E2B
Sprites
DedalusPair runtime and machine lane in one account. Spin up persistent workers with tools, then watch the fleet from one dashboard.
Provision Agentson any Substrate.
Pick runtime, provider, and model from one account. Provision persistent workers with loadout, state, console, logs, usage, cron, and artifacts.
Hermes
hermes-prod · Dedalus
CPU
0.3 vCPU
MEM
128 MB
DISK
2.1 GB
OpenClaw
openclaw-browser · Dedalus
CPU
0.5 vCPU
MEM
256 MB
DISK
1.4 GB
Claude Code
claude-code-ci · Dedalus
CPU
1.0 vCPU
MEM
512 MB
DISK
4.7 GB
Codex CLI
codex-sandbox · Dedalus
CPU
0.5 vCPU
MEM
256 MB
DISK
1.8 GB
ACTIVITY -- 6 MONTHS
149 active daystap a cell . click a chip to filter
filter by agent . 6
filter by service . 23
SDK
Create the worker in code.
Create a persistent worker with one typed recipe. Choose the agent, substrate, model, and state policy. The control plane handles boot, gateway, logs, and usage.
01import { AgentMachines } from "agent-machines"0203const am = new AgentMachines()0405const agent = await am.create({06agent: "hermes",07sandbox: "e2b",08model: "claude-opus-4.8",09persistent: true,10})1112await agent.run("review my code")Import the client
Use the SDK from any server.
Shape the worker
Pick runtime, substrate, and model.
Keep the worker
Persist files, skills, and memory.
Run real work
Send the prompt when ready.
WORKFLOW
One worker, one recipe.
Account settings compile into a runnable worker: runtime, substrate, model path, environment, loadout, and observable state.
AGENT MACHINES · CONTROL PLANE
stage 01Configure once, supervise the whole fleet.
Setup, lifecycle, terminal, logs, artifacts, usage, and chat all read from the same account objects. The dashboard shows what is configured, running, and saved.
- 01Pair
runtime + substratefrom one account settings model - 02Worker lifecycle:
wake · sleep · destroy(per substrate capability) - 03Persistent state in
/home/machine/.agent-machines
ClerkVercelAGENT MACHINES · RUNTIME LANES
stage 02Four runtimes, two operation models, one worker.
Hermes and OpenClaw run as longer-lived agent drivers. Claude Code and Codex run as task CLIs. Each lands inside the same machine record and persistent runtime root.
- 01Always-on agents:
Hermes · OpenClaw - 02Task CLIs:
Claude Code · Codex - 03Reusable per-account
agent profiles
Nous
OpenClawAGENT MACHINES · LOADOUT
stage 03Skills, MCP servers, CLIs, and plugins — one harness.
Loadout is the active stack on a machine: skills, MCPs, service lanes, CLIs, plugins, and source entries. Registry is where new entries get added.
- 01
161 skillsin SKILL.md protocol - 02
27 service lanes· MCP → CLI → skills - 03
24+ CLIs· closed-loop verification - 04Custom loadout:
skill · tool · mcp · cli · plugin
Agent MachinesAGENT MACHINES · SUBSTRATE LANES
stage 04Four substrate lanes — E2B, Sprites, Dedalus, and Vercel.
Each lane implements the same MachineProvider shape. The UI shows only the lifecycle actions and streaming behavior that provider supports.
- 01
E2B— sandbox with pause/resume - 02
Sprites— persistent microVM on Sprites.dev - 03
Dedalus Machines— strong default on boot and sleep/wake - 04
Vercel Sandbox— persistent microVMs with auto-snapshots
AGENT MACHINES · ENVIRONMENT
stage 05Gateway and env profiles follow every new worker.
Gateway modes, named variable sets, and bootstrap presets are account-level objects a new worker inherits on deploy.
- 01Gateway modes:
tunnel · ai gateway · byo - 02Named variable sets with
env profiles - 03Phase-tracked
bootstrappresets
CloudflareAGENT SURFACES
Every agent gets a full worker UI.
Each agent lands in the same worker frame: command surface, loadout, observable state, and saved runtime lanes.
╭─────────────────╮
│ ◈ H E R M E S │
│ ╱╲ ╱╲ │
│ ╱ ╲──╱ ╲ │
│ ╱ ╲╱ ╲ │
│ ╲ ╱╲ ╱ │
│ ╲──╱ ╲──╱ │
│ nous research │
╰─────────────────╯HermesSelf-improving agent with persistent memory, cron scheduling, session history, MCP host, subagents, and FTS5 search. Works with any OpenAI-compatible endpoint -- 30+ providers out of the box.
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bashhermeshermes gatewayhttps://hermes-agent.nousresearch.com/docs/https://github.com/NousResearch/hermes-agentLOADOUT
Loadout is the worker's complete kit.
Mirrors tool-hierarchy.mdc: built-ins fire immediately, MCP servers spawn at bootstrap, service entries pick the best lane, and tasks choose the right skill or tool.
callable tools
267
skills
161
services
27
task categories
12
built-in tools . 23
view >- Browser6
- Filesystem4
- Memory3
- Search3
- Shell1
- Vision1
- Image1
- Audio1
- Code1
- Delegate1
- Schedule1
services . 27
view >- Vercel
- Stripe
- Supabase
- Clerk
- Firebase
- Figma
- PostHog
- Sentry
- Datadog
- Linear
- Slack
- Shopify
tasks . 12
view >- Browser automationagent-browser
- Frontend verificationagent-browser diff
- Generative UIjson-render
- Code reviewcode-review
- Design reviewdesign-review
- QA + testingqa
hermes/openclaw/
claude code/
codexMirrors tool-hierarchy.mdc
129 trusted add-ons available
MCPs, CLIs, skills, sources, and providers composable into custom presets.
