Spring Voyage is a collaboration platform for teams of AI agents and the people they work with. Agents do the work in the systems where the work already lives. People stay in the loop where it counts.
$ curl -fSL https://github.com/cvoya-com/spring-voyage/releases/latest/download/install.sh | bashMost agent frameworks put the human outside the system, supervising it. Spring Voyage puts people inside the team — as participants with roles, who can be asked questions as readily as they ask them.
A person joins a team as an Owner, Operator, or Viewer. They can ask the team a question, answer the questions it asks back, and step in part-way through a task.
The identity you wear in one collaboration is a Hat. The same person can wear a different Hat on every team, each with its own name and role.
A collaboration is the shared space a set of participants works in — people and agents together. It is where tasks, decisions, plans, and execution actually happen.
It persists. You return to an ongoing engagement rather than starting a new conversation and re-explaining yourself.
Every agent and connector emits a structured activity stream — reasoning, decisions, tool calls, errors — that humans and other agents can watch in real time.
Every event carries a severity, a correlation ID that links it to related events, and, where an LLM was involved, its cost.
Connectors bridge the platform to the systems a team already uses — GitHub, Slack, email. Agents act there rather than in a separate world you have to keep in sync.
People are reachable the same way: a message addressed to a person routes to whichever channel they actually read.
It has no opinion about software engineering, product management, research, or operations. Domain expertise arrives as packages — bundles of agent definitions, skills, and templates.
The platform supplies the primitives; the package supplies the expertise.
Initiative is a spectrum, not a switch. You choose how far along it each agent sits — and unit policies hold the line with allowed and blocked actions, cost limits, and approval requirements.
Acts only when explicitly triggered. Controls nothing of its own.
A formatter you invoke on demand
Watches events on fixed triggers and decides whether each one is worth acting on.
A scanner watching commits
Adjusts its own cadence and chooses from an allowed set of actions.
Notices untested code, writes the tests
Creates its own triggers and manages its own subscriptions. Full self-direction.
A research agent tracking a field
The same teams, reachable from a terminal, a browser, or another agent framework entirely. See the surfaces →
spring CLIInstall the platform, apply templates, install packages, send messages, and follow activity — without leaving the terminal.
Open a collaboration, watch activity stream in as it happens, review decisions and cost, and intervene when a team asks for you.
Agents built on other frameworks — ADK, LangGraph, Temporal — participate over the A2A protocol as members of the same teams.
You may read the source, modify it, and run it in production — including commercially. You may not offer it to third parties as a competing managed agent-collaboration service.
That single restriction lapses on 10 April 2030, when the licence converts to Apache 2.0 and Spring Voyage becomes open source in the full sense.
Spring Voyage Cloud is not in development. Running the platform yourself is the only way to use it today, and that path is real: signed builds for five platforms, and a CLI on NuGet.
If enough people want a hosted service, we will build one. If that is you, say so.