The engine, as an API.
One call pattern for every tool. Call Novo's in-silico tools over HTTP from a script or your own AI agent. One base URL, one Bearer key, usage-based credits. The same engine behind the MCP connectors, exposed directly to your code, plus a few tools available only here.
One pattern, every tool.
Every tool is POST /v1/tools/{name} with a JSON body and a Bearer key. Plain HTTP, no SDK.
curl -s https://api.novomcp.com/v1/tools/validate_target \
-H "Authorization: Bearer $NOVOMCP_API_KEY" \
-H "Content-Type: application/json" \
-d '{"arguments": {"target": "EGFR", "disease": "lung cancer"}}'Responses come back as { result, usage }. The tool's payload plus the credits spent and your remaining balance. Full curl, Python, and JavaScript examples are in the API docs.
Bring your own AI agent.
NovoMCP is the engine. You supply the model. Expose Novo's tools to your own agent (Claude API, GPT, anything) and let it call them in its loop, with live tool schemas from GET /v1/tools.
# Expose Novo's tools to your own model and let it call them in its loop.
def call_novomcp(tool, arguments):
r = requests.post(
f"https://api.novomcp.com/v1/tools/{tool}",
headers={"Authorization": f"Bearer {KEY}"},
json={"arguments": arguments},
)
return r.json() # { "result": {...}, "usage": {...} }Each hostname, one job.
api.novomcp.com is the hosted REST API you call from your code. localhost:8018/mcp is the local MCP surface for any MCP-compatible assistant. localhost:8018 is the same open engine over plain HTTP. One engine underneath all three.
api.novomcp.comHosted REST API, call tools over HTTP (FAVES-consumption surface)hostedlocalhost:8018/mcpLocal MCP surface, install the open engine and add it to any MCP-compatible AI assistantself-hostlocalhost:8018Local REST surface, the same OSS engine over HTTP, keys optionalself-hostNovo is the engine. Reach it the way you work.
Chat, your own code, the desktop, the browser, your editor, the terminal. One engine and one audit trail, whichever door you use.
Wire 69 in-silico tools into your stack.
Self-host the engine on your own hardware, or request a cloud deployment with GPU compute and the FAVES compliance API.