ForgeOps + MCP Knowledge Base
ForgeOps gives your agent project memory, scoped work, handoffs and guardrails, so no session starts from zero. It comes with its own local memory, the MCP Knowledge Base.
> /forge-load my-app ━━━ MY-APP SESSION LOADED ━━━ Phase: v2, checkout flow Last session: cart totals fixed, tests pass Next: wire the receipt email Is this still accurate?
ForgeOps moves that work out of your head and into files your agent reads every time.
forge-load, forge-scope, forge-build, forge-pr-loop and forge-close, plus align-prd, align-roadmap and the RISEN prompt builder.
A local, private memory server (SQLite, runs on your machine) that the skills query and write.
Installation for the skills and the Knowledge Base.
Project-state and handoff templates.
A filled-in .forge/ folder showing exactly what a governed project looks like.
How the skills pull memory in and write it back, documented.
The whole Forge method, written down.
The agent reads the project's state and asks if it's still right.
One bounded piece of work, with what's allowed and what isn't.
A build prompt grounded in your project's own files.
Build, test, review. You approve each merge.
A handoff is written, so tomorrow starts warm.
"I didn't know what I was doing with Claude. As sessions pile up with context, everything slows down, and starting fresh meant manually writing countless prompts trying to compile everything core to my project. ForgeOps has that built in. It makes the handoff for you, backs it up on your machine, and loading the project into a new session is all I do now. I forget what I'm doing quite often, especially after a few days. I just ask what we're working on and it lays it out."
"A session starts by reading my two files back to me and asking if that is still right. The next day, or a different AI, it starts where I left off instead of where it guesses I left off."
Why I built ForgeOps
I knew I was ready to get into agentic systems, but I didn't know how. After a long process of isolating good habits and removing bad ones, ForgeOps was born. If someone else knows the time is right to get into agents, ForgeOps is the system that gives them a real place to begin.
$197
No. It's a workflow layer: project state, source-of-truth docs, scoped loops, handoffs and guardrails, backed by a real local memory store.
No. The Knowledge Base is included and sets up in three steps. No accounts, no cloud, nothing leaves your machine.
No. It improves continuity, scope control and handoff discipline. You stay in the loop.
ForgeOps uses visible project files so state is inspectable, editable, portable, and tied to the project instead of living only in vague recall.
Yes, at first. When you're new you tell the AI to build five features in one prompt, with no roadmap and no idea what a PRD is, and you can go very fast. With the better models you even get results. Adding structure takes time: planning, scoping, phases, versions. You make it back by not fixing and re-debugging everything a Frankenstein prompt produced. The speed you lose up front comes back as time you stop spending on cleanup.
The whole loop is five steps: load, scope, build, pr-loop, close. Start with the included example state files. Setup instructions are included in the download.