Create agents that grow around a purpose
Each agent remembers across sessions, runs on schedules, reacts to events, and extends itself over time.
$npm i -g wolli
Each agent remembers across sessions, runs on schedules, reacts to events, and extends itself over time.
A wolli agent is born empty. Tell it what it's for and it writes who it is into its SOUL.md.
Scout needs issues delivered the moment they open, so it writes its own GitHub integration, the workflow that wakes it, and the tool to talk back.
An issue opens and the workflow wakes scout in a fresh session. It triages, flags what matters, and reports back.
Routines the agent repeats get written down as skills, authored by the agent itself.
Memory, autonomy, sandboxing, and scheduling are built into every agent. Focus on its purpose.
The agent works out its purpose with you in its first conversation and writes it as the first line of its SOUL.md.
The agent authors and installs its own skills, tools, and integrations. It grows more capable at its job instead of staying a fixed tool.
Sessions are an append-only JSONL tree, the agent's lifetime memory. Nothing is rewritten; the latest leaf resumes by default.
SOUL.md, MEMORY.md, and USER.md are frozen into the system prompt each session, maintained by the agent itself.
Runs sandboxed by default: Apple Seatbelt, bubblewrap, or Docker. Your real machine is an approval-gated escalation away.
Multi-provider via OAuth login: Anthropic, OpenAI, and others.