The vaas-x SDK gives Python programs, machines and AI agents persistent, outcome-grounded memory. Send it what happened, ask what is similar, and record whether it worked, so the next answer favours what actually worked.
pip install vaas-xPython 3.10 or newer. Free plan: 1 device, 10,000 episodes and 500 queries a day, no card.
Five lines. This is the same code as the Quickstart, which explains each step.
from vaasx import Bootstrap
brain = Bootstrap(api_key="YOUR_API_KEY", device_id="my_device")
brain.connect("readings.jsonl") # or send episodes yourself: brain.ingest([...])
hits = brain.query("engine running hot", k=5)
brain.outcome(episode_id=hits[0]["id"], success=True) # tell it what happenedSeven connectors: HTTP, WebSocket, MQTT, files (.jsonl, .csv, .json), pandas DataFrames, Python generators and pipes. Bootstrap profiles each field and classifies the data itself, with no schema to design. All of that runs on your own hardware; only the resulting episodes cross the network.
Results are ranked by similarity and, by default, weighted toward episodes with a recorded successful outcome (prefer_success=True). It finds what worked before, not only what looks alike.
Record whether an action worked, either afterwards with outcome() or all at once as state, action and outcome when your agent already knows. Later queries surface what worked.
Every device keeps its own memory. On Developer plans and above, HiveMind fans one query out across all of them and merges the ranked results.
Rate-limit and quota problems come back as a VaasxAPIError with a status code and a machine-readable detail, and the REST API is documented for non-Python clients.
If you only want an AI agent to remember, you do not need the SDK at all. Add the connector address to any agent that supports MCP.
Everything the SDK stores lands in your account, and app.vaasx.com is where you look after it. Browse and search the notes, ask your memory a question, mark what worked, see when each device last wrote, set silence alerts, invite your team, and download or delete your data. The app never shows your API key.
Sign in with your email. The app has no password: we email you a code. Use the same address you used for your key, and your account is there waiting for you to claim it.
Start free. Prices and checkout are on the pricing page.
Step-by-step guides on public datasets, with your own key.