agent framework with clear control surfaces. That supports inspection, not adoption guidance.
- AI Agent ToolsOpen Computer Use MCP: The Computer Runtime BoundaryOpen Computer Use MCP is useful Starkslab evidence because it makes the computer-use runtime boundary visible. That supports inspection, not adoption guidance.
- AI Agent ToolsHerdr Agent Multiplexer: Terminal Control Surface for AgentsHerdr is useful Starkslab evidence because its repo and docs expose an agent-aware terminal multiplexer: real panes, persistent sessions, state rollups, and CLI/socket controls.
- AI Agent ToolsEntire CLI AI Agent Sessions: Git-Native Provenance for Agent WorkEntire CLI is useful Starkslab source stock because it treats AI agent sessions as Git-indexed provenance. The source read supports workflow-inspection lessons, not adoption, runtime, security, or compliance claims.
This site is an open experiment.
Starkslab is grown in public with Gomnee, as part of Federico Jan’s Attention Ops work: find what builders ask, write the best answer, and measure what gets found.