Replicate

What it does

Safeguard reads an inventory of what your Replicate account runs: its deployments (the model and the pinned version each serves, the hardware and the instance range), its trainings (fine-tunes, with the base model and version), and the models those name, with each model's licence link. The results appear as AI asset inventory entries and continuous tests (model provenance, licence risk, pinned model versions). Find the Replicate card under Integrations on the Project/Product tab (the AI Model filter shows only AI model cards). Nothing it reads is sent to a model: no customer data goes to third-party LLMs.

Public vs private

Supported auth methods

Permissions required

Safeguard only reads: it sends GET requests and nothing else. Replicate documents no token scopes, so the token you give can do everything your account can (run models, change deployments, start trainings). Treat it as an elevated grant, keep it only in Safeguard, and delete it to disconnect.

Replicate API token

Least privilege: Replicate offers nothing narrower than an account token. Use a token of the organization that owns the deployments, not a personal token with access to other organizations.

Token types

Expiry: rotate the token on your own schedule, for example every 90 days. When a token is deleted, Check access fails with the rejected token message below.

What we never need: predictions, their inputs or outputs, files, webhooks, billing or member management. Safeguard never calls the predictions API, because it returns each prediction's inputs and outputs.

Step-by-step setup

Create the credential in Replicate

1. Sign in to Replicate. For an organization, switch to it in the account menu first.
2. Open replicate.com/account/api-tokens and create a token named Safeguard (read-only use).
3. Copy the token.

Or with the Replicate CLI

Replicate has no CLI for creating API tokens.

Connect it in Safeguard

1. Open Integrations and choose the Project/Product tab (optionally the AI Model filter).
2. Click Connect on the Replicate card.
3. Give the connection a name and paste the token into the API token field. The Setup steps section repeats the token steps above, with a copy button.
4. Save, then click Check access. Safeguard calls GET /v1/account with the token. Success reads The token reads Replicate <user or organization> <username>. Replicate API tokens cannot be limited to reads: the token can also run models and change deployments; Safeguard only reads.

Webhooks and triggers

This integration has no webhooks. The inventory is read each time the connection's tests run, on their schedule or when you run them by hand.

What we read and store

• Read: the account's username and type, its deployments and their current release, its trainings, and the public or owned model records those name.
• Stored: the inventory (deployment and model names, versions, hardware, instance range, licence links, training status and base model) in the AI asset inventory, and the test results. An asset that disappears is marked inactive, not deleted. The token is encrypted at rest and never shown again after you save it.
• Not stored or read: predictions, prompts, inputs, outputs, training data and weight files. The training list also returns each training's input, logs and weight URL; Safeguard drops them unread.
• Replicate documents no way to list your own models directly, so a private model that no deployment or training uses is not in the inventory.
• No customer data from this integration is sent to third-party LLMs.

Troubleshooting

Rotation and disconnect

• Rotate: create a new token at replicate.com/account/api-tokens, edit the connection in Safeguard, paste it and save, then delete the old token.
• Disconnect: delete the connection in Safeguard, then delete the token in Replicate.

Vendor references

• Replicate HTTP API reference, checked 2026-10-16.
• Replicate OpenAPI schema, checked 2026-10-16.
• Replicate API tokens, checked 2026-10-16.