ChapsVision

Sinequa MCP Server

A standardized bridge connecting AI agents to enterprise search — enabling secure, intelligent document retrieval without custom integrations

External Agents

Claude
ChatGPT
Copilot
ChapsAgent *Q2 2026

Any MCP-compatible agent can connect

Sinequa Search Platform

MCP Server

Standardized translation layer exposing search capabilities to any agent

Company Internal Documentation
Content & Structured Metadata
User-Level Security
Company-Specific Jargon

Full retrieval capabilities with hybrid search: a combination of statistical, semantic and multimodal search

Five Focused Tools

Each tool has a single responsibility — the agent orchestrates them to accomplish complex tasks

Search

The core retrieval tool that enables agents to access your company's internal documentation with precision and flexibility. This tool transforms any agent into a search expert with capabilities that go beyond traditional search interfaces.
Parallel queries: Run multiple searches simultaneously to explore different angles
Configurable output: Get metadata-only lists or full document content based on your needs
Scoped filtering: Reduce search scope with precise filters to target specific document sets
Relevance tuning: Iteratively fine-tune which content to prioritize

Under the Hood

Powered by Sinequa's proven search technology with advanced hybrid search combining statistical, semantic and multimodal approaches for exceptional relevance. Security is guaranteed at the platform level — users only see documents they have permission to access, respecting the original source permissions.

Get Search Instructions

Returns orchestration rules, catalogs of labels, filters, and sources available in your data
Call First

Get Filter Definition

Returns syntax and validation rules for a specific filter field
Before Filtering

Validate Filter

Normalizes user input to canonical index values with fuzzy matching
If Required

Clarify Jargon

Resolves company-specific terminology before querying
When Ambiguous

Typical Orchestration Flow

1

Initialize

Call Get Search Instructions once at session start

2

Clarify

Use Clarify Jargon if query contains unknown terms

3

Prepare Filters

Get definition → validate value → apply to search

4

Search

Execute with queries, labels (soft), filters (hard)

5

Iterate

Refine and re-search until results satisfy the question

Design Principles

Why the MCP server is architected this way

Separation of Concerns

Retrieval stays in Sinequa. Reasoning stays with the agent. The MCP server is a pure retrieval layer that exposes the full power of the search platform without making any LLM calls. Swap agents anytime without rebuilding your search infrastructure.

Cautious Filtering, Aggressive Guidance

Labels boost and demote but never exclude documents. Filters are hard constraints applied only when explicitly justified and validated. This prevents agents from accidentally filtering out the right answer while still allowing precise scoping when needed.

Use Case Agnostic

The same MCP server can be plugged into different data sets, powering various use cases. The server uses no hard-coded values — everything available comes directly from the indexed data and adapts dynamically as you index more documents.

Key Benefits of Sinequa MCP Server

1

Agent Agnostic

Works seamlessly with Claude, ChatGPT, Copilot, or any MCP-compatible agent. One integration serves all agents, no custom development needed.

2

Turns Agents into Search Experts

Any agent can leverage your internal data with expert-level search capabilities. Agents can adapt search scope and fine-tune relevance to find exactly what's needed.

3

Security Enforced by Sinequa

Security reflects the original source permissions. Users only see what they're authorized to access, guaranteed at the search platform level.