Koinoflow

Why Koinoflow

The first open-source platform for
AI skills, built for everyone

Existing solutions are built for developers who live in Git. But most of the people who own business processes don't. Koinoflow is the first platform purpose-built to let every team, technical or not, create, govern, and deploy AI skills at scale. Admins also decide which automated agents can access which skills. It is open source under the MIT license and developed by Visionect.

View on GitHubTalk to us

The problem

Existing solutions are built for people who already know how to code

Today's skill tools are fast and elegant for engineering teams. Set up a .skills manifest, push to Git, done. But that simplicity is only simple if you already know Git, Markdown, and your way around a CLI. For everyone else, it is a wall.

Developer-first tools assume

  • Git fluency across the org
  • Comfort with Markdown and YAML
  • Existing CI/CD pipelines
  • Dev team bandwidth for every skill change

Koinoflow gives you

  • A visual editor anyone can use
  • No Git, no CLI, no syntax
  • Built-in governance and review
  • Self-service for every department

What makes it different

Empower the people who actually own the work

The people closest to a process should be able to turn it into an AI skill themselves, without filing a ticket and waiting for engineering.

No Git required

Create, edit, and publish skills from a visual interface. No terminals, no repos, no pull requests.

No coding needed

Write skills in plain language. If you can describe a process, you can build a skill an AI agent follows.

Built for every team

Ops, HR, finance, support. Anyone who owns a process can turn it into a governed, reusable skill.

Visual skill builder

A guided editor walks you through workflows, guardrails, and ownership. No YAML, no Markdown syntax.

Governance built in

Versioning, ownership, review cycles, and usage analytics come standard. Not bolted on after the fact.

From idea to production in minutes

Publish a skill, deploy it to the agents that need it, and keep the agent code unchanged. No deploy pipeline to configure.

How teams use it

Three teams, one platform, zero engineering tickets

Pick the story that matches your role. Each team owns the process end-to-end, from draft to governed, versioned skill delivered over MCP.

Operations lead

"SOP to skill in 10 minutes, without filing an engineering ticket."

You own the vendor-onboarding SOP. Right now it lives in a Confluence page that nobody can find, and when a buyer asks an AI agent to "onboard vendor X," the agent invents its own steps.

In Koinoflow

  1. 1Open the visual editor and paste the SOP. Koinoflow structures it into workflow steps and guardrails.
  2. 2Assign yourself as owner and set a 60-day review cadence. Staleness alerts fire to your inbox.
  3. 3Publish. The skill is live on the MCP endpoint, and you choose whether every agent or only selected agents can use it.

Outcome

Every agent onboarding a vendor now follows the same approved workflow. Usage analytics show which teams adopted it.

People / HR partner

"Onboarding runbooks stay accurate even when the team changes."

New-hire onboarding touches IT, payroll, legal, and the buddy program. Every new hire gets a slightly different experience because each handoff is copy-pasted from last time's Slack thread.

In Koinoflow

  1. 1Turn the onboarding checklist into a single governed process with day-1, week-1, and month-1 sections.
  2. 2Tag the IT-provisioning and payroll sections with owners in those teams, and they get review reminders.
  3. 3When an AI agent runs onboarding for a new hire, it follows the current version and logs the run.

Outcome

Audit-ready history of every onboarding run. When the policy changes, every agent picks it up on the next call.

Support lead

"Refund policy is the same across chat, email, and AI tier-1."

Refund rules live in one doc for humans, another for a chatbot, and a third copy baked into an internal AI prompt. They drift. Customers get inconsistent answers; finance gets angry.

In Koinoflow

  1. 1Publish the refund policy once in Koinoflow with clear guardrails ("escalate if > $500", "no refunds after 90 days").
  2. 2Connect every client to the same MCP endpoint, then deploy the refund skill only to the human-facing help desk AI and internal tier-1 bot.
  3. 3Review usage analytics monthly: which refunds escalated, which agents consulted the current version.

Outcome

One source of truth. Consistent answers across every channel. Policy changes propagate in a single publish.

Side by side

Koinoflow vs. dev-first skill registries

The open-source dev-first registries are great if your team lives in Git. Koinoflow is what you reach for when the people who actually own the process don't.

Primary user

CLI- and Git-fluent developers

Process owners across ops, HR, finance, support, plus developers

Editing

Markdown + YAML in an IDE, `arctl skill init`, PR review

Guided visual editor, in-app review and approval

Ownership

Git blame + CODEOWNERS file

Named accountable owner per skill with escalation paths

Freshness

Manual; nobody remembers to check

Staleness alerts with configurable rules (e.g. "90 days since review", "500+ calls since update")

Analytics

DIY on gateway logs or per-server metrics

Built-in dashboard: adoption, health scores, per-client and per-version breakdowns

Delivery

Self-host gateway, Docker/OCI publishing, your own runtime

Self-host via Docker Compose or let Visionect host it: MCP endpoint live in 30 minutes either way

Agent distribution

Usually handled in prompts, code, gateway policy, or separate deployments

Create agent identities, deploy skills to all or selected agents, and rotate access without touching agent code

Setup burden

Kubernetes + Helm + OAuth provider + registry + CI/CD

Sign in, publish, paste MCP config

Open source

Yes, but you assemble the stack yourself

Yes (MIT licensed, developed by Visionect), plus an optional managed-hosting offering when you want someone else to run it

Enterprise ops

Assemble from OSS + your own IAM

Managed hosting helps you scope private deployment, IAM, and commercial support needs; self-hosting lets you wire those controls into your own stack

Dig into the alternatives

We respect the open-source work in this space. If you're a platform team evaluating options, here are the projects worth looking at alongside Koinoflow:

The bottom line

When the people who own your processes can also publish and govern the AI skills behind them, the entire organization moves faster.

Not just engineering. That's the wedge, and that's why Koinoflow is built process-owner-first.

Ready to empower your whole team?

Create governed AI skills without Git, code, or developer bottlenecks. Open source (MIT): clone the repo and self-host, or let Visionect host it for you.

View on GitHubTalk to us

Open source (MIT) · free to self-host · managed hosting by Visionect