OpenAPI, SPARQL, MCP & Custom Integrations

hub Overview

Beyond the built-in tools, LexiChat lets you connect external data sources as callable tools. There are four ways to do this:

description OpenAPI Import

LexiChat parses OpenAPI 3.x specifications (JSON or YAML). Each operation (GET /users/{id}, etc.) becomes one tool the AI can call.

Example OpenAPI spec

Here's a minimal spec that exposes a weather lookup endpoint. Paste this to try it out:

{
  "openapi": "3.0.0",
  "info": {
    "title": "Weather API",
    "version": "1.0"
  },
  "servers": [
    { "url": "https://api.open-meteo.com/v1" }
  ],
  "paths": {
    "/forecast": {
      "get": {
        "operationId": "get_weather",
        "summary": "Get weather forecast for a location",
        "parameters": [
          {
            "name": "latitude",
            "in": "query",
            "required": true,
            "schema": { "type": "number" },
            "description": "Latitude of the location"
          },
          {
            "name": "longitude",
            "in": "query",
            "required": true,
            "schema": { "type": "number" },
            "description": "Longitude of the location"
          },
          {
            "name": "current_weather",
            "in": "query",
            "required": false,
            "schema": { "type": "boolean" },
            "description": "Include current conditions"
          }
        ]
      }
    }
  }
}

After importing, the AI can answer questions like "What's the weather in London?" by calling get_weather with the coordinates it looks up.

build Custom Tools

For simple HTTP integrations that don't have an OpenAPI spec, use the Custom Tool builder to define an endpoint by hand.

Example: GitHub Repository Info

A custom tool to fetch public GitHub repo metadata:

Once registered, you can ask: "How many stars does the ollama/ollama repo have?" and the AI will call the tool automatically.

// AI generates this call
{
  "name": "get_github_repo",
  "arguments": {
    "owner": "ollama",
    "repo": "ollama"
  }
}

// LexiChat makes: GET https://api.github.com/repos/ollama/ollama
// Returns JSON → fed back to model

hub SPARQL Endpoints

SPARQL endpoints expose linked data (RDF) — UK government open data, scientific datasets, knowledge graphs, and more. Unlike REST APIs, a SPARQL service has no machine-readable contract describing its data, so instead of importing a spec you assemble a small bundle that tells the model how to query it: the endpoint URL, namespace prefixes, a short schema summary, and a few example queries.

Endpoint fields

Example: HM Land Registry

The built-in HM Land Registry endpoint (https://landregistry.data.gov.uk/landregistry/query) is set up to answer UK property questions. With it enabled you can ask:

"Use HM Land Registry to find recent sold prices in postcode PL6 8RU."

The AI calls hm_land_registry_query with a query like its bundled example:

PREFIX lrppi: <http://landregistry.data.gov.uk/def/ppi/>
PREFIX lrcommon: <http://landregistry.data.gov.uk/def/common/>
SELECT ?date ?price ?street ?town WHERE {
  ?txn lrppi:propertyAddress ?addr ;
       lrppi:pricePaid ?price ;
       lrppi:transactionDate ?date .
  ?addr lrcommon:postcode "PL6 8RU" .
  OPTIONAL { ?addr lrcommon:street ?street }
  OPTIONAL { ?addr lrcommon:town ?town }
} ORDER BY DESC(?date) LIMIT 20

If the model is unsure of the vocabulary it first calls hm_land_registry_schema to retrieve the prefixes and examples, then writes the query.

dashboard_customize MCP Servers

Model Context Protocol (MCP) servers expose whole toolsets — filesystems, databases, SaaS integrations, and more. LexiChat connects to them over stdio (a local command) or HTTP, discovers their tools, and makes them callable by the model.

Expand any server to enable/disable individual tools, and remember: MCP selections are per-profile. See Managing Tools.

MCP Apps interactive UI

MCP servers can go beyond text: with MCP Apps (the SEP-1865 extension, built with the MCP-UI and OpenAI Apps SDK teams) a tool can return an interactive HTML interface — a form, dashboard, chart, viewer, or generator — that renders directly in the conversation. LexiChat supports both the official ext-apps and the community MCP-UI dialects.

Try it with the ready-to-run examples in the MCP ext-apps repo (e.g. a QR generator, PDF viewer, or map). Add the server, enable apps, and ask the model to use its tool.

lock Authentication

When adding an OpenAPI spec, SPARQL endpoint, or MCP connection, you can configure authentication. LexiChat supports four methods:

settings_applications Managing Tools

All tools (built-in, OpenAPI, and MCP) are managed in the Admin panel. From here you can: