Model Context Protocol

Ground every assistant. Ship every agent.

SupportLogic exposes its 16 agents, 40+ signals, and full customer context to Claude, ChatGPT, Gemini, and any AI assistant that speaks MCP. That means the answers your team and customers get are real, cited, and current. Plus a Snowflake-native Data Cloud and a full REST API for everything else you’ll build.

System architecture
🤖
AI Client
Claude, ChatGPT…
🔐
MCP Gateway
Auth + policy
🗄️
Support Data Lake
Enriched signals
Zero-trust auth Audit logging Granular scoping JSON-RPC 2.0
Claude
cURL
Python
Snowflake SQL
// In any MCP-compatible assistant, just ask:

"What's going on with the ACME-Corp account right now?"

// → Claude calls the SupportLogic MCP server
// → Gets grounded, cited context
// → Responds with real signals

tool: "supportlogic.account.health"
input: { account: "acme-corp" }

→ returns:
{
  health_score: 42,
  trend: "declining",
  open_escalations: 3,
  top_signals: ["frustration", "competitive_threat"],
  recommended_action: "executive_outreach"
}
# Direct REST call — no MCP client needed

curl -X POST https://api.supportlogic.com/v1/accounts/health \
  -H "Authorization: Bearer $SL_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "account": "acme-corp",
    "window": "last_30d",
    "include": ["signals", "escalations", "recommendations"]
  }'
# Python SDK

from supportlogic import SupportLogic

sl = SupportLogic(api_key=os.environ["SL_API_KEY"])

# Subscribe to escalation signals
for alert in sl.alerts.subscribe(
    signal_types=["escalation_predicted", "churn_risk"],
    min_score=75
):
    notify_slack(alert.account, alert.recommended_action)
-- Query SupportLogic signals directly from Snowflake

SELECT
    account_name,
    health_score,
    escalation_probability,
    top_signal,
    last_updated
FROM supportlogic.account_signals
WHERE health_score < 50
  AND escalation_probability > 0.7
ORDER BY last_updated DESC;
Native MCP support for
Claude ChatGPT Gemini Microsoft Copilot Cursor Custom Apps

AI assistants without context are just guessing.

A general-purpose LLM doesn’t know your customers, your cases, or your account history. MCP changes that — and SupportLogic gives you the cleanest signal layer in the industry to plug into it.

01

Grounded answers

Every response cites the case, call, signal, or KB article it came from. No hallucinated customer history. No invented sentiment scores.

02

Persistent context

The Context Agent maintains memory across years of relationship — not just the current LLM token window. Every assistant gets the long view.

03

Permissioned access

MCP respects the access controls of your source systems. Account executives see their accounts. Engineers see their cases. Nothing leaks.

04

Single truth

Claude, ChatGPT, your in-house apps, and your CRM widgets all see the same source of truth. No more “which dashboard is right?”

Every agent, every signal. One protocol.

The SupportLogic MCP Server exposes a curated set of tools that map cleanly to how AI assistants think about customer context. Browse the catalog. Pick what your assistant needs.

Account & Customer Context
supportlogic.account.health read
Real-time account health score, trend direction, top contributing signals, and recommended actions.
supportlogic.account.summary read
Executive summary of an account across all open and recent cases, calls, and conversations.
supportlogic.account.escalations read
List of active and predicted escalations for an account, with severity and timeline context.
supportlogic.account.history read
Long-form relationship history across years, channels, contacts, and systems of record.
Case Intelligence
supportlogic.case.summary read
GenAI summary of a case with current status, resolution path, next steps, and tags.
supportlogic.case.signals read
All extracted signals from a case across customer, commercial, product, and agent dimensions.
supportlogic.case.related read
Related cases the Knowledge Agent identifies as similar in symptom, root cause, or resolution.
supportlogic.case.route write
Trigger the Routing Agent to recommend or assign the best engineer for a case.
Knowledge & Resolution
supportlogic.knowledge.search read
Precision RAG search across internal and external KB with source citations on every result.
supportlogic.knowledge.answer read
Natural-language answer to a customer or agent question, cited from KB and case history.
supportlogic.kb.draft write
Draft a new KB article from a resolved case, ready for human review and publication.
supportlogic.reply.draft write
Draft a personalized reply to a case in the customer’s voice and the agent’s tone.
Signals & Alerts
supportlogic.signal.list read
Query signals across the install base — by type, score, account, or time window.
supportlogic.alert.subscribe write
Subscribe to proactive alerts on escalation prediction, churn risk, or any custom rubric.
supportlogic.trend.detect read
Surface emerging trends and themes across thousands of cases — product feedback, root causes, feature requests.
supportlogic.voice.analyze read
Analyze tone, sentiment, hold time, and resolution from a voice call recording or transcript.

