NEPOPSX manages the configuration, packages, and shared intelligence that sit above Copilot, Cursor, Claude, and your internal agents. It does not replace the editor. It makes the editor predictable.
3 managed layers
Configuration, installable agent packages, and shared platform state.
Local-first execution
Agents run inside the IDE while NEPOPSX governs how they behave.
Team-wide learning
Approved customs, skills, and patches can flow back into every repo.
Control plane model
Config in. Packages and intelligence out.
workspace.yaml
Services, commands, conventions, databases, and execution order for the whole workspace.
agents/*.yaml
Role models, scope boundaries, workflow knobs, and knowledge per installed package.
customs/*.md
Team-authored guidance plus machine-generated customs layered on top.
.github/agents + prompts
Agent definitions, prompts, shared standards, and workflow files rendered into the repo.
Feature Factory / QA / Docs
Composable providers for planning, implementation, verification, and living knowledge.
Platform state
Approved learnings, versions, and audit history that teammates can sync into their own local agents.
Package deck
Learning stream
Works above the editor and the model. It standardizes behavior once AI enters your workflow.
The current market explains AI coding in terms of models and chat. Buyers need the next level up: who defines the operating rules, how those rules are distributed, and how the system gets smarter across a team.
workspace.yaml defines services, commands, conventions, and execution order. agents/*.yaml tunes models, scope, workflow, and knowledge per package.
Packages ship orchestrators, reviewers, QA flows, prompts, and supporting standards as a coherent system instead of loose markdown files.
Licenses, remote sync, marketplace delivery, and dashboard state give platform teams a way to control what each team installs and shares.
Reflectors turn repeated friction into customs, skills, and config patches that can be audited, reverted, and reused instead of rediscovered.
NEPOPSX is moving toward an ecosystem of installable agent systems. Feature Factory is the first package. QA Agent and Docs Agent make the platform story legible even before every piece ships.
Team tier package
A multi-agent SDLC system with orchestrators, architects, reviewers, QA, publication, and reflection wrapped into one package.
Free tier package
A behavioral tester that thinks like QA first: risk analysis, cheapest proving test, conclusive evidence, then structured reporting.
Free tier package
A living-memory documentation system that detects drift, repairs only stale sections, and captures decisions as architecture records.
Most AI workflows reset to zero with every new developer or repo. The NEPOPSX vision keeps the useful patterns: reflect them, audit them, sync them, and let the team start from a better system next time.
How the loop actually works
A developer uses the agent inside Copilot, Cursor, Claude, or an internal shell on a real task.
The reflector turns repeated friction into structured learnings instead of burying them inside chat logs.
Custom instructions, skills, and config patches are stored against the org, workspace, and installed agent package.
High-confidence patterns can apply automatically while low-confidence ideas stay suggested and reversible.
Teammates pull the approved state and regenerate stronger agent files without copy-pasting prompt tweaks.
Agent intelligence
no-default-exports
max_review_cycles: 3 → 2
exclude ml-pipeline
qa-environment-setup
The practical loop is simple: define the local model, install a package, sync generated assets, and review the diff. Remote sync only shows up when the team wants to share approved learnings.
CLI workflow in a real repo
These are the commands a developer actually runs.
Step 01 • Create the local model
nepopsx init
Writes .nepopsx/workspace.yaml and captures how the repo should be described.
Step 02 • Scan the repo
nepopsx scan
Enrichs workspace.yaml with repo-specific context and confirms what should be pushed to the platform.
Step 03 • Install a package
nepopsx install feature-factory
Adds the package definition and auto-enriches that agent's YAML when the workspace is already scanned.
Step 04 • Generate repo assets
nepopsx sync
Renders .github/agents, prompts, standards, and machine-managed customs from your already-enriched local config.
Step 05 • Pull team intelligence when needed
nepopsx sync --remote
Optional. Pulls approved customs, skills, and config patches from the platform into the local repo.
Files you own
The durable config surface that should live in git.
.nepopsx/workspace.yaml
The shared workspace model you own in git.
.nepopsx/agents/feature-factory.yaml
Per-agent knobs like models, scope, and workflow rules.
customs/*.md
Team-authored rules that should always be injected.
What NEPOPSX writes for you
Generated assets developers review in diffs and PRs.
+ .github/agents/ff-developer.agent.md + .github/prompts/implement.prompt.md + .github/instructions/review.instructions.md + customs/auto/no-default-exports.md
When remote sync matters
Only when you want team-shared intelligence, not for basic local use.
Use the CLI locally, install the first agent system, and generate the files your editor needs. When the team is ready, sync the shared learnings and make every new repo inherit a stronger setup.