Skip to Content
MCP Server

MCP Server

The Strise MCP server exposes Strise’s compliance intelligence to AI assistants that speak the Model Context Protocol . Connect Claude or any other MCP-aware client to your Strise tenant and run KYB/KYC, screening, monitoring, and review workflows directly from chat.

The server uses the same data, authorization, and audit trail as the Strise web app. Every tool call goes through your Strise team’s permissions and is logged in Compliance Memory.

Server endpoint

FieldValue
Streamable HTTP URLhttps://mcp.strise.ai/mcp
Discoveryhttps://mcp.strise.ai/.well-known/oauth-authorization-server
Tool reference (markdown)https://mcp.strise.ai/docs
Health checkhttps://mcp.strise.ai/healthz

Authentication

The server implements OAuth 2.0 with Dynamic Client Registration (RFC 7591) and PKCE (S256). MCP clients register themselves at runtime — there is no pre-shared client_id or client_secret.

OAuth propertyValue
Grant typesauthorization_code, refresh_token
Token endpoint authclient_secret_post, none (public PKCE)
Scopesopenid, profile, email
Identity providerAuth0 (same login as the Strise web app)

Each request to /mcp must include a verified Strise access token as Authorization: Bearer <token>. The user must belong to at least one Strise team. The first team is used by default; pass X-Strise-Team: <teamId> to target a specific team.

Setup

Claude (web and desktop)

  1. Open the Claude connector directory .
  2. Add the Strise connector.
  3. Sign in with your Strise account when prompted.
  4. The Strise tools become available in Claude.

Claude Desktop (manual config)

Add the server to claude_desktop_config.json:

{
  "mcpServers": {
    "strise": {
      "url": "https://mcp.strise.ai/mcp"
    }
  }
}

Claude Desktop handles the OAuth flow automatically the first time the server is contacted.

Cursor / other MCP clients

Any client that supports remote MCP servers with OAuth can connect to https://mcp.strise.ai/mcp. Refer to your client’s documentation for the exact configuration format. Clients that only support local stdio servers can run the server from source — see the README .

Capabilities

Tools are grouped by surface area. The full reference with parameters, return shapes, and per-tool example prompts lives at mcp.strise.ai/docs .

CategoryWhat it covers
Search & lookupSearch companies, business persons, and private persons by name or registration number; lookup by national identifier
Entity detailsCompany, person, and private-person profiles — addresses, identifiers, citizenship, industry, corporate purpose
Risk & complianceCalculated Risk score, sanctions screening, PEP screening, adverse media (AMS) events, review history
Company dataFinancials, ownership, holdings, real estate, collateral, credit and CCJ reports, registers, documents, branches
Entity structureBeneficial ownership (UBO) with configurable threshold, ownership charts, board roles, signatories, connected entities
Monitoring & alertsList alerts, get alert change details, portfolio briefing, resolve alerts
Review managementCreate company and private-person reviews with audit risk and review status
Adverse mediaDisposition AMS events as confirmed true/false; create custom AMS events
EditingEdit ownership and roles, manage custom registers, connect or disconnect custom entities, update entity status
WorkflowsOrder credit and CCJ reports; send customer-facing forms; log team activities

Example workflows

Compliance review of a Norwegian company

Look up the Norwegian company with org number 923609016. Give me the calculated risk score, beneficial owners at 25% threshold, and any sanctions or PEP hits.

The model calls lookup_company, get_entity_risk_summary, and get_entity_ownership, then synthesizes a written summary.

Sanctions and PEP screening of a person

Search for “Ivan Petrov” and check whether any match has sanctions or PEP flags.

The model calls search_entities (or search_private_persons), then get_entity_risk_summary for each promising match.

Portfolio alert triage

Show this week’s high-severity monitoring alerts and explain what triggered each.

The model calls get_portfolio_briefing and then get_alert_changes for individual alerts to surface the underlying change.

Onboard and approve a customer

Approve company abc-123 as low risk with the comment “KYC verified” and set status to ACCOUNT.

The model calls get_entity_risk_summary to confirm the current risk level, create_company_review with reviewStatus: APPROVED, auditRisk: LOW, and update_entity_status to set the final status.

Add a private person to perpetual monitoring

Search for “Jane Doe” in Sweden, screen her, and add her to perpetual monitoring.

The model calls search_private_persons, get_entity_risk_summary, and add_private_person_to_monitoring.

Working with the MCP and Connect API together

The MCP server is an opinionated wrapper around the Strise Connect GraphQL API. Use the MCP server when an LLM agent should drive the workflow; use Connect directly when you need full control over the GraphQL query shape, custom batching, or response fields not exposed via tools. The two are interchangeable for the underlying data — the same monitoring rules, screening sources, and audit trail apply.

Limits and behaviour

  • Tool surface is curated. The MCP server exposes a subset of Connect that is safe for agentic use. Operations that change tenant-wide configuration are not exposed.
  • Destructive tools are marked. Tools that delete data (custom registers, custom entity connections, ownership/role resets, status removal, monitoring removal) are flagged as destructive: true in the tool reference. Models should require explicit confirmation before calling them.
  • Sessions are stateless across reconnects. Each initialize produces a new session ID. Tokens and team context are resolved per request.
  • Read-after-write consistency. Mutations write to Spanner and return the updated resource; subsequent reads observe the change immediately.

Support

Last updated on
Strise Connect APISettings Pages