Skip to content

The agent-native knowledge API

The knowledge base your app and your AI agents share

Define a schema — FoxNose instantly generates the embeddings, the hybrid search index, and a Knowledge MCP server, all on one endpoint. Plug in Claude, ChatGPT, Cursor, or any agent framework — no Pinecone+Postgres glue, no hand-written tool catalog.

For support agents, product copilots, and internal knowledge assistants.

Works with any LLM, agent framework, or application

OpenAIAnthropicGoogleLangChainOllama
1store.pyraw data in
# Store — embeddings generated automatically
mgmt.create_resource("help_articles", {"data": {
"title": "Refund policy for EU customers",
"body": "Items can be returned within 30 days…"
}})
auto-embed · index · ready
2search.pyhybrid query out
# Hybrid search — vector + keyword, weighted fusion
results = flux.hybrid_search(
"help_articles",
query="can I return after 14 days?",
find_text={"query": "refund policy"},
vector_weight=0.7, text_weight=0.3,
)
# → top match: "Refund policy for EU customers"
The Problem

Building an AI agent over real data is a stack problem, not a model problem

Most teams shipping a vertical agent end up maintaining a vector DB, a primary store, a full-text index, an embedding pipeline, sync scripts, and a hand-written MCP server with tool definitions that drift every time the schema changes. FoxNose collapses all of it.

Typical AI data stack
Your Backend
API endpoints, validation, auth, rate limits, scaling...
Vector DB
Pinecone, Qdrant...
Embedding API
OpenAI, Cohere...
Search Engine
Elasticsearch...
Database
Postgres, Mongo...
+ Sync scripts
ETL pipelines, cron jobs, retry logic, error handling...
+ Hand-written MCP server
Tool definitions, auth, schema drift every time the data model changes...

Five services, five bills, sync scripts to keep them consistent, and an MCP tool catalog that drifts.

With FoxNose
One Managed Service
REST + MCP from the same endpoint
Instant API
from your schema
Auto-embeddings
on every save
Knowledge MCP
auto-generated, no drift
Multi-tenant
by data model

Define a schema, get REST endpoints, a hybrid search index, and a Knowledge MCP — same auth, same data, no parallel surface to maintain.

Developer Experience

One API from schema to search

Define your data model, store records, and query with hybrid search. Auto-embeddings and indexing happen behind the scenes.

# mcp config — Claude Code, Cursor & other HTTP MCP clients
# (stdio-only clients such as Claude Desktop connect via mcp-proxy)
{
  "mcpServers": {
    "foxnose-support": {
      "type": "http",
      "url": "https://<env_key>.fxns.io/support/_mcp",
      "headers": {
        "Authorization": "Simple <public_key>:<private_key>"
      }
    }
  }
}
# → 5 tools instantly available:
#   discover_resources · describe_resource · query_records · get_record · search_records
# Same auth as REST. Same data. No tool catalog to hand-write.

Define a schema in the dashboard. A Knowledge MCP server appears at /<prefix>/_mcp over Streamable HTTP — auto-generated, no server to deploy. Works with Claude Code, Claude Desktop, OpenAI Responses API, Cursor, Cline, Continue, and any MCP-aware client. How it works →

From signup to your first MCP query

Sign up. Define a schema. Connect a Flux API. Drop the MCP server URL into Claude or Cursor. That's the whole setup.