Core as Infrastructure

Use Isomer Core directly via API, MCP, or CLI to power your own AI agents with Isomer's ingestion, ontology, and claim graph.

Isomer Core is available as infrastructure for teams building their own AI agents. If you have existing detection or workflow logic you want to keep, you can consume Core directly — same ingestion pipeline, same ACORD-aligned ontology, same claim graph, same security posture — without using Signals or Actions.

What you get

Access methods

Core exposes three access methods.

API

HTTP endpoints for programmatic access from any language or system. Use the API to query the claim graph, retrieve documents, list parties, and access timeline events.

# Retrieve a claim graph
curl -X GET https://api.isomer.ai/v1/claims/{claim_id} \
  -H "Authorization: Bearer {api_key}"

# List all parties on a claim
curl -X GET https://api.isomer.ai/v1/claims/{claim_id}/parties \
  -H "Authorization: Bearer {api_key}"

# Retrieve documents classified as legal correspondence
curl -X GET "https://api.isomer.ai/v1/claims/{claim_id}/documents?classification=legal" \
  -H "Authorization: Bearer {api_key}"

# Semantic search across claims using a natural-language query
curl -X POST https://api.isomer.ai/v1/claims/semantic-search \
  -H "Authorization: Bearer {api_key}" \
  -H "Content-Type: application/json" \
  -d '{"query": "rear-end collision with deferred medical treatment", "limit": 20}'

Semantic search returns claims ranked by relevance to a natural-language query, scored against an embedding of each claim's loss description. Use it when an identity match (claim number, claimant name, policy number) isn't available or when you want to surface conceptually similar claims. Identity-based lookup remains available at POST /v1/claims/search.

MCP

Model Context Protocol is the canonical way to wire an external AI agent into Core. The MCP server exposes Core's capabilities as tools — your agent calls them directly without managing API auth in application code.

# Install the Isomer MCP server
npm install @isomer-ai/mcp

# Or via pip
pip install isomer-mcp

Configure your agent to connect to the MCP server:

{
  "mcpServers": {
    "isomer": {
      "command": "npx",
      "args": ["@isomer-ai/mcp"],
      "env": {
        "ISOMER_API_KEY": "your_api_key"
      }
    }
  }
}

Available MCP tools mirror the Core API: get_claim, list_parties, list_documents, get_timeline, search_claims, and more.

The search_claims tool accepts a mode of either identity (weighted match on claim number, policy number, claimant name, loss date, and policyholder) or semantic (natural-language query against the loss description). The mode is selected per call:

// Identity search
{
  "mode": "identity",
  "searchParams": { "claimNumber": "CLM-00123" }
}

// Semantic search
{
  "mode": "semantic",
  "searchParams": { "query": "rear-end collision with deferred treatment", "limit": 20 }
}

CLI

For scripted workflows, one-off operations, and terminal-driven access.

# Install the CLI
npm install -g @isomer-ai/cli

# Authenticate
isomer auth login

# Fetch a claim graph
isomer claims get CLM-00123

# List recent signals on a claim
isomer signals list --claim CLM-00123

# Export claim data
isomer claims export CLM-00123 --format json

Authentication

All three access methods use the same API key. Generate keys in Settings → API keys.

Access tokens are scoped and time-limited. Isomer does not store persistent service-account credentials — see Security for details.

When to use Core directly

Core as infrastructure is designed for teams that:

• Have existing detection models or rules they want to run on structured claim data
• Are building AI agents in-house and want Isomer's ingestion and ontology without adopting Signals or Actions
• Need programmatic access to the claim graph for custom reporting, integrations, or downstream systems

If you want pre-built risk detection, use Signals. If you want pre-built workflow execution, use Actions. Core is the right choice when you're bringing your own logic.

Next steps