Integrations

MCP — Claude Integration

Reliant has a native MCP (Model Context Protocol) server. Connect it to Claude Desktop or Claude Code and control your entire reliability layer — schemas, guards, SLAs, contracts, and metrics — directly from chat, without writing code.


Installation

Install the Reliant MCP server globally via npm:

terminal
npm install -g reliant-mcp

Configuration

Claude Desktop

Add the following to your claude_desktop_config.json:

json
{ "mcpServers": { "reliant": { "command": "reliant-mcp", "env": { "RELIANT_API_KEY": "your-api-key", "RELIANT_USER_ID": "your-user-id" } } } }

Claude Code

Add the following to your .mcp.json in the project root:

json
{ "mcpServers": { "reliant": { "command": "reliant-mcp", "env": { "RELIANT_API_KEY": "your-api-key", "RELIANT_USER_ID": "your-user-id" } } } }
VariableWhere to find
RELIANT_API_KEYDashboard → Settings
RELIANT_USER_IDDashboard → Settings
RELIANT_API_URLOptional. Default: https://reliant-production.up.railway.app

Available tools

Once connected, Claude has access to 19 tools organized by category:

Execute

ToolDescription
reliant_executeExecute a prompt through a schema — returns guaranteed structured JSON with automatic retry and fallback.
reliant_list_executionsList recent executions with status, latency, provider, and schema.
reliant_inspect_executionInspect a specific execution in detail — full output, validation errors, and metadata.

Schemas

ToolDescription
reliant_list_schemasList all schemas in the project.
reliant_get_schemaGet details of a specific schema, including JSON definition and version history.
reliant_create_schemaCreate a new schema by describing the desired structure in natural language.

Guard

ToolDescription
reliant_guard_chatSend a user message through a Guard — validates topic, PII, and tone before returning the AI response.
reliant_list_guardsList all Guard configurations.
reliant_create_guardCreate a Guard by describing allowed topics, blocked keywords, and tone rules in plain English.
reliant_guard_usageCheck usage statistics for a Guard — executions and top end-users.

Contracts

ToolDescription
reliant_list_contractsList all Business Contracts in the project.
reliant_create_contractCreate a Business Contract with field rules and semantic rules evaluated by Claude.

SLA

ToolDescription
reliant_sla_statusCheck the SLA status of a schema — success rate, latency, and whether targets are being met.
reliant_sla_configureConfigure SLA targets for a schema and set a webhook for breach alerts.
reliant_sla_listList all SLA configurations and their current status.

Audit

ToolDescription
reliant_audit_recentShow recent audit trail events — executions, guard checks, contract violations, and limit events.

Metrics

ToolDescription
reliant_metricsGet project metrics — total executions, success rate, average latency, and token usage.
reliant_analyticsDetailed analytics broken down by schema and provider.
reliant_health_checkQuick health check — API status, schema count, today's success rate, and active SLA breaches.

Usage examples

Create and test a schema

chat with Claude
You: Create a schema called "Lead Scoring" with fields: score (number), tier (string: hot/warm/cold), reason (string) Claude: [calls reliant_create_schema] ✅ Schema created — ID: cmp... You: Test it with this text: "The lead opened 5 emails and visited pricing twice" Claude: [calls reliant_execute] { "score": 87, "tier": "hot", "reason": "High email engagement and pricing interest" }

Monitor your pipelines

chat with Claude
You: How are my AI pipelines doing today? Claude: [calls reliant_health_check + reliant_metrics] API: ✅ Online Schemas: 11 active Today: 243 executions, 98.8% success rate Avg latency: 612ms SLA breaches: none

Set up a Guard for a chatbot

chat with Claude
You: Create a Guard for our customer support bot. It can only discuss billing, shipping, and returns. Block competitor names. Always be professional. Claude: [calls reliant_create_guard] ✅ Guard created — ID: grd... Allowed: billing, shipping, returns Blocked keywords: [competitor names] Tone: professional

Tip: You can ask Claude to do everything in natural language — create schemas, configure SLAs, inspect failures, or run health checks — without opening the Dashboard or writing a single line of code.