Use Relewise with an AI Assistant

BETA

Agent Gateway is in BETA, and we're continuing to expand its capabilities. Your feedback helps shape what we build next—reach out to us through Relewise Support.

The Relewise plugin is the recommended way to let a supported AI assistant work with your Relewise configuration, analytics, merchandising, and optimization. The plugin discovers the Datasets and capabilities available to you, so you do not need to find Dataset IDs or assemble API requests yourself.

The plugin is open source, follows the official Agent Plugins specification, and is maintained in the Relewise Agent Plugins repository.

The currently supported clients are OpenAI Codex, Claude Code, Claude Cowork, and GitHub Copilot CLI.

If your AI client is not listed, installation and local command support may differ. Contact Relewise Support before relying on the plugin with that client.

Before You Begin

You need:

• Access to My Relewise.
• The Use Agent Gateway permission for at least one Dataset. Administrators have this access by default.
• The REST API Connection Method enabled for that Dataset.
• The Allowed Areas required for the work you want the assistant to do.

If you cannot change the Agent Gateway Configuration, ask an Administrator or someone with the Manage Agent Gateway permission to enable the required access. See Security and RBAC for details.

Create a Personal Access Token

In My Relewise, open User > Personal Access Tokens, then select Create New Token.

Choose the smallest practical Dataset Scope. Use Selected Datasets when the assistant only needs to work with specific Datasets.

My Relewise shows the token value once. Store it immediately in a secure credential provider or another secure location. Never paste the token into an AI conversation, command argument, repository file, log, or transcript.

See Personal Access Tokens for token status, scope, regeneration, and revocation details.

Install the Relewise Plugin

Install the Relewise plugin from the public Relewise Agent Plugins repository. The repository includes the plugin and the relewise-agent runtime for supported operating systems.

OpenAI Codex

Add https://github.com/Relewise/relewise-agent-plugins.git as a plugin marketplace. Use main as the Git ref, leave Sparse paths empty, and then install Relewise from that marketplace.

Claude Code

Run:

claude plugin marketplace add Relewise/relewise-agent-plugins
claude plugin install relewise@relewise

Claude Cowork

In Claude, open Cowork, then open Customize > Plugins. Under Personal plugins, add a marketplace from https://github.com/Relewise/relewise-agent-plugins, then install Relewise from that marketplace.

GitHub Copilot CLI

Run:

copilot plugin marketplace add Relewise/relewise-agent-plugins
copilot plugin install relewise@relewise

Installation Instructions Can Change

AI client interfaces and installation steps can change. If these instructions no longer match what you see, please contact Relewise Support so we can update the guide.

Connect to Relewise

After installation, ask your AI assistant:

Help me connect Relewise.

The plugin's relewise-setup skill first checks whether authentication already works. If setup or repair is needed, it guides you through a setup that keeps working the next time you use Relewise:

• A secure credential provider available to your AI client. The client retrieves the token and passes it only to Relewise commands as RELEWISE_AGENT_GATEWAY_TOKEN.
• A persistent user or system environment variable named RELEWISE_AGENT_GATEWAY_TOKEN.

Follow the assistant's local setup guidance without sending the Personal Access Token in the conversation. Restart the AI client after changing a persistent environment variable so the client can read the new value.

Confirm the Connection

After setup, the assistant should verify authentication and list the Relewise Datasets you can access. Setup is complete when the assistant confirms the connection and can name at least one Dataset you expect to access.

If authentication succeeds but the assistant lists no Datasets, follow No Datasets Are Visible.

After the connection is confirmed, try prompts such as:

• What Relewise Datasets do I have access to?
• Review search performance for my Relewise Dataset.
• Compare search performance across two Relewise Datasets.
• Audit merchandising rules and point out conflicts or gaps.
• Show which features drive the most consumption.

Your user permissions, Personal Access Token Dataset Scope, Connection Methods, and Allowed Areas continue to control what the assistant can access.

Troubleshooting

The Token Is Missing or Rejected

Ask the assistant to help you reconnect. Confirm that RELEWISE_AGENT_GATEWAY_TOKEN is available to the AI client and that the token is active. Regenerate the token if the value was lost or may have been exposed. Authentication failures return 401 Unauthorized and do not appear in Agent Gateway Request Logs.

See Personal Access Tokens and Security and RBAC.

No Datasets Are Visible

Confirm that the token's Dataset Scope includes the expected Dataset and that your user has Use Agent Gateway for it. A valid token can authenticate successfully even when no Datasets are available to it.

See Agent Gateway Configuration and Security and RBAC.

A Capability Is Unavailable

Confirm that the Dataset has the REST API Connection Method and the required Allowed Area enabled. For requests that pass authentication, use Request Logs to see which access check blocked the request.

Advanced and Custom Integrations

The Relewise plugin is the recommended route for supported AI assistants. For applications that need direct HTTP access, use the Agent Gateway REST API. For a custom Dataset-specific AI connection, use the MCP endpoint.