Overview

Reltio MCP Server is a lightweight, plugin-based Model Context Protocol (MCP) server built to bring language-model-assisted entity matching into Reltio deployments. It exposes a set of tools that translate high-level matching workflows (search, compare, merge, reject) into Reltio API calls, and it can be extended with custom plugins to modify or enrich matching logic. The server acts as a bridge between an MCP-capable client (for example, a desktop LLM interface or a custom agent) and the Reltio platform, allowing faster, contextual, and more accurate matching decisions powered by language models.

For developers, the project provides a predictable surface of callable tools and a small operational footprint so you can iterate quickly. The plugin architecture lets teams add domain-specific logic or connect additional data sources without changing core MCP behavior. Typical uses include automated duplicate detection, assisted merge/reject workflows, and integration with Reltio workflows and audit trails.

Features

Plugin-based MCP server for extensible matching logic
Pre-built tools for common Reltio operations (search, get, update, merge, relations)
Support for retrieving match history, potential matches, and merge trees
Tools for tenant and business-configuration inspection (data model, entity types)
Workflow and task utilities (list/reassign tasks, start process instances)
Lightweight HTTP server suitable for local development or container deployment
Designed to be used with MCP clients (LLM UIs, Claude Desktop, custom agents)

Installation / Configuration

Prerequisites:

Git and Docker (recommended) or a compatible runtime present on your machine
Reltio API credentials and tenancy information

Quick start (Docker):

# clone the repository
git clone https://github.com/reltio-ai/reltio-mcp-server.git
cd reltio-mcp-server

# build docker image
docker build -t reltio-mcp-server .

# run with environment variables (example)
docker run -p 8080:8080 \
  -e RELTIO_BASE_URL="https://api.us.reltio.com" \
  -e RELTIO_CLIENT_ID="your-client-id" \
  -e RELTIO_CLIENT_SECRET="your-client-secret" \
  -e MCP_PORT=8080 \
  reltio-mcp-server

Local configuration (example .env):

RELTIO_BASE_URL=https://api.reltio.com
RELTIO_CLIENT_ID=your-client-id
RELTIO_CLIENT_SECRET=your-client-secret
MCP_PORT=8080
LOG_LEVEL=info

After starting the server, verify health:

curl http://localhost:8080/health
# expected: 200 OK or a JSON status response

Notes:

Keep credentials secure and rotate as required by your organization.
API usage counts against your Reltio subscription; monitor entitlements to avoid overages.

Available Tools

Key tools exposed by the MCP server (full list available in the repository):

Tool	Purpose
search_entities_tool	Search entities with filters and scoring
get_entity_tool	Fetch an entity by Reltio ID
update_entity_attributes_tool	Update attributes on an entity
find_potential_matches_tool	Find potential duplicates by rule/score/confidence
get_entity_match_history_tool	Retrieve match history for an entity
merge_entities_tool	Merge multiple entities into one
reject_entity_match_tool	Reject a potential match (mark as not-duplicate)
get_entity_graph_tool	Traverse entity graph / relationships
create_entity_tool	Create entities in a tenant
get_merge_activities_tool	Retrieve merge-related activity events

The server also includes tools for tenant metadata, RDM lookups, workflow tasks, and relation administration. Each tool is exposed to MCP clients as an action the client can request; the implementation delegates to Reltio APIs and optional plugins.

Use Cases

Assisted duplicate resolution: An LLM-based assistant queries potential matches, presents top candidates, and triggers merge or reject actions after human confirmation.
- Flow: client -> search_entities_tool -> find_potential_matches_tool -> human-in-the-loop -> merge_entities_tool
Bulk rule-based cleanup: Run scheduled jobs that call find_potential_matches_tool with a confidence threshold, then either auto-merge or produce a review queue for operators.
Context-enriched matching: Add a plugin that augments entity records with external reference data (CRM, billing) before scoring; the MCP server calls the plugin during the search flow.
Workflow integration: Trigger Reltio workflow process instances after merges or create interactions for audit purposes using start_process_instance_tool and create_interaction_tool.

Extensibility and Integration

Plugins: Implement new providers to augment data, modify scoring, or add custom tool handlers without changing core server code.
Clients: The MCP server is compatible with MCP-capable clients (desktop LLM UIs or custom agents). For local testing use any HTTP client to invoke the MCP endpoints or the provided sample clients in the repo.
Testing: The repository contains integration and unit tests; run the provided test scripts or containerized test targets to validate behavior in your environment.

For full tool descriptions, API examples, and plugin interfaces, consult the repository README and code comments in the GitHub project: https://github.com/reltio-ai/reltio-mcp-server.

Reltio MCP Server for Language Model Entity Matching