Calypso Context
Multimodal RAG your users can check.
Load PDFs, images, web pages, and data into buckets, and get cited answers through an OpenAI-compatible API, MCP, n8n, or a web widget. Hosted, so there is no vector database to run.
Calypso Context is the knowledge layer behind Calypso WhatsApp. Developers use it directly.
Live demo
Pick a corpus, ask, and read the citations.
Quick start
Point the OpenAI SDK at Calypso. That is the integration.
Same client, same Responses call. The model is your agent; the answer comes back with the sources it drew on.
- 01Create a project API key in Project → API Keys.
- 02Put your sources in a bucket and wait for ready.
- 03Call calypso-agent, or a named agent, from any OpenAI SDK.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["CALYPSO_API_KEY"],
base_url="https://api.calypso.so/v1",
)
response = client.responses.create(
model="calypso-agent",
input="What does the refund policy say about opened items?",
)
print(response.output_text)
# Sources arrive as file_search_call.results[] with source_index,
# label, page_number, and access_url.What it reads
PDFs, images, pages, and data. Read, not just stored.
A source is accepted if Calypso can actually read it, and rejected at upload if it cannot. Every source lands in a bucket and reports one ready bit when it is searchable.
PDFs
Read as text and as page images, so tables, charts, and scanned pages count.
Citations point at the page.
Images
PNG and JPEG are read for what they depict. A product photo answers a question about the product.
Citations point at the image.
Documents
DOCX, Markdown, and plain text.
Citations point at the document.
Web pages
Point at a URL. The page is crawled, analyzed, and kept as a living source, one per team.
Citations point at the page URL.
Data files
CSV and JSON. Export Excel to CSV before uploading.
Citations point at the file.
Code
Python, TypeScript, Go, Rust, SQL, HTML, YAML, and more.
Citations point at the file.
Up to 25 MB per file, 1 to 100 files per batch. The full list of accepted types, and how each gets in.
The shape
One source layer. Named agents. Answers everywhere.
01 · Buckets
Scoped, durable source memory.
A bucket is a named collection of sources. Ingestion writes into buckets, agents search buckets, and nothing crosses between them unless you bind it. Create one on first use, or provision it by slug.
The doors
Four ways in. One answer.
API
OpenAI-compatible API
Responses, Chat Completions, and Search at api.calypso.so/v1, with the standard SDKs.
MCP
MCP server
Ask, search, upload, and manage buckets from Cursor, Claude Desktop, and other MCP clients.
n8n
n8n community node
Ask a grounded agent and upload knowledge from workflows.
Website
Web widget
One script tag, bound to the agent policy you already tested.
Proof
Ask a real corpus yourself.
Ask the SpaceX IPO filing like an analyst.
Grounded across indexed filing sources, including prospectus summaries, risk factors, MD&A, launch vehicle pages, Starlink materials, xAI/X references, charts, and image exhibits.
- Official filing corpus
- 84 sources
- 142.4 MB
- Multimodal
The World Cup is live. So is 150 years of soccer history — ready to be questioned.
Grounded in international men's soccer results from 1872 to 2026 — match outcomes, tournaments, venues, shootouts, goalscorers, and historical team names. Fast mode is the default; switch to Medium or Extended anytime for deeper, citation-rich analysis.
- 49k+ results
- 1872-2026
- Shootouts
- Goalscorers
Controls
Scoped, traceable, yours.
Scoped per bucket
An agent answers only from the buckets it is bound to. Support never quotes the price list; Sales never reads the return policy.
A source on every answer
Inline citations plus a structured source list with page, modality, and a signed link to open the original.
Yours, isolated
Your knowledge stays in your workspace, never trains shared models, and can be exported or deleted any time.
Plans
Included in every plan.
There is no separate Context SKU. Pick the capacity: knowledge storage, answers per month, and agents.
Context pricingStarter
$19/ month
100 MB of knowledge · 1,000 answers a month
Pro
$49/ month
1 GB of knowledge · 5,000 answers a month
Scale
$199/ month
10 GB of knowledge · 25,000 answers a month
Questions and answers
Before you load your first bucket.
New to the idea? Read what multimodal RAG is.
Start today
Load a bucket. Call the agent. Ship.
Start the 14-day trial on Pro limits, create a project API key, and get a cited answer in your first hour.