Calypso MCP
Calypso MCP brings multimodal RAG to your AI tools and workflows with a grounded answer layer built for websites, AI agents, support copilots, internal knowledge assistants, and product experiences.
Instead of relying on generic model memory, Calypso retrieves from your actual source material, including PDFs, docs, screenshots, diagrams, charts, help content, and structured knowledge, so answers stay source-backed, explainable, and production-ready.
Why teams use Calypso MCP
-
Grounded answers, not generic guesses
Query your knowledge with retrieval-backed responses instead of unverified model output. -
Multimodal RAG support
Work across PDFs, documentation, help center content, screenshots, diagrams, images, and internal files. -
Built for AI agents and real product surfaces
Use the same answer layer across website chat, support, sales assistants, workflow automation, and custom agent experiences. -
Agent-store and knowledge-store uploads
Upload files directly to a specific RAG agent store for retrieval-backed chat, or ingest them into a broader durable knowledge store for indexing. -
RAG-first file semantics
Agent file attachments use retrieval-backed behavior withrag_policy, so uploaded files are treated as searchable context rather than default inline attachments.
Included MCP tools
calypso-rag-agent
Query the Calypso RAG agent for grounded answers over your configured knowledge base.
calypso-upload-agent-file
Upload a file into the agent store and receive a compatible file_id for retrieval-backed RAG chat.
calypso-upload-knowledge-file
Upload a file into the knowledge store for durable indexing and shared corpus ingestion.
Best fit for
- AI support assistants
- Sales enablement copilots
- Internal knowledge agents
- Website answer widgets
- Workflow automation
- Product-embedded RAG experiences
What makes it different
Calypso is designed as a grounded answer layer. That means:
- retrieval over real source material
- reusable deployment across multiple surfaces
- multimodal knowledge handling
- strong fit for operational and customer-facing AI systems
If you need source-backed answers from PDFs, docs, visuals, and knowledge files inside an MCP-compatible workflow, Calypso MCP gives you a clean path from upload to retrieval to grounded response.
服务器配置
{
"mcpServers": {
"Calypso Multimodal RAG": {
"command": "npx",
"args": [
"-y",
"@calypso-rag/calypso-mcp"
],
"env": {
"CALYPSO_API_KEY": "sk-your-calypso-api-key",
"CALYPSO_API_BASE_URL": "https://api.calypso.so/v1"
}
}
}
}