> ## Documentation Index > Fetch the complete documentation index at: https://docs.mutagent.io/llms.txt > Use this file to discover all available pages before exploring further. # Three ways to run Helix > One binary, three working modes — the orchestrator, a single agent, or the RLM-based Prime loop. What each gives you and what to type. You install one binary, `helix`. How you start it decides what kind of session you get. All three modes read your project's `AGENTS.md` / `CLAUDE.md`, use the provider you signed in with, and share the session tools — `/trace`, the sub-agent views, and the rest (see [Sessions](/helix/sessions)). | You type | You get | Reach for it when | | ---------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- | | `helix` | **The orchestrator.** The Helix dashboard, the lifecycle stages, and a crew of specialist agents it dispatches for you. | You want the spec → build → evaluate → diagnose → optimize loop, or you'd rather describe the job and be routed. | | `helix agent` | **A plain coding agent.** A capable general-purpose agent with no orchestrator and no lifecycle framing. | You want to work on code directly, with Helix's harness but none of its ceremony. | | `helix agent --name ` | **One agent definition, as the whole session.** | You wrote an agent (or Helix built one) and want to run or test it on its own. | | `helix --prime` | **Prime — the RLM-based agent loop.** Instead of picking tools one call at a time, the model writes and runs code against your repository, recursively, in an isolated runtime. | Data work, experiments, anything where you want the model to decompose work into code rather than into briefs. | Not sure? Start with `helix`. It is the full product; the other modes are subtractions from it. ## `helix` — the orchestrator ```bash theme={null} cd your-project helix ``` You land on the Helix dashboard: the lifecycle stages, the skills that loaded, and the command roster. The Helix boot dashboard: the gradient wordmark, the five-stage lifecycle box, the six mounted skills, setup and state panels, and the full command roster. From here you drive the loop in plain English or with a `/command` — both reach the same stage: ``` evaluate the support-triage agent against last week's traces ``` ``` /evaluate support-triage ``` When a stage needs parallel work, Helix dispatches sub-agents. They appear as a list under your prompt — open one and its conversation replaces the chat; typing steers that agent. That surface is described in [Sessions](/helix/sessions). `/help` repaints the dashboard at any time; `/clear` starts a fresh conversation without leaving. This is the only mode with the lifecycle commands (`/spec`, `/build`, `/evaluate` …). The other modes do not load the orchestrator, so those commands do not exist there — by design. ## `helix agent` — a single agent as the session ```bash theme={null} helix agent # a general-purpose coding agent, interactive helix agent --name reviewer # one agent definition IS the session helix agent -p "list the TODOs" # headless: one task, answer on stdout ``` No dashboard, no Helix persona, no lifecycle commands: a lean coding agent on the same harness. Sub-agents, Monitor, `/goal`, `/trace` and your project context all still work. Definitions resolve by `--name` (from `.mutagent/agents/` and `.pi/agents/`), `--file`, `--prompt`, or a bare quoted prompt. Interactive, loading a specific definition, headless with `-p` — plus the capability list and the sub-agent views, with real captures. ## `helix --prime` — the RLM-based agent loop ```bash theme={null} helix --prime ``` Prime changes the working mode itself. Instead of the model selecting tools one call at a time, **it writes and runs JavaScript** against your repository, in an isolated runtime. The structure is recursive: the model decomposes a problem by writing smaller code, not smaller briefs, and when it needs help it spawns child agents from code — each of them a Prime in turn, bounded in depth. What is gone relative to `helix`: the orchestrator persona, the dashboard, the lifecycle skills, and the `Agent` tool. What is there instead: one tool, `run`, that executes code in a **namespace that persists for the whole session** — variables, functions and classes defined in one call are still defined in the next, so the model accumulates state instead of recomputing it. Every file write and shell command still passes a host gate, so "the agent ran code" still means something. Two commands belong to this mode: | Command | Does | | --------- | -------------------------------------------------------------------------------------------------------------------- | | `/prime` | Show the namespace — the identifiers defined so far, how many cells ran, and whether any runaway cell was recovered. | | `/refine` | Record a learning from this session for the next Prime session to read. It won't accept a claim without evidence. | Combine it with agent mode to run a specific definition on the Prime loop — same agent, a code-interpreter as its only tool: ```bash theme={null} helix agent --name reviewer --prime ``` Prime is experimental. ## The three modes side by side | | `helix` | `helix agent` | `helix --prime` | | ------------------------------------------ | ------- | -------------- | --------------- | | Helix dashboard + persona | yes | no | no | | Lifecycle commands (`/spec` … `/optimize`) | yes | no | no | | Sub-agent crew (fleet view, agent viewer) | yes | available | from code only | | Persistent `run` namespace | no | with `--prime` | yes | | `AGENTS.md` / `CLAUDE.md` loaded | yes | yes | yes | | `/trace` | yes | yes | yes | What the dashboard shows, the slash commands, the fleet view and the agent viewer.