MCP Server

MCP Server Configuration via UI

Integrate with Fynapse MCP server

Fynapse MCP Server allows you to connect MCP-compatible AI assistants, such as Claude Code, directly to your Fynapse instance so they can work with your data through the Model Context Protocol (MCP). You control what can connect through ‘Consumer Profiles’. Until you create a profile and switch it on, only AI clients running locally can reach your instance.

Each instance gets its own secure /mcp endpoint where you register and authenticate via OAuth.

What Can the AI assistant Do?

The AI assistant can answer questions about your data model and operational health using seven read-only tools that can:

  1. Browse namespaces
  2. List entities
  3. Retrieve full entity schemas
  4. Query operational metrics for ingestion, processing, and throughput
  5. Report per-day journal counts over a date range
  6. Fetch processing statistics for an ingestion batch
  7. Check the status of a specific ingestion

In practice this means you can ask e.g. “what entities exist in this namespace and how are they structured?” or “how did yesterday’s ingestion perform?” and get answers grounded in live instance data — without writing a query or opening a dashboard.

MCP Settings Screen

The MCP Settings screen allows you to browse and edit existing consumer profiles and create new ones.

To create, edit and browse consumer profiles, your Fynapse users has to have either the Fynapse System Administrator or the Fynapse Security Administrator role assigned.

This article covers the Fynapse side of the configuration. For information how to enable, AI clients, e.g. Claude, go to the client’s documentation, e.g. https://claude.com/docs/connectors/building/authentication#network-reference.

Consumer Profile

Each consumer profile has the following details:

  • Name - name of the consumer profile
  • Allowed switch - determines if the AI client can connect, enabled, or if the access is disabled. Turn it off to pause access without deleting the consumer profile.
  • Redirect URIs - the web address the AI client returns to after a user signs in. Paste the callback URL given by your AI client here — it has to match
    exactly. Use the + Add additional URL button if a client needs more than one.
  • IP egress ranges - Limits connections to specific networks, entered
    as CIDR ranges (for example, 10.0.0.0/24). Only clients connecting from a
    listed range can get in. Use the + Add additional CIDR button to add more if required.
  • Description - a note to help you remember what the profile is for

The redirect URIs can also accept wildacrds. This can be useful e.g. for OpenAI, where you need to define https://closeai.ai/oauth/*/redirect, since OpenAI generates a unique identifier that is used a part of the URI path.

Each consumer profile has own Delete profile and Save changes buttons.

Tutorials

  1. Go to System Configurations > MCP Settings.
  2. On the MCP Settings screen click the + New profile button.
  3. Type a meaningful name for the consumer profile.
  4. You can add an optional Description.
  5. The Allowed switch is by default enabled. This means your AI client can connect to the MCP server.
  6. Click the Create button to save changes to your Business Event Type, or the Cancel button to discard your changes.
  7. Open the tab with your consumer profile.
  8. Add Redirect URIs.
  9. Add the IP egress ranges.
  10. Click the Save changes button to save your configuration.

First you need to configure Claude.AI Redirect URIs and IP ranges. These are documented in th Anthropic documentation: here and here.

  1. Go to System Configurations > MCP Settings.
  2. On the MCP Settings screen click the + New profile button.
  3. Type “Anthropic” as the name for the consumer profile.
  4. You can add an optional Description.
  5. Leave the Allowed switch enabled. This means your AI client can connect to the MCP server.
  6. Click the Create button to save changes to your Business Event Type, or the Cancel button to discard your changes.
  7. Open the tab with your consumer profile and fill in the details from the Anthropic documentation.
    1. Add Redirect URIs as per Anthropic documentation.
    2. Add the IP egress ranges as per Anthropic documentation.
  8. Click the Save changes button to save your configuration.
  9. Navigate to claude.ai and follow the path + > Connectors > + Add connector > Add custom connector.
  10. Type name Fynapse and [https://your.fynapse.address/mcp]
  11. Once added, open the connectors and click on the newly added connector to authenticate.
  12. The authentication will open a new browser window and navigate to Fynapse login page.
  13. Once your credentials are provided (or SSO option is chosen) and the authentication is sucesfull, you’ll be navigated back to Claude where you can start using Fynapse MCP server, by mentioning it in the prompt: e.g. What attributes are defined for my Journal entity in Fynapse.