OBSERVABILITY -- DASHBOARD
The worker state you can inspect.
Runtime files, gateway health, usage, logs, loadout, and cron all map to dashboard APIs. These panels show the shape, not decorative placeholder art.
illustrative panels, matched to the dashboard APIs for gateway, logs, usage, loadout, and cron
WORKER CONTRACT
Runtime choice is saved on the worker.
The control plane stores agent, model, provider, sandbox, env, loadout, and persistence together. Switching agents changes the runtime, not the machine record.
memory . cron . sessions . MCP-native
hermeshermes gateway6 optionscomputer use . browser . shell . vision
openclawopenclaw gateway run6 optionsedit repos . run shell . SDK . headless
claudeclaude -p "task description"2 optionsship tasks . sandbox . JSONL . CI
codexcodex exec "task description"2 optionsFAQ
Common questions about Agent Machines.
13 answers
- 01
Can I run multiple agents for different jobs?
- Yes. Provision specialist machines from opinionated presets: Hermes for memory and scheduled work, OpenClaw for browser work, Claude Code or Codex for coding tasks. Each preset bundles runtime, model path, memory, and loadout. One dashboard supervises activity, chat, cron, logs, usage, and artifacts.
- 02
What is Agent Machines?
- Agent Machines is the product layer above sandboxes: a control plane that provisions a persistent agent worker as one unit — runtime, model path, skills, MCP, integrations, cron, observation, and fleet management — on the machine provider you choose. Pick Hermes, OpenClaw, Claude Code, or Codex, then pick E2B, Sprites.dev, Dedalus Machines, or Vercel Sandbox. Provision specialist workers from opinionated presets (Hermes, OpenClaw, Claude Code, Codex). Each preset is runtime + model path + memory bundle + loadout, visible from one fleet dashboard. The dashboard supervises the fleet. The long-term control surface is dashboard for humans, MCP/CLI for agent-to-agent orchestration.
- 03
How is this different from a regular chatbot?
- A regular chatbot mostly returns messages. Agent Machines gives the agent a machine record, runtime root, terminal, filesystem, logs, usage, cron schedules, sessions, artifacts, and installable tools. State lives with the worker instead of disappearing after one request.
- 04
Which agents can I run?
- Hermes, OpenClaw, Claude Code, and Codex are supported. Hermes is the default memory, cron, sessions, and MCP-native runtime. OpenClaw is the computer-use runtime. Claude Code and Codex are task-driven CLIs. All persist state under ~/.agent-machines/.
- 05
Which providers can host the machine?
- E2B Sandbox, Sprites.dev, Dedalus Machines, and Vercel Sandbox are live provider implementations. Each plugs into the same MachineProvider abstraction for provision, state, lifecycle, command streaming where available, and public URLs where supported.
- 06
How is this different from a sandbox like E2B or Daytona?
- Those are machine substrates. Agent Machines is the product layer above them: pick E2B, Sprites.dev, Vercel Sandbox, or Dedalus and get runtime install, loadout, gateway, cron, logs, usage, artifacts, and the browser console in one worker. Provider-specific features like sleep, snapshots, and public URLs are surfaced when the selected lane supports them.
- 07
How do I get my own machine today?
- Sign in with Clerk, add provider credentials in /dashboard/setup, pick the agent, provider, spec, and model, then provision the machine record. The browser flow creates the provider machine and stores it in your fleet; the reliable agent bootstrap path is still the matching root CLI deploy command until browser-driven bootstrap lands.
- 08
What tools and skills come pre-installed?
- The harness ships 161 SKILL.md files, 27 ranked service lanes (MCP → CLI → skills per vendor), 39 MCP catalog entries (2 core + 32 bundled + 4 IDE), 24+ closed-loop CLIs, and 9–23 agent-native tools depending on runtime (Hermes, OpenClaw, Claude Code, Codex). The loadout registry — not static marketing copy — is the source of truth.
- 09
Is Cursor required?
- No. Cursor is optional delegation for code edits through cursor-bridge and @cursor/sdk. Without CURSOR_API_KEY, the rest of the machine still runs: chat, files, browser automation, closed-loop tools, skills, cron, memory, dashboard polling, artifacts, and provider lifecycle controls.
- 10
What is ~/.agent-machines?
- ~/.agent-machines is the unified runtime root for Agent Machines. It holds all agent state -- skills, crons, sessions, logs, MEMORY.md, USER.md, config, chats, and artifacts. The repo checkout at /home/machine/agent-machines is used by reload-from-git.sh to sync knowledge from GitHub.
- 11
What inference providers are supported?
- Models can use any OpenAI-compatible /v1 endpoint. The CLI defaults to a vendor-agnostic inference URL; override with DEDALUS_CHAT_BASE_URL or configure model.base_url on the machine. The dashboard stores a model slug per machine.
- 12
What happens when a machine sleeps?
- On supported providers, sleep pauses compute while preserving the persistent volume. The next wake resumes from disk: app artifacts, agent runtime state, skills, cron schedules, sessions, and the venv remain available.
- 13
Where does my data live?
- Provider credentials and gateway bearers live in Clerk private metadata. Machine state lives on the provider machine under /home/machine, with all agent runtime data and app state under ~/.agent-machines. The public client only sees redacted provider and machine status.