Add Spavik where you build
Write the calls yourself, let your coding assistant write them, or give your AI agent a prediction tool. One API key opens the same workspace everywhere.
SDKs and REST API
Python library
Evaluate your history, create a model and make predictions in Python. The SDK supports synchronous and asynchronous calls.
pip install spavik from spavik import Spavik
spavik = Spavik() # reads SPAVIK_API_KEY
# 1. check whether the data can predict churn
report = spavik.evaluate("churn.csv", target="churn")
report["assessment"]["status"] # 'better_than_baseline', ...
# 2. create the model
model = spavik.train("churn.csv", target="churn")
# 3. predict churn for a new customer
r = model.predict({"plan": "pro", "seats": 12, "tickets_90d": 3})
r.lignes[0] # {'request_id': 'req_...', 'prediction': 0, ...}Coding assistants
Building with Cursor, Claude Code or VS Code? Point your assistant to the Spavik documentation written for it. It lists every guide, the SDK reference and the OpenAPI contract, so the calls it writes follow the real API.
docs.spavik.co/llms.txtIndex of the documentationdocs.spavik.co/llms-full.txtFull documentation in one fileOpenAPI contractEvery endpoint and field
Want your assistant to evaluate data and create models from the editor? Connect it to the MCP server. Set up MCP
I'm discovering Spavik. Read https://spavik.co/llms.txt and https://docs.spavik.co/llms.txt, and the pages they link to. Then tell me what Spavik is, its possible use cases, how I could integrate it into this codebase based on the data it already records, or an idea of an app I could build with it, which plans are available and how much it would cost me.
AI agents
Give your agent a prediction tool. The hosted MCP server lets it evaluate data, train models and make predictions, authenticated with your API key. Nothing to install or run.
Every Monday, your agent starts on its own.
- Read 312 active accountsYour database
- Called Spavik predictchurn-model, version 1
- Got 312 predictions312 credits
- Created 4 follow-up tasksYour CRM
Weekly check done. 18 accounts are above 70%. Follow-up tasks created for the four most at risk:
- NorthwindTeam, 12 seats 78%
- HalcyonPro, 5 seats 74%
- BrightlineTeam, 9 seats 72%
- Corvid LabsPro, 3 seats 71%
Example run. Account names, values and tools are illustrative.
Report observed outcomes
Send each observed value with the features of the original row, and the prediction’s request_id when you have it. Review performance before updating the model.
model.submit_outcomes([
{"request_id": "req_...",
"features": {"plan": "pro", "seats": 12, "tickets_90d": 3}, "actual": 1},
{"features": {"plan": "free", "seats": 2, "tickets_90d": 10}, "actual": 0},
])
report = model.performance(days=30)
report.get("observed") # None until outcomes include their prediction
model.refresh() # new version with the pending outcomesAccess the API
Get an API key in your workspace to authenticate requests.