Caelis intelligence, in the tools you already use.
Caelis exposes its read-only intelligence layer over the Model Context Protocol (MCP) — the same open standard protocol Claude Desktop and Claude Code use to connect to external data sources.
Early access — not self-serve yetAn internal tool, not a finished customer integration.
The MCP adapter runs locally over stdio, with no authentication of its own beyond what the underlying functions already enforce — built for Caelis' own developer team and FounderOS tooling today, not for customers to connect their own Claude to their own account. That's the next step, not something live yet.
Twelve read-only queries, grouped by domain.
No write access. Every query points at a real, existing function in Caelis — nothing has been rebuilt or simplified for this surface.
AssetService
- Asset360 — unified lifecycle, freshness, citations and engagement for one asset
- Latest A/B experiment results for one asset
OpportunityService
- Opportunities snapshot for one customer
- Growth audit (expansion opportunities)
- Highest-value opportunity for one customer
- Prospects ranked by Close Opportunity Score
- Full company diagnosis (health, audit, competitor standing merged)
- Cross-customer portfolio (operator)
VisibilityService
- Per-question competitor standing
- What changed in visibility since the last measurement
- Competitor comparison with diagnosis
OutcomeService
- Results snapshot for one customer
Pure protocol translation, no new logic.
The adapter does one thing: validates input against a schema, calls the real function in Caelis' own codebase, and wraps the response in a standard envelope. No business logic lives in the adapter itself.
Tell us what you'd want to build, and we'll talk through what's possible today.