Skip to content

Welcome to Vonzio

Vonzio is an open-source, self-hosted runtime for AI agents. Each conversation runs in its own isolated Docker container, on your machine, with the model provider you bring. You give agents work — bug fixes, inbox triage, research, code reviews — and they execute autonomously, calling tools, integrating with your data sources, and notifying you when they’re done.

This is the reference. If you’re looking for the marketing pitch, start at vonzio.com. Everything below assumes you have a running instance — install it with the one-line installer from the vonzio repo, then open the dashboard at http://localhost:3000.

ConceptOne-line definition
Agent profileA configuration: which model, which tools, which MCP servers, which secrets. Reusable.
WorkspaceA running session — an agent profile + an isolated Docker container + conversation history.
TaskA single unit of agent work. Either fire-and-forget (mode: batch) or interactive (mode: session).
PlaybookA scheduled task. Cron triggers, optional notification on completion, multi-chain execution.
MCP serverA capability surface the agent can call: memory, notifications, mail, calendar, platform operations.
IntegrationA connected third-party account (Slack workspace, Telegram bot, IMAP/SMTP mailbox, CalDAV calendar). Scoped per-agent.

This documentation is hand-written and matches the code at the most recent release. The OpenAPI spec (/v1/openapi.json) and Swagger UI (/v1/docs) on your instance are machine-generated from the same Zod schemas the server uses at runtime — use those for code generation, use this for understanding.

When the prose docs and the OpenAPI spec disagree, the spec wins — please open an issue so we can fix the prose.