Skip to main content
Glama
GAIIA Expert Proxy (MCP Server)

GAIIA Expert Proxy (MCP Server)

GAIIA Unified MCP Server (v2.0.0)

The GAIIA Unified MCP Server is an enterprise-grade Model Context Protocol (MCP) server that connects AI coding assistants and autonomous agents to the GAIIA ecosystem. It enables expert-guided code audits, project transformations, BPMN process triggering, and multi-protocol API interrogation.


🌟 Capabilities & Features

  • Expert Selection & Manifest Routing: Browse and select from a registry of Proxy Experts (e.g. Clean Architecture, Security, Performance, Transit domain).

  • Code Transformation & Auditing: Transform individual code blocks or conduct repository-wide audits against immutable architectural standards.

  • Multi-Protocol API Interrogation: Fuzz and discover schemas across diverse protocols:

    • REST (JSON): OpenAPI v3 synthesis.

    • GraphQL: Schema introspection and automatic tool generation.

    • gRPC (Protobuf): Server reflection inspection (src/mock/mock-grpc-server.ts).

    • ISO 8583 (TCP Sockets): Raw socket communication for payment gateways (src/mock/mock-tcp-server.ts).

    • XML / SOAP: Automatic .xsd and .wsdl parsing.

    • EDI (ANSI X12): Segment repair for legacy enterprise supply chain formats.

    • OData & ERP: Enterprise metadata resolution (src/mock/mock-erp-server.ts).

  • BPMS Process Orchestration: List BPMN blueprints and trigger workflow executions in gaiia-process-management.

  • Telemetry & Context Retention: In-memory context tracking and telemetry logging via src/services/telemetry-service.ts.


Related MCP server: docsgrep

βš™οΈ Installation & Build

Prerequisites

  • Node.js: v20+ LTS

  • npm: v9+

  • A registered GAIIA account for cloud registry sync

Setup

# 1. Navigate to directory
cd gaiia-mcp-server

# 2. Install dependencies
npm install

# 3. Build TypeScript
npm run build

πŸ” Authentication

Authenticate with the GAIIA platform to cache session tokens locally:

npm run login

Tokens are cached securely at ~/.gaiia/auth.json.


πŸ”§ Environment Configuration (.env)

GAIIA_GRAPHQL_ENDPOINT=https://<api-id>.appsync-api.ap-southeast-2.amazonaws.com/graphql
AWS_REGION=ap-southeast-2
USER_POOL_ID=ap-southeast-2_xxxxxxxxx
APP_CLIENT_ID=xxxxxxxxxxxxxxxxxxxxxxxxxx

πŸ”Œ MCP Client Configuration (e.g., Claude Desktop)

Add the server to your claude_desktop_config.json:

{
  "mcpServers": {
    "gaiia-unified-mcp": {
      "command": "node",
      "args": ["c:/programming/aiia/gaiia-mcp-server/build/index.js"]
    }
  }
}

πŸ› οΈ Available MCP Tools Reference

Tool Name

Purpose

Key Arguments

gaiia_list_experts

Lists all available AI experts in the registry

query (optional string: name, email, or architecture keywords)

gaiia_set_active_expert

Sets the active expert for transformations

email (required string)

gaiia_transform

Audits, refactors, or generates code with the active expert

code (string), instructions (string)

gaiia_analyze_project

Performs deep architectural audit or automated refactor on a directory

directory_path (string), mode ("audit" | "refactor")

gaiia_ingest_project

Ingests directory code samples into active expert memory

directory_path (string)

interrogate_endpoint

Multi-protocol API fuzzing and schema synthesis

url (string), method (string), auth_header (optional), base_payload (optional)

sync_specs

Synchronizes locally discovered schemas in specs/ to cloud registry

None

gaiia_list_processes

Lists registered BPMN templates in the workspace

None

gaiia_start_process

Triggers a BPMN workflow run with initial context

process_type (string), payload (optional object)


πŸ§ͺ Local Testing & Multi-Protocol Mock Servers

The server includes standalone mock servers in src/mock/ for testing complex protocols without external dependencies:

# Start mock gRPC server
npx tsx src/mock/mock-grpc-server.ts

# Start mock ISO 8583 TCP socket server
npx tsx src/mock/mock-tcp-server.ts

# Start mock ERP / OData server
npx tsx src/mock/mock-erp-server.ts

πŸ“„ License

MIT Β© 2026 GAIIA Engineering

Available Tools

8 tools
gaiia_analyze_projectC

Perform a deep architectural audit of an entire local project directory.

Inputs: directory_path (required), mode.

gaiia_list_expertsC

List available AI experts and their specialties.

Inputs: query.

gaiia_list_processesA

List all registered BPMN process blueprints/templates in the GAIIA registry for the authenticated tenant.

Inputs: no named parameters.

gaiia_set_active_expertC

Select an expert to use for code transformations.

Inputs: email (required).

gaiia_start_processB

Starts/triggers an instance of a registered BPMN process blueprint with an optional custom payload.

Inputs: process_type (required), payload.

gaiia_transformC

Audit, refactor, or generate code using the active GAIIA expert.

Inputs: code (required), instructions (required).

interrogate_endpointB

Intelligently interrogates a REST endpoint to discover its schema via reinforcement learning loop.

Inputs: url (required), method (required), auth_header, base_payload, +1 more.

sync_specsA

Synchronizes all locally discovered API specifications in the 'specs/' directory to the GAIIA Registry.

Inputs: no named parameters.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

    TDQS

    B3.4/5.0

    Scored across 8 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: project analysis, expert listing/selection, process listing/starting, code transformation, API interrogation, and spec syncing. No significant overlap.

    Naming Consistency3/5

    Six of eight tools use the consistent 'gaiia_' prefix with verb_noun pattern, but two tools ('interrogate_endpoint', 'sync_specs') lack the prefix, creating a mixed naming convention.

    Tool Count5/5

    With 8 tools, the set is well-scoped for a proxy server providing GAIIA platform operations, covering multiple areas without being overwhelming.

    Completeness4/5

    The tool surface covers core workflows (expert management, project analysis, process execution, code transformation, API spec handling), but lacks a get_active_expert tool and some CRUD operations, leaving minor gaps.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers