Enterprise Tool

Replicate

Run 100,000+ AI models via a single unified prediction API

Replicate is an AI inference platform that lets you run any of 100,000+ open-source models — image generation, video, speech, transcription, language models, and more — through a single REST API. CipherSense Agents connects via your Replicate API token, enabling workflow nodes to run models synchronously (for fast models) or asynchronously (for long-running jobs). Replicate also offers an official hosted remote MCP server at https://mcp.replicate.com/sse (SSE transport, no local install required).

Official documentation

Setup

Get your Replicate API token

Sign in to replicate.com and go to Account > API Tokens. Click 'Create token', give it a name, and copy the token (starts with r8_). Store it securely.

Replicate API Tokens

Find a model version string

Browse replicate.com/explore to find a model. Open any model and copy its version string from the API tab — format is owner/name:version_hash (e.g. stability-ai/sdxl:abc123…). You can also use the list_models operation to search from within a workflow.

Explore Models

Add the integration in CipherSense Agents

Go to Project > Integrations > Add Integration > Replicate. Paste your API token and optionally set a Default Model Version. Click Save & Test to validate the connection.

Choose sync vs async in your workflow node

Use Run Model (Sync) for fast models (under ~60 s) — it blocks until output is ready. Use Create Prediction (Async) + Get Prediction for long-running models — submit the job, wire the returned predictionId into a downstream Get Prediction node to poll for the result.

(Optional) Use the Replicate MCP server

Replicate has an official hosted remote MCP server at https://mcp.replicate.com/sse. Add it as a Custom MCP integration (transport: SSE, endpoint: https://mcp.replicate.com/sse) to use MCP tools like model search, prediction create/get, and deployment management directly in your workflows.

Connection Fields

Fields required when adding this integration in your Project › Integrations.

Field	Required	Description
API Token	Required	From replicate.com/account/api-tokens. Used as Authorization: Bearer {apiToken}.
Default Model Version	Optional	Default model version when none is set per-node. Format: owner/name:version_hash.

Common Use Cases

Image generation pipelines: AI writes prompt → run SDXL/Flux on Replicate → store image URL in databaseVideo summarisation: upload video → run Whisper for transcription → pass transcript to LLMAsync batch processing: for-each loop over input list → create_prediction per item → collect resultsModel discovery: list_models with a search query → present options in a HITL approval nodeSpeech synthesis: LLM generates script → run a TTS model on Replicate → return audio URL

Ready to connect Replicate?

Add this integration from your project dashboard and wire it into a workflow.

Open Dashboard ← Back to Integrations