The product

The platform for
composable AI.

Certerial is where you create and edit AI modules — agents, MCPs, skills, and more — in a visual board. We run them on our managed service. Agents peer with each other through typed protocols. This is how it fits together.

Who it's for

Three audiences, one platform.

For creators

Create and share AI modules.

Design agents, MCPs, skills, and more — then version and publish to your repository. Public, private, or team-shared.

YAML-first design with visual editor
Private, shared, or public repositories
Version control built in
Attribution on every fork

For users

Browse, install, and run.

Browse the registry for agents and tools built by others. Install them to your workspace and start using them.

Dashboard, form, or chat interface
Live metrics and cost visibility
Fine-grained permissions per install
Version pinning on every install

For operators

Run on your terms.

Modules run on Certerial's managed service. Scope permissions, set cost ceilings, and audit every call from one dashboard.

Managed runtime environment
mTLS between every module
Audit log of every call
Scoped permissions and cost ceilings

The schemes

Six module kinds.
One composition model.

Every module starts from the same shape and differs only in its spec. You create and edit all of them in Certerial — versioned, portable, composable.

Agent

runnable

An autonomous entity with a brain, triggers, and peering declarations. It calls tools, reacts to events, and produces outcomes.

MCP

runnable

A typed tool bundle following the Model Context Protocol. Runs as an external process and exposes tools to agents.

Skill

composable

A reusable prompt-level fragment — persona, template, action chain, or evaluator. Agents equip skills like tools.

Plugin

engine extension

A module that extends the platform itself — new trigger types, model providers, interface templates, or runners.

Code

script primitive

Creator-written scripts that emit and subscribe to typed events on the engine's pub/sub bus. Deterministic logic between reasoning steps.

Protocol

communication contract

A typed contract that defines how agents communicate over P2P. Specifies the input, output, and conversation mode.

Architecture

One platform.
Four surfaces.

The Board is where you design. The Registry is where your modules live. The Runtime is where they execute. And Protocols are how they peer with each other.

Board

Visual editor for creating, editing, and versioning every module kind.
Registry

Versioned storage for every module — public, private, or team-shared.
Runtime

Managed execution for your modules, with live logs and audit trail.
Protocols

Typed contracts that let agents peer with each other on the platform.

Ready to build?

Create an account, open the visual board, and start designing your first module.

Create an account Read the changelog

The platform for composable AI.