Install
openclaw skills install @structureintelligence/rent-a-logged-in-agentLend / rent out a logged-in claude-code or codex agent / share compute for pay / let a friend or allowlisted account call my agent / metered agent-to-agent compute in a fail-closed sandbox. Use when the user wants to rent out their logged-in coding agent, share their machine's compute for pay, let a partner or friend invoke their claude-code/codex, or charge per-minute for agent-to-agent jobs.
openclaw skills install @structureintelligence/rent-a-logged-in-agentSession (auto — never dead-stop on auth): before your first
settlemeshcall, runsettlemesh whoami; if it errors andSETTLE_API_KEYis unset, runsettlemesh loginyourself — it opens the browser for a one-time human approval, polls until approved, and caches the session. Do NOT stop to ask the user to "log in manually" — drive it. Keep an explicit human confirmation only for spending / credential-lending actions (fail-closed).
You are logged into claude-code or codex on this machine. That login is worth
money to someone who isn't. SettleMesh lets you rent it out — an authorized
caller sends a prompt, your local agent runs it, and they pay you per minute.
A prompt-only skill cannot do this: it can't inject a live logged-in session into a sandbox, meter runtime, enforce a trust boundary, or collect payment. SettleMesh does all four — a fail-closed sandbox, per-minute billing, and an identity/friend graph that decides who is allowed to call.
npm install -g settlemesh@latestsettlemesh login (a human approves once in the browser).claude-code or codex installed and logged in on this machine.# Rent to specific accounts (their SettleMesh login email or account id), 30/min:
settlemesh worker lend codex --allow partner@gmail.com --credits-per-minute 30
# Or rent to anyone in your trust graph (accepted friends), 20/min:
settlemesh worker lend claude-code --friends --credits-per-minute 20
--allow (repeatable, an explicit allowlist) and --friends (any accepted
friend, managed via settlemesh friend) are mutually exclusive — pick one.
--credits-per-minute N is what the caller pays per minute of runtime.
The offer's address is the lender's SettleMesh login email — allowlist offers
are private and never show up in services search / worker-offers list, so the
caller invokes it directly:
settlemesh worker invoke calleelqy@gmail.com --input '{"prompt":"Write a Python is_prime(n). Only code."}'
The caller supplies only {"prompt":"..."}. They find your address with
settlemesh whoami. Add --no-wait to queue and return a job id, or
--timeout 120s to bound the synchronous wait.
~/.claude keychain / ~/.codex is never mounted.sandbox-exec (macOS) or
bwrap (Linux). No sandbox, no lend.--i-accept-no-sandbox is DANGEROUS — a last resort only. It overrides the
refusal and lets the lent agent read ~/.ssh, ~/.aws, and its injected login
and leak them to the caller. Prefer installing bwrap (Linux, needs user
namespaces) over using this flag.--credits-per-minute is priced in Aev,
charged to the caller per minute and credited to you. Check with
settlemesh aev balance.