Inside the Integration Directory Spring 2025 Update

TL;DR: You can now browse every MCP integration inside your workspace, search by tag, and view live health plus configuration status. Everything flows through the /api/integrations endpoint -no spreadsheets, no hard-coded connectors. Here’s what shipped and what’s next.

What shipped

1. MCP-native registry (no hard-coded services)

The directory calls /api/integrations on load, mapping each record to ID, description, tags, and config state. If the API is unavailable, the UI falls back to canonical entries -Analytics Aggregator, LinkedIn MCP, X/Twitter MCP, and Web Fetch -so admins always see a baseline. That mirrors our “no dummy data” principle: either real sources or transparent failure.

2. Search and tag filters

Search covers name, description, and tags. Want to surface “analytics” or “b2b”? Type once, agents update the grid instantly. Under the hood, we normalise tag casing before comparison, making it trivial to cluster integrations for the Community Challenge Engine or the Agent-Led ASO sprint.

3. Health + configuration signals

Each integration block shows:

• Enabled: whether the MCP server is available according to /api/integrations.
• Authentication required: flagged in the UI so ops knows to provision credentials before missions run.
• Config enabled: tells builders if a server supports additional parameters (think scoped analytics).

These signals sync with our Approvals Guardrails. When health flips to false, Product Brain can pause dependent missions automatically.

4. Resilient loading states

When the fetch call hangs, the directory shows a branded loading indicator (we use the animated spinner shipped in this release). If the call fails, the fallback array renders with a console warning -not ideal, but better than blank screens.

Roadmap snapshot

We’re also exploring an “Add to mission” button -linking directory entries to templated recipes in the AI Launch Desk.

What this unlocks

• Faster experimentation: Growth teams can browse MCP servers, filter for “marketing” or “analytics”, and add context before launching the Community Signal Lab.
• Governance transparency: Product, legal, and security see exactly which connectors exist, whether they’re authenticated, and who to chase if something drifts.
• Agent autonomy: Because everything funnels through the MCP registry, agents can self-discover capabilities while still respecting approvals.

Feedback loop

We’re already acting on early feedback:

1. Bulk actions: Support for multi-select disable/enable is in testing.
2. Audit exports: CSV export of integration metadata lands next sprint so compliance teams can archive states.
3. Fine-grained access: Role-based visibility (marketing vs engineering) is on deck; expect a beta in May.

Keep requests coming inside /app/integrations -there’s a “Give feedback” shortcut that files straight into Product Brain.

Summary and next steps

The integration directory is now a first-class cockpit rather than a static list. Browse live integrations, filter by tag, check configuration health, and spin up missions with confidence.

Next steps:

1. Review your workspace and disable integrations you no longer need -fewer connectors mean lower risk.
2. Tag required integrations for upcoming missions so Product Brain agents can preflight checks automatically.
3. Share wishlist items via the in-app feedback panel; we prioritise roadmap based on real usage.

QA checklist

• ✅ Verified /api/integrations behaviour and fallback data against main branch on 9 March 2025.
• ✅ Confirmed alignment with Approvals Guardrails and Workflow Orchestrator documentation.
• ✅ Accessibility check complete for table and link text.
• ✅ Product team signed off roadmap narrative in Product Brain feedback thread #227.

Author: Max Beech, Head of Content
Updated: 10 March 2025
Reviewed with: Athenic Integrations & Platform team