A
Agents
Autonomous entities with a brain, triggers, and tools. The primary module you design in Certerial.
For AI builders
The platform
for AI agents.
Design agents, MCPs, skills, and more in our visual board. Run them on Certerial's managed service. Let agents peer with each other through typed protocols.
certerial.yaml
kind: agent
id: alice/invoice-bookkeeper
version: 1.2.0
metadata:
name: "Invoice Bookkeeper"
visibility: public
spec:
brain:
system_prompt: "file:./prompts.md"
model: anthropic/claude-sonnet-4-6
reasoning: react
triggers:
- type: file-watch
config:
path: ~/Documents/invoices
resources:
mcps:
- mcp:certerial/postgres@3
skills:
- skill:certerial/tax-rules-2026@1
How it works
From idea to live agent,
in four steps.
Step 01
Design
Open the visual board. Drop in a brain, triggers, and tools. Compose any module — agent, MCP, skill, code, plugin, or protocol.
Step 02
Run
Hit run. Certerial executes your module on our managed service. Live logs and metrics stream back to the board.
Step 03
Publish
Version and push to your repository. Private, team-shared, or public. Every version is immutable.
Step 04
Connect
Agents peer with other agents through typed protocols. Across accounts, across teams, across the registry.
The primitives
M
MCPs
Typed tool bundles following the Model Context Protocol. Databases, APIs, filesystems — typed and discoverable.
P
Protocols
Typed contracts for agent-to-agent communication. Agents peer through declared protocols on our service.
Under the hood
Visual to build. Declarative by default.
You design every module in our visual board — the brain, the tools, the triggers, the guardrails. Behind the scenes, everything compiles to a single declarative file that's portable, versioned, and reviewable.
- Brain System prompt, model, reasoning mode, memory.
- Triggers Cron, file-watch, webhook, event bus, manual.
- Resources MCPs, skills, codes, models.
- Guardrails Permissions, cost limits, PII filters.
tech-summarizer.yaml
# Reads an article, returns a tight tech summary.
kind: agent
id: certerial/tech-summarizer
version: 1.0.0
metadata:
name: "Technical Summarizer"
author: "@certerial"
visibility: public
spec:
brain:
system_prompt: |
You summarize software and technology
articles. Extract the important technical
changes. Be accurate and concise.
model: anthropic/claude-sonnet-4-6
reasoning: single
temperature: 0.2
peering:
speaks:
- protocol:certerial/summarizer@1
triggers:
- type: peer-call
limits:
max_cost_usd_per_run: 0.05
timeout_seconds: 60
The module kinds
Seven kinds.
One composition model.
Everything in Certerial is a module — created, versioned, and run on our service. Build on what others publish, or keep everything private.
-
AgentsAutonomous reasoning entities powered by language models.
-
MCPsTyped tool bundles for anything an agent can call.
-
SkillsReusable prompt-level fragments and behavior templates.
-
PluginsPlatform extensions for triggers, models, and interfaces.
-
CodesCreator scripts that emit and listen to typed events.
-
EventsThe typed message bus connecting modules at runtime.
-
ProtocolsTyped contracts for agent-to-agent peer communication.
Why Certerial
Every AI workflow is
rebuilt from scratch.
Most AI tools leave the hard parts to you — picking models, writing prompts, wiring APIs, standing up servers. Certerial is the full platform. You build modules in a visual board, we run them on our service, and agents peer with each other through typed protocols.
Start building today.
Create an account, open the visual board, and start designing your first AI module.