Integrations

Settings → AI configures the AI provider behind drafting, natural-language search, completeness suggestions and impact narratives — each workspace brings its own provider and key rather than relying on a shared deployment-wide key. Choose Anthropic, OpenAI, Google (Gemini), or a custom OpenAI-compatible endpoint (a self-hosted model server like Ollama or vLLM, or a compatible gateway — not Azure OpenAI, which uses a different API shape), enter an API key, and optionally override the default model. "Test connection" makes one small real call to confirm the key and model work before you switch AI on for the workspace. Every AI feature keeps working with AI turned off or unconfigured — natural-language search falls back to keyword search, impact narratives fall back to a deterministic structural summary — so nothing hard-depends on it.

Settings → Connectors lets admins pull in records from an external CMDB/ITSM system as a source of truth for Applications/Technologies/Risks/Capabilities. Configure field mapping and an optional auto-sync schedule; when the remote system's data drifts from what's in Eidon, the drift is surfaced for an admin to apply the remote value or dismiss it, rather than being silently overwritten. The endoflife.date connector is a named preset for Technology lifecycle dates specifically — no URL, auth or field mapping to configure, since it's a public, unauthenticated API keyed off each Technology's name.

Choosing "ServiceNow" as the source type pre-fills the resource path with the matching CMDB Table API endpoint for the entity type you're syncing (e.g. cmdb_ci_appl for Applications), asking for display values so ServiceNow's choice fields come back as readable labels. Only extend the field mapping to a ServiceNow field that shares its vocabulary with the Eidon field it's mapped to — mapping a ServiceNow choice field like "operational_status" onto an enum-typed Eidon field like lifecycle_status will generally show up as drift you can't apply, since the two systems rarely share the same set of status values; applying a value that isn't valid for that field returns a clear error rather than corrupting the record.

"View history" on a connector shows the last 50 sync attempts (manual or scheduled), each with its start time, success/error status, record count, and — for a failed run — the error. This is a log across runs, unlike the connector list's own "Last sync" column, which only ever shows the most recent attempt.

Settings → LeanIX Import connects to a SAP LeanIX workspace via an API token and lets you preview, then import, fact sheets (applications, capabilities, IT components, data objects) into the catalog. Re-running an import is safe — records are matched and updated by their LeanIX id rather than duplicated.

Settings → Cloud Accounts connects AWS, Azure or GCP via a discovery endpoint and suggests matches against existing Applications/Technologies based on discovered resources. Nothing is auto-linked — an admin reviews each suggestion and chooses to match it to an existing record, create a new one, or ignore it.

Cost sync is a separate step on each connected account: configure that provider's billing credentials (an AWS access key pair, an Azure app registration's client credentials, or a GCP service account key plus its billing export dataset/table), then run "Sync costs now" to pull trailing-30-day spend from the provider's actual billing API. It only attributes cost to resources you've already confirmed a match for — an unmatched or merely-suggested resource has nothing to record cost against. Each sync adds a data point to Reports → Cost Trend rather than overwriting the record's own Annual Cost field, which stays under your manual control.

Settings → Shadow IT Discovery is different from SSO (Settings → SSO): SSO covers who can log into Eidon itself, while this queries your identity provider's own audit log (Okta System Log, or Microsoft Entra ID sign-in logs via Microsoft Graph) for *other* SaaS apps employees are actually signing into — visibility the catalog wouldn't otherwise have. Configure the provider's API credentials, then run "Scan now" to pull the trailing 30 days of cross-app sign-ins. Each app is auto-suggested a match against existing Applications by name, same as Cloud Accounts' resource discovery; review each one and confirm a match, create a new Application record, or ignore it. Discovered apps land in the workspace your SSO configuration's default workspace points to.

Settings → Webhooks sends outbound notifications when matching events occur (e.g. a specific entity change, or "*" for everything). Choose a channel type: generic (raw JSON, HMAC-signed if a secret is set) or Slack/Teams (a formatted chat message posted to an incoming webhook URL). A channel can subscribe to specific notification categories — recertifications, contract expiry, approvals, rule actions — so it only receives what it's meant to. Failed deliveries retry automatically with backoff, visible in a per-webhook delivery log; Slack/Teams channels may require the connectors plan feature.

Settings → SIEM Streaming forwards audit log entries in near-real-time to Splunk, Microsoft Sentinel, IBM QRadar, or a generic HTTP collector, in either JSON or CEF format — for organisations that centralise security event monitoring outside Eidon itself.

Settings → Confluence connects a Confluence Cloud (or Server/Data Center) site so reports and entity detail pages can be pushed there as a page that stays updated, rather than a one-time export. Configure the site's base URL, an account email, and an API token (Confluence Cloud's standard authentication), plus an optional default space key used when a page doesn't specify one of its own. "Test connection" confirms the credentials work before you rely on them.

Once configured, supported pages show an "Export to Confluence" action that sends the same content you'd otherwise download as Markdown or print to PDF. The first export creates a new page in the target space; every export after that updates that same page in place instead of creating a duplicate, and the button shows when it was last synced with a link to the live page. Removing a page's export link (via the API) stops tracking it — a later export then creates a fresh page rather than updating the old one. Exporting requires the connectors plan feature, the same one Jira/Azure DevOps work item linking uses.

Eidon exposes a Model Context Protocol (MCP) server at /mcp, so an AI assistant that supports MCP (Claude, Copilot, ...) can query a workspace's architecture catalog, relationships, and roadmap directly from a chat interface — read-only, the same data the public /v1 API offers under a different protocol. To connect one, create an API key under Settings → API Keys and make sure its allowed paths include /mcp (a key already scoped to /v1 for other tooling needs /mcp added explicitly), then point your assistant's MCP client configuration at https://your-eidon-host/mcp with that key as the X-API-Key header. No separate setup on the Eidon side — the same key that works for the REST API works here too.