Engineer. Maker. Builder.
Making AI-assisted software development more reliable.
I build software, explore agentic engineering workflows and share what works, what fails and why.
latest writing
- I don't want a smarter coding agent. I want one that consistently follows how I work.Consistency goes further than cleverness: how session context in Markdown turned into a workflow my agents follow.post
the problem
My agent had my workflow and still skipped steps. The more rules and context I added, the less consistently it followed them.
Start over
Open a new session and let the agent work it out again.Patch it
Keep growing an instruction file, keep notes, paste summaries into the next session.Build it into the workflow
Give each step its own setup and carry what was learned into the next session.now building
nemonik · the workflow layer for coding agents
Give coding agents a repeatable workflow, the right tools for each step and the context to continue where the last session left off.
I'm building nemonik around the third approach. It breaks the workflow into kits, one per step, each with its own system prompt, skills, MCP servers and hooks, so the agent only gets what that step needs.
- > Kits. A system prompt, skills, MCP servers and hooks, tied together by a workflow.
- > Workflows. The steps in order and which skills to use. A workflow can start the next kit: you run the plan kit, approve the plan and the build kit opens in a new session on its own, then runs the review kit automatically when it's done.
- > Memory. A running memory, journal and status that carry what each session learned into the next, in files in your repo or in nemonik's hosted store.
Use Claude Code or Codex every day? Try nemonik on one real ticket.
Run one ticket your usual way and one with a kit and compare.
Join the beta
With the MCP, your kits, context and memory live in one place you can share, your agents search it by meaning and you can reach it from Claude or GPT on your phone.
Docs →