Runtime
Run a hosted RAG agent
Use a managed answer runtime that resolves policy, searches Buckets, and returns grounded responses.
Create hosted RAG agents that search selected Buckets, follow scoped answer policy, and return source-backed responses with citations through an OpenAI-compatible interface.
Agent live
calypso-rag-agent:support
Bucket memory
Answer policy
Search selected Buckets, answer concisely, and expose source evidence when citations are available.
Sample output
“Install the widget script, pass your team id, then verify the cited source references.”
API
callable
MCP
ready
n8n
usable
Runtime
Use a managed answer runtime that resolves policy, searches Buckets, and returns grounded responses.
Memory
Attach the exact source sets each agent is allowed to use, so retrieval stays scoped and repeatable.
Policy
Define retrieval effort, citation visibility, tone, and answer rules once, then reuse them.
Evidence
Expose source-backed answers users can verify instead of asking them to trust a black-box response.
Agent profiles
Give support, pricing, onboarding, and internal teams named agents with their own Bucket memory, answer policy, and evidence behavior.
calypso-rag-agent:supportAnswer troubleshooting questions from approved support sources with citations.
calypso-rag-agent:pricingExplain plans, limits, packaging, and procurement rules from approved sources.
calypso-rag-agent:onboardingGuide setup and implementation from curated onboarding knowledge.
API workflow
Use an OpenAI-compatible request, choose a Calypso agent model id, and let the agent handle retrieval, policy, and source-backed output.
Responses API
Ask a named agent
# Ask a named Calypso RAG agent
curl https://api.calypso.so/v1/responses \
-H "Authorization: Bearer sk_..." \
-H "Content-Type: application/json" \
-d '{
"model": "calypso-rag-agent:support",
"input": "How do I connect the widget?",
"store": false
}'Select the source memory this agent is allowed to search.
Control retrieval effort, citation behavior, and answer rules.
Use the same model id from API, MCP, n8n, widget, or product UI.
Send grounded answers with sources, citations, and metadata.
Deploy everywhere
Configure the agent once, then reuse the same Bucket memory and citation behavior wherever users or automations ask questions.
Get started with Agents
Attach Buckets, set answer policy, and launch source-backed agents across your website, API, MCP clients, n8n workflows, and product experiences.