Symptoms

Users can access the AgentFlow page, but the workspace displays the following message:

No models available, please contact support

This message appears when no active LLM configuration exists for the tenant. The chat input area is disabled, and users cannot start or continue AgentFlow conversations until an active model configuration is available.

Root Cause

The tenant's physical configuration had AgentFlow-related settings partially enabled:

"mcpConfig": {
  "enabled": true
},
"agentsConfig": {
  "aiEnabled": true
}

However, the following required AgentFlow enablement flag was missing:

"agentFlowEnabled": true

Expected configuration:

"mcpConfig": {
  "enabled": true
},
"agentsConfig": {
  "aiEnabled": true,
  "agentFlowEnabled": true
}

In addition, users accessing AgentFlow must have the required tenant roles and permissions. Reltio documents that AgentFlow uses dedicated execution permissions for both MCP and agents, and that MCP execution rights are a prerequisite for using agents.

Resolution 

Confirm tenant physical configuration

Verify that the tenant's physical configuration includes both MCP and AgentFlow enablement:

"mcpConfig": {
  "enabled": true
},
"agentsConfig": {
  "aiEnabled": true,
  "agentFlowEnabled": true
}

If agentsConfig.agentFlowEnabled is missing, update the tenant's physical configuration or engage the appropriate Reltio internal team to enable it.

2. Confirm user role assignments

Make sure each user accessing AgentFlow has the following base roles on the applicable tenant:

ROLE_USER
ROLE_API
ROLE_EXECUTE_MCP
ROLE_EXECUTE_AGENTS

ROLE_EXECUTE_MCP grants execute-only access for MCP tool calls through agentflow.mcp.EXECUTE.

ROLE_EXECUTE_AGENTS grants execute-only access to run MCP and AgentFlow agents through both agentflow.mcp.EXECUTE and agentflow.agents.EXECUTE.

Confirm LLM/model configuration status

If the user still sees:

No models available, please contact support

Ask a System Administrator to verify the LLM configuration status in Admin Settings. Reltio documentation states that this message appears when no active LLM configuration exists for the tenant. If all configurations are inactive, the admin should raise a request with Reltio Support to re-enable them.

Validate AgentFlow / MCP access

For MCP-backed access, validate that the user has a valid OAuth access token and the required MCP permissions. All MCP tool interactions require a valid access token issued by the Reltio Authentication Server, and only users with ROLE_EXECUTE_MCP or a role containing agentflow.mcp.EXECUTE can invoke MCP tools.

You can validate available MCP tools with:

GET https://<environment>.reltio.com/ai/tools/metadata/list
Authorization: Bearer <access_token>

The Metadata API returns available MCP tools, including each tool’s name, title, description, and input schema. The returned tool list depends on the user’s role and access permissions.

Example environment URLs:

https://dev.reltio.com/ai/tools/metadata/list
https://test.reltio.com/ai/tools/metadata/list
https://prod.reltio.com/ai/tools/metadata/list

Claude / MCP Integration Notes

When configuring Claude to connect to the Reltio AgentFlow MCP Server, Reltio documentation lists the following prerequisites:

MCP feature enabled in the tenant physical configuration
ROLE_EXECUTE_MCP or ROLE_ADMIN_MCP
AgentFlow MCP server endpoint
Claude desktop app
Permission to edit Claude configuration
Node.js installed

The documented MCP endpoint format is:

https://<reltio-namespace>.reltio.com/ai/tools/mcp/

For example:

https://dev.reltio.com/ai/tools/mcp/
https://test.reltio.com/ai/tools/mcp/
https://prod.reltio.com/ai/tools/mcp/

Reltio’s Claude configuration documentation notes that the Bearer token must be valid for the tenant, and an expired token can cause authentication failures.

Authentication Guidance

For interactive AgentFlow or Claude MCP access, use OAuth 2.0 Authorization Code Flow with PKCE, where supported by the MCP client. Reltio documentation states that the AgentFlow MCP Server uses OAuth 2.0 Authorization Code Flow with PKCE and that the MCP client handles the OAuth flow, including PKCE and token exchange.

For service-to-service or API flows, a token can be obtained separately and used to access MCP tools, provided it has the required permissions and tenant scope.

Security and Access Behavior

AgentFlow operates on live Reltio data while honoring RBAC, data masking, and audit policies. Reltio documentation states that AgentFlow agents use governed APIs and that each interaction is scoped to the user’s access level and auditable.

For Data Explorer Agent usage, users must have read access to the underlying entities, relationships, and attributes they are exploring. For future write-enabled agent use cases, additional write privileges or product-specific roles may be required depending on the downstream action.

Checklist

Use this checklist when triaging similar tickets:

[ ] Confirm tenant ID and environment.
[ ] Confirm mcpConfig.enabled = true.
[ ] Confirm agentsConfig.aiEnabled = true.
[ ] Confirm agentsConfig.agentFlowEnabled = true.
[ ] Confirm active LLM/model configuration exists.
[ ] Confirm user has ROLE_USER.
[ ] Confirm user has ROLE_API.
[ ] Confirm user has ROLE_EXECUTE_MCP.
[ ] Confirm user has ROLE_EXECUTE_AGENTS.
[ ] Confirm the user has required read permissions for Data Explorer.
[ ] Validate MCP metadata endpoint with the user’s token.
[ ] If Claude is involved, confirm endpoint, token validity, and Claude MCP configuration.