MCP & AI Agent Integration

Holded
automation.

17 AI agent actions for Holded — callable from any MCP-compatible runtime, Claude, Cursor, or Cerebral OS workflow. Use Holded to process orders, manage inventory, and handle customer requests autonomously. Full governance, audit trail, and dry-run safety on every execution.

No credit card required
Live in production environments
<200ms median execution
Dry-run before production
Execution trace
live
17
actions
100%
governed
<200ms
latency
17
AI agent actions
8
Read operations
9
Write operations
1
High-risk actions (approval gated)
E-Commerce Holded is a E-Commerce integration — use it to automate order management, fulfillment, and customer operations from any AI agent or MCP-compatible runtime.
Actions

What you can do
with Holded.

Every action below is available as an MCP tool and a verb in Cerebral OS — callable from any AI agent, Claude, Cursor, Windsurf, or your own runtime via the BYOA API. All executions are governed, audited, and dry-run safe.

Get Contact
holded:get_contact
Fetch a single contact by ID with all details including billing info.
Read Low risk
Get Expense
holded:get_expense
Fetch a single expense by ID with all details including attachments.
Read Low risk
Get Invoice
holded:get_invoice
Fetch a single invoice by ID with all details including line items and payment status.
Read Low risk
Get Product
holded:get_product
Fetch a single product by ID with all details including pricing and stock.
Read Low risk
List Contacts
holded:list_contacts
List contacts with optional filtering and pagination.
Read Low risk
List Expenses
holded:list_expenses
List expenses with optional filtering and pagination.
Read Low risk
List Invoices
holded:list_invoices
List invoices with optional filtering and pagination.
Read Low risk
List Products
holded:list_products
List products with optional filtering and pagination.
Read Low risk
Create Contact
holded:create_contact
Create a new contact in Holded.
Write Medium risk
Create Expense
holded:create_expense
Create a new expense in Holded.
Write Medium risk
Create Invoice
holded:create_invoice
Create a new invoice in Holded.
Write Medium risk
Create Product
holded:create_product
Create a new product in Holded.
Write Medium risk
Delete Contact
holded:delete_contact
Delete a contact from Holded. WARNING: This action cannot be undone.
Write High risk
Mark Invoice Paid
holded:mark_invoice_paid
Mark an invoice as paid by recording a payment.
Write Medium risk
Send Invoice
holded:send_invoice
Send an invoice via email to the contact or specified email address.
Write Medium risk
Update Contact
holded:update_contact
Update an existing contact in Holded.
Write Medium risk
Update Product
holded:update_product
Update an existing product in Holded.
Write Medium risk
MCP & Runtime API

Call Holded
from any AI agent.

Any AI agent — Claude, Cursor, LangChain, AutoGen, or your own — can call Holded actions through the Cerebral OS Runtime API. Governance, credentials, and audit trail fire automatically.

holded:get_contact READ
# Call via Runtime API
curl
-X POST \
  "https://api.cerebralos.com/v1/runtime/actions/run"
  -H "X-API-Key: YOUR_KEY" \
  -d '{
    "verb": "holded:get_contact",
    "args": {},
    "execution_id": "agent-001"
  }'
holded:create_contact WRITE
# Dry-run first — no production risk
curl
-X POST \
  "https://api.cerebralos.com/v1/runtime/actions/run"
  -H "X-API-Key: YOUR_KEY" \
  -d '{
    "verb": "holded:create_contact",
    "args": {},
    "execution_id": "agent-001",
    "metadata": {"dryRun": true}
  }'
Get your Runtime API key at app.cerebralos.com/signup — 1,000 free executions, no credit card required.
AI agent examples

What your AI agent
can do with Holded.

Real patterns your AI agent can execute via MCP or the Runtime API. Every action governed, dry-run safe, and fully audited.

Trigger
AI agent needs Holded data
Call holded:get_contact via MCP or Runtime API
AI processes result and takes next action
Full execution logged to audit trail automatically
Trigger
Workflow needs to write to Holded
Dry-run validates holded:create_contact before execution
Approval gate fires if risk level is high
Action executes with full governance — logged, audited, reversible
Trigger
Order or fulfillment event
Process with AI
Update customer records
Notify team via Slack
How it works

Every Holded action
governed end-to-end.

Cerebral OS isn't a connector. It's the execution layer that sits in front of Holded — adding governance, dry-run safety, and a full audit trail to every operation.

Governance first
Every verb carries a risk classification. High-risk writes require explicit approval gates before they execute in production.
Dry-run safe
Simulate any Holded action before it touches production. See exactly what would happen before a single real call is made.
Immutable audit trail
Every Holded action is logged — what ran, what changed, who approved it, when it happened. Full history on every verb, forever.
Connect with

Holded works best
alongside these.

Build multi-step workflows that connect Holded to the rest of your stack. All governed. All audited.

Holded integration

Start free.
No credit card required.

Start free with 1,000 runs — no credit card required. Connect Holded in minutes, dry-run every action before it touches production, full audit trail on everything.

Start free — 1,000 runs Browse all integrations →