OpenClaw, running all day.
An agent that lives in your Slack, Discord and codebase, wakes on a heartbeat, runs cron jobs and answers whenever you message it. It rebuilds its whole prompt on every run, which is why it goes through tokens so fast on a per-token bill.
Pre-launch. Accounts are open; the API and checkout open soon.
How it runs.
01
Point OpenClaw at Voidstone.
Add Voidstone as a custom provider in openclaw.json, with Kimi K3 as the main model and a quick one as the fallback.
02
Let it live in your chats.
Message it from Slack, Discord or the terminal; it works in your codebase and your tools.
03
Give it a schedule.
Heartbeats and cron jobs run without you: morning briefings, inbox triage, checks on your servers.
Why lanes fit it.
Every run resends the context.
Its prompt, tools, skills and workspace files go out with every run, plus the conversation so far. On lanes none of it is billed per token.
No need to starve it.
On a per-token bill people cap its context and prune sessions to save money. On lanes it can use your plan’s full context.
Heartbeats cost nothing extra.
A check every hour holds a lane while it runs and never adds to the bill.
Works with what you use.
OpenClaw
{ env: { vars: { VOIDSTONE_API_KEY: "vs_your_key" } }, agents: { defaults: { model: { primary: "voidstone/kimi-k3", fallbacks: ["voidstone/deepseek-v4.1-flash"], }, }, }, models: { mode: "merge", providers: { voidstone: { baseUrl: "https://api.voidstone.net/v1", apiKey: "${VOIDSTONE_API_KEY}", api: "openai-completions", models: [ { id: "kimi-k3", name: "Kimi K3" }, { id: "deepseek-v4.1-flash", name: "DeepSeek V4.1 Flash" }, ], }, }, },}