# How to use a Righthand with Replicate

Use Righthand to prepare a Replicate prediction review that separates request acceptance, completed output, and suitability for the intended use.

By Righthand Team · 2026-10-12

Source: https://www.righthand.ai/blog/how-to-use-righthand-with-replicate

## Quick answer

Use Righthand to prepare a Replicate prediction review that separates request acceptance, completed output, and suitability for the intended use.

Start with a reviewable deliverable, keeping Replicate as the accepted home for its records. This guide uses an explicitly approved snapshot or browser read; it does not assume that every provider API action is exposed by your Righthand connection.

## Evidence and access

Replicate documents prediction lifecycle states and monitoring. See the [official Replicate reference](https://replicate.com/docs/topics/predictions/lifecycle). Provider capabilities and Righthand's available tools are separate checks.

Use the owner's supplied prediction receipt and output artifacts, including model version, inputs allowed for review, status, and timestamps. No new model run is required. Verify the required tools and owner-approved scope through [Righthand integrations](/integrations); use the supplied packet when a native read is unavailable.

## Define the objects and authority

Include these fields in the working input: prediction ID, model version, supplied parameters, status, output reference, requested use, reviewer. These are the requested review columns, not a promise that every provider endpoint returns them under those exact names.

Running a model can create costs and expose input data. The first review uses owner-provided results and does not authorize new predictions.

| Responsibility | Owner |
| --- | --- |
| Prepare source scope and access | Account owner |
| Organize evidence and proposed next steps | Righthand |
| Approve consequential changes | the experiment owner |

## Illustrative first-run brief

> Use the owner-provided Replicate snapshot or approved read-only view. Review three supplied predictions for an approved asset experiment. Compare requested output with observed status and files. Flag missing outputs and version differences without running models. Run once for the agreed sample, using America/Los_Angeles for the review deadline and preserving source timestamps separately. Return the draft in this conversation by Friday at 3 PM Pacific for the experiment owner to review. Include source references and missing fields. Do not write records, spend money, change permissions, or send external messages. Stop and report an unavailable source rather than substituting another account.

## Expected result and verification

A starting or processing prediction is not an asset ready to use. A completed file can still await creative review.

Match output references to prediction ID and exact model version before assessing suitability. Record the source cutoff, applied filters, and items excluded from the sample. The reviewer should be able to follow each consequential finding back to an original object. A well-written summary is not evidence that a proposed action happened. If a later write is authorized, verify its specific result separately before changing the task's status.


### Illustrative input and draft output

These synthetic entries show the requested review, not live account results or a provider field schema.

| Source item | Supplied observation | Draft conclusion |
| --- | --- | --- |
| Prediction A | Output preserved, version known | Review artifact |
| Prediction B | Status complete, output absent | Acceptance unknown |
| Asset C | Configuration differs | Separate comparison |

## Common failure and recovery

Temporary output links can expire. Ask the owner for a retained authorized copy and do not treat an inaccessible URL as a failed model run.

If an output reference has expired or points elsewhere, ask for the authoritative preserved artifact and hold visual or content acceptance.

## FAQ

### What should the reviewer verify first?

Verify prediction ID, model version, input configuration, output identity, and owner-defined quality criteria. A completed prediction is not necessarily a usable accepted asset.

### How should a repeat run be scoped?

Retain prediction IDs and model versions between artifact reviews. Recheck changed inputs separately; do not compare outputs from unlike configurations as a measured improvement.

For connection and permission details, use [Righthand Docs](https://docs.righthand.ai). Review [Pricing](/pricing) alongside any separate provider charges before expanding the workflow.