--- title: "How to Enable YOLO Mode in localharness | AgentRQ Guidelines" url: https://agentrq.com/guidelines/yolo-mode/localharness description: "localharness has no YOLO switch because its ACP agent never asks: it gets no shell or file tools. What AgentRQ still adds, and what each prompt costs." --- ![](https://agentrq.com/assets/agent-icons/localharness.png) Checked October 30, 2026 | [All YOLO Mode guides](https://agentrq.com/guidelines/yolo-mode/) # How to Enable YOLO Mode in localharness There is nothing to enable. localharness's ACP agent never asks for permission, so it has no YOLO mode and needs none. It also gets no shell and no access to your files: it is a conversational agent with read-only on-chain tools. AgentRQ's YOLO toggle changes nothing for it, but AgentRQ still gives you a task, a record and a way to reply from anywhere. This page is one of our [YOLO mode guides](https://agentrq.com/guidelines/yolo-mode/), and part of our guide to [YOLO mode in every coding agent](https://agentrq.com/blog/what-is-yolo-mode-and-how-to-turn-it-on-in-every-coding-agent). Everything below was checked against the localharness source (compusophy/localharness, version 0.82.0, commit 7bcfe7d) on October 30, 2026. ![Animated sketch of YOLO mode in AgentRQ: claude waits to run go test, YOLO is flipped on for the task, and every later edit and test run lands auto-allowed in the Trajectory view](https://agentrq.com/assets/sketches/yolo-mode.svg) ## Does localharness Have a YOLO Mode? No, and the reason is different from most agents. localharness is a Rust agent SDK whose agents each own an on-chain identity and wallet at `.localharness.xyz`. Its ACP server, `localharness acp`, runs one of those agents in a headless session. The source is explicit about what that session can do: - → **No permission requests.** The ACP server sends none, so there is nothing for anyone to approve. - → **No filesystem and no terminal.** The working directory an ACP client passes is accepted and ignored. - → **No client MCP servers.** MCP servers your ACP client offers are not connected. - → **No session modes.** There is nothing to switch. What the agent does have is its published persona, lessons and skills, and read-only tools for the chain it runs on. ## What localharness Costs Per Prompt The cost to watch is not a dangerous command. Every ACP prompt is a metered turn paid in `$LH`, localharness's in-system credit, from the identity you run it as. An approval prompt would not have stopped that anyway, since it happens per prompt, not per tool call. Keep an eye on the balance of the identity you connect. ## Using localharness With AgentRQ ### Connect localharness Install the CLI with its wallet feature, create or import an identity, and point the [ACP Gateway](https://agentrq.com/docs/connect-acp-agent) at its ACP server ([localharness setup guide](https://agentrq.com/docs/agents/localharness)): bash ```bash cargo install localharness --features wallet npx @agentrq/acp-gateway@latest -- localharness acp --as ``` `--as` picks the identity whose persona and meter the session uses. `--model ` picks the model. ### What the YOLO Toggle Does Here The **YOLO** toggle on a task, and the workspace-wide **YOLO Mode (Execute All)** under **Settings → Automations**, auto-approve the permission requests an agent sends. localharness sends none, so [YOLO mode](https://agentrq.com/features/yolo-mode) has nothing to act on, on or off. A task runs the same either way. ### What AgentRQ Still Gives You - → **A task per job.** You hand localharness work as AgentRQ tasks, and follow them on the web, your phone or Slack. - → **Remote replies.** The agent's answers land in the task, and you can reply from wherever you are. - → **A record.** The [tool call history](https://agentrq.com/features/tool-call-history) lists the tool calls the agent made in each task, so you can see what it looked up. - → **Scheduling.** [Scheduled tasks](https://agentrq.com/features/task-scheduling) and [event triggers](https://agentrq.com/features/events) can run it unattended, and nothing will stop to wait for approval. ### How to Restrict localharness Instead Since there is no approval step, control comes from what you connect: - → **Fund the identity sparingly.** The `$LH` balance is the real limit on what a session can spend. - → **Use a dedicated identity.** Run ACP sessions under an identity made for the job, not the one you use for everything else. - → **Watch the task record.** It is the place to see what the agent did. ## FAQ **Does localharness have a YOLO mode?** No. Its ACP agent never asks for permission, so there is nothing to auto-approve. **Can localharness run shell commands or edit files over ACP?** No. Its ACP session gets no filesystem and no terminal, only conversation and read-only on-chain tools. **Does AgentRQ's YOLO toggle change anything for localharness?** No. YOLO approves permission requests, and localharness sends none. **Does localharness cost money per prompt?** Yes. Each ACP prompt is a metered turn paid in $LH from the identity you run it as. **How do I connect localharness to AgentRQ?** Run `npx @agentrq/acp-gateway@latest -- localharness acp --as ` from your project root, with your workspace's `.mcp.json` there. For every other agent, see [what YOLO mode is and how to turn it on in every coding agent](https://agentrq.com/blog/what-is-yolo-mode-and-how-to-turn-it-on-in-every-coding-agent).