Platform
Platform Overview Maps Memory Shadow Mode Widget Governance Actions Integrations
Developers
Overview Catalog Guardrails Dry Run Idempotency Credentials Audit Trail API Reference
Company
About Blog Contact Book a Demo
MCP Integrations Replicate
MCP & AI Agent Integration

Replicate
automation.

12 AI agent actions for Replicate — callable from any MCP-compatible runtime, Claude, Cursor, or Cerebral OS workflow. Use Replicate to automate business processes and connect tools in your stack. Full governance, audit trail, and dry-run safety on every execution.

Connect Replicate free Browse all integrations →
No credit card required
Live in production environments
<200ms median execution
Dry-run before production
Execution trace
live
12
actions
100%
governed
<200ms
latency
12
AI agent actions
8
Read operations
4
Write operations
2
High-risk actions (approval gated)
Business Tools Replicate is a Business Tools integration — use it to automate business operations and workflow automation from any AI agent or MCP-compatible runtime.
Actions

What you can do
with Replicate.

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 Collection
replicate:get_collection
Get details about a specific model collection and its models.
Read Low risk
Get Model
replicate:get_model
Get details about a specific model including its versions and schema.
Read Low risk
Get Prediction
replicate:get_prediction
Get the current status and results of a prediction.
Read Low risk
Get Training
replicate:get_training
Get the current status and results of a training job.
Read Low risk
List Collections
replicate:list_collections
List featured model collections on Replicate.
Read Low risk
List Models
replicate:list_models
List available models on Replicate with optional filtering.
Read Low risk
List Predictions
replicate:list_predictions
List your predictions with pagination support.
Read Low risk
List Trainings
replicate:list_trainings
List your training jobs with pagination support.
Read Low risk
Cancel Prediction
replicate:cancel_prediction
Cancel a running prediction to stop processing and avoid further charges.
Write High risk
Cancel Training
replicate:cancel_training
Cancel a running training job to stop processing and avoid further charges.
Write High risk
Create Prediction
replicate:create_prediction
Create a new prediction (run a model) with the specified inputs.
Write Medium risk
Create Training
replicate:create_training
Create a new training job to fine-tune a model with your data.
Write Medium risk
MCP & Runtime API

Call Replicate
from any AI agent.

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

replicate:get_collection READ
# Call via Runtime API
curl
-X POST \
  "https://api.cerebralos.com/v1/runtime/actions/run"
  -H "X-API-Key: YOUR_KEY" \
  -d '{
    "verb": "replicate:get_collection",
    "args": {},
    "execution_id": "agent-001"
  }'
replicate:cancel_prediction 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": "replicate:cancel_prediction",
    "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 Replicate.

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 Replicate data
Call replicate:get_collection 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 Replicate
Dry-run validates replicate:cancel_prediction before execution
Approval gate fires if risk level is high
Action executes with full governance — logged, audited, reversible
Trigger
Event in Replicate
Process with AI agent
Take governed action
Log to audit trail
How it works

Every Replicate action
governed end-to-end.

Cerebral OS isn't a connector. It's the execution layer that sits in front of Replicate — 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 Replicate action before it touches production. See exactly what would happen before a single real call is made.
Immutable audit trail
Every Replicate action is logged — what ran, what changed, who approved it, when it happened. Full history on every verb, forever.
Connect with

Replicate works best
alongside these.

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

Slack Gmail Hubspot Google Sheets Notion
Replicate integration

Start free.
No credit card required.

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

Start free — 1,000 runs Browse all integrations →