Claude Enterprise connector

Connect Claresia to Anthropic Claude Enterprise. End users invoke @claresia.<skill> from the Claude desktop app, Claude.ai web, or any Claude SDK-based client your org has built.

Prerequisites

Claude Enterprise tier (Team, Enterprise, or Anthropic API enterprise plan)
An Anthropic admin with rights to mint Admin API keys
Zero-retention enabled in your Anthropic Console
Your tenant_slug from Claresia (e.g., dainese)

Part 1 — Generate the Admin API key

Sign in to console.anthropic.com as an organization admin.
Navigate to Settings → API keys → Admin keys.
Click Create admin key.
- Name: claresia-distribution-{tenant_slug}
- Scopes: ensure Skills:write, Skills:read, Audit:read, Workspaces:read are checked.
Copy the key (shown once). Store securely; you’ll paste it into Claresia in 60 seconds.

Part 2 — Connect in Claresia

In Claresia Onboarding Portal, click Connect LLM (Step 6). Pick the Anthropic Claude Enterprise card.
Paste your Admin API key.
Click Test connection — Claresia calls GET https://api.anthropic.com/v1/organizations/me and shows your organization name + workspace count + zero-retention status.
Click Save & continue.

If Test connection returns:

✅ Success — you’re done.
⚠️ Zero-retention not enabled — Claresia refuses to publish. Enable zero-retention in Anthropic Console → Settings → Privacy.
❌ 401 Unauthorized — re-paste the key (most common: trailing space).
❌ 403 Forbidden — the key lacks one of the required scopes. Re-mint with the scopes listed above.

Part 3 — Skill publish flow

Once connected, Claresia automatically publishes your tenant’s skill catalog:

Reads the Skill IR JSON for every skill granted to one of your archetypes
Transpiles each into Claude’s Skill JSON format
POSTs to https://api.anthropic.com/v1/organizations/{org_id}/skills under the namespace claresia-{tenant_slug}-{skill_id}
Sets workspace + group ACLs to match your archetype RBAC
End users see the new skills appear in Claude’s @ mention picker within ~30s (publish lag SLO: <60s p99)

Part 4 — Telemetry ingest

Claresia polls Anthropic’s Audit API every 60s:

GET https://api.anthropic.com/v1/organizations/{org_id}/audit_logs
  ?event_types=skill.invoked,skill.failed
  &after_id=<last_seen_id>

For each event, we normalize into fn_telemetry_event and write to the Hub telemetry_event table. Surfaces in Command Center within 5 min p95.

End-user experience inside Claude

The user types @claresia in Claude’s input box. A picker shows the skills entitled to their archetype:

@claresia.gatespic.incident-postmortem
@claresia.steve.prd-review
@claresia.boss.standup-notes

User picks one, Claude prompts for parameters, then runs the skill. The reply is wrapped in Claresia’s Adaptive Card-equivalent (Claude renders it as markdown with a structured footer):

[skill output body]

---
*Claresia · Incident Postmortem · $0.04 · 2.3s · [View record](https://hub.dainese.claresia.com/records/rec_abc123)*

Click the View record link to open the Hub record with full provenance.

What if I have multiple Anthropic workspaces?

Claresia can publish to one or many workspaces in your Anthropic org. Map each workspace to a tenant or sub-tenant in Onboarding Portal → Hub → Workspace mapping. Common pattern:

Anthropic workspace	Claresia tenant
`dainese-engineering`	`dainese-eng` (sub-tenant of `dainese`)
`dainese-corporate`	`dainese-corp` (sub-tenant)
`dainese-sales`	`dainese-sales`

Sub-tenants share the same Hub but get separate skill catalogs + audit logs.

Common gotchas

Symptom	Cause	Fix
Skills don’t appear in `@` picker	Distribution publish lag	Wait 90s; if still missing, hit Republish all in Command Center
Telemetry never appears	Anthropic Audit API hasn’t ingested yet	Up to 5 min p95 — if longer, check Anthropic status page first
`403` on publish	Workspace ACL excludes Claresia’s service principal	Add the Admin key’s principal to the target workspace
End user sees `@claresia` skills they shouldn’t	Cached old publish	Run Republish all to force a fresh sync with current archetype grants