Executive briefings

Ask Claude or ChatGPT for a Monday-morning rundown of the top 5 at-risk accounts. Get a cited, current answer in 30 seconds — not a 90-minute prep call.

“Top 5 churn-risk accounts this week, with the signals driving the score and the recommended action for each.”

Slack copilots

Drop a bot into your support channels that answers questions about cases, accounts, and resolutions — using your real data, not a generic LLM’s guess.

“@supportlogic what changed on case 134791 in the last 24 hours?”

Vibe-coded tools

An engineering manager builds a custom dashboard in Cursor on a Friday afternoon. The MCP Server makes the data layer trivial. The agent layer comes free.

“Build me a Streamlit app that shows escalations by region and lets me drill in by signal type.”

Customer-facing assistants

Power a help-portal assistant that answers customer questions using your KB and their own case history — cited, accurate, on-brand.

“What’s the status of my support cases, and what should I expect this week?”

CRM widgets

Embed a SupportLogic widget inside Salesforce, Zendesk, ServiceNow, or Jira. Agents get the same MCP-grounded answers without leaving their primary tool.

“Summarize this account. Tell me the next best action.”

Automated workflows

Subscribe to signal streams and trigger downstream actions — escalation review, exec alerts, account-team handoff — with no engineering overhead.

“When escalation_probability > 0.7 AND deal_size > $250K, page the account team.”

Two more open surfaces for everything else you’ll build.

MCP is the right protocol for AI assistants. But not everything is an AI assistant. The Data Cloud and REST API give your data team and your engineering team first-class access to the same substrate.

Data Cloud

A Snowflake-native system of record for every signal, score, and prediction SupportLogic generates. Query it from your warehouse. Join it with your billing data. Power your BI tools, your CDP, your CS workflow engine.

  • Snowflake-native. Use your own instance or one we provide.
  • Read + write. Push signals back into Salesforce, Gainsight, anywhere.
  • BI-ready. Tableau, Looker, ThoughtSpot, Hex — all work out of the box.
  • Customer-facing. Power a hyper-personalized customer portal.
Learn about Data Cloud →

REST API

For everything that isn’t an AI assistant: webhook subscriptions, batch jobs, automation platforms, custom backend systems. Every MCP tool has a REST equivalent. Auth is OAuth or bearer token. Rate limits scale with your tier.

  • Full coverage. Every signal, score, and agent action is API-accessible.
  • Webhooks. Subscribe to events — no polling required.
  • SDKs. Python, Node, Go, Ruby. TypeScript types for everything.
  • Versioned. Deprecations announced 12 months in advance.
View API reference →

Single-Tenant VPC

Models run inside a Virtual Private Cloud dedicated to your data. Inference never leaves your boundary. No multi-tenant LLM shortcuts.

Zero-Copy Architecture

Source data is never replicated to shared infrastructure. The Data Extraction Agent reads, normalizes, and processes — without copying out.

Advanced Data Protection

Row- and column-level filtration. PII and sensitive field redaction. SHA-256 / RSA encryption. OAuth and TLS 1.2+ throughout.

SOC II Type 2 ISO 27001 GDPR HIPAA FIPS 140-2 2FA Bastion Host Access

From zero to grounded AI in three steps.

If SupportLogic is already deployed on your data, you can wire the MCP Server into Claude or ChatGPT in under five minutes.

01

Get your endpoint

In the SupportLogic Command Center, navigate to Settings → MCP Server → Generate Credentials.

mcp.supportlogic.com/
v1/your-tenant-id
02

Add to your assistant

Paste the endpoint and bearer token into Claude, ChatGPT, or any MCP client. Done — no engineering required.

claude mcp add supportlogic
  –url …
  –token …
03

Ask anything

“What’s going on with ACME-Corp?” “Top 5 at-risk accounts this week.” “Draft a reply to case 70577.” Claude does the rest.

# tools auto-discovered
16 tools available
✓ ready

Stop building data wrappers for every AI feature.

SupportLogic gives you the substrate. Your team builds on top — in Claude, in Cursor, in Slack, in whatever framework ships next quarter.