E2B
Sprites
DedalusRoute runtime and substrate in one account — deploy persistent agent workers with tools and supervise your fleet from one dashboard.
Persistent Agentson any substrate.
Route runtime and substrate in one account. Deploy persistent agent workers with tools and supervise your fleet from one dashboard.
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
145 active daystap a cell . click a chip to filter
filter by agent . 6
filter by service . 23
CONTROL PLANE
OpenRouter for agents and containers.
Route the agent runtime and the container substrate from one account, then deploy a persistent worker with its full harness — supervised from a single dashboard.
Route runtime and substrate
OpenRouter-style dual routing in one account. Pick an agent runtime, pick a substrate lane, deploy a persistent worker — Vercel on AWS for sandboxes.
Supervise a specialist fleet
One Clerk sign-in. Preset workers for code, design, and ops. Same account on every device — keys, runtime choice, and fleet metadata travel with you.
One worker, full harness
Not a bare sandbox — runtime, skills, MCP, cron, gateway, and observation in one deployable unit. State persists under /home/machine across sleep cycles.
WORKFLOW
Route runtime, substrate, and loadout in one surface
The control plane unifies dashboard, agent runtimes, tools, substrate lanes, and environment into one account — OpenRouter for agents and containers.
AGENT MACHINES · CONTROL PLANE
Configure once, supervise the whole fleet.
Settings, setup, worker lifecycle, terminal, logs, artifacts, and chat all read from the same account objects — not one-off wizard state.
- →→Route
runtime + substratefrom one account settings model - →→Worker lifecycle:
wake · sleep · destroy(per substrate capability) - →→Persistent state in
/home/machine/.agent-machines
Clerk
VercelAGENT MACHINES · RUNTIME ROUTES
Four runtimes, two operation models, one worker.
Autonomous agents wake on schedule with built-in drivers. Task-driven CLIs run per job but automate via headless flags and cron — same gateway and disk boundary.
- →→Autonomous routes:
Hermes · OpenClaw - →→Task-driven routes:
Claude Code · Codex - →→Reusable per-account
agent profiles
AGENT MACHINES · LOADOUT
Skills, MCP servers, CLIs, and plugins — one harness.
Built-ins and custom loadout entries live in the same account settings model so terminal edits sync back into the dashboard.
- →→
161 skillsin SKILL.md protocol - →→
27 service routes· MCP → CLI → skills - →→
24+ CLIs· closed-loop verification - →→Custom loadout:
skill · tool · mcp · cli · plugin
Agent MachinesAGENT MACHINES · SUBSTRATE ROUTES
Four substrate lanes — E2B, Sprites, Dedalus, and Vercel.
OpenRouter-style substrate routing — same bootstrap, gateway, and dashboard surfaces. Pick the host shape; lifecycle actions follow what each lane supports.
- →→
E2B— sandbox with pause/resume - →→
Sprites— persistent microVM on Sprites.dev - →→
Dedalus Machines— strong default on boot and sleep/wake - →→
Vercel Sandbox— persistent microVMs with auto-snapshots
VercelAGENT MACHINES · ENVIRONMENT
Gateway 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.
- →→Gateway modes:
tunnel · ai gateway · byo - →→Named variable sets with
env profiles - →→Phase-tracked
bootstrappresets
VercelCloudflare
AGENTS
Chat and terminal commands for every agent.
Autonomous agents have built-in drivers that wake up on schedule. Task-driven CLIs run per-task but can be automated via headless flags and cron.
╭─────────────────╮
│ ◈ H E R M E S │
│ ╱╲ ╱╲ │
│ ╱ ╲──╱ ╲ │
│ ╱ ╲╱ ╲ │
│ ╲ ╱╲ ╱ │
│ ╲──╱ ╲──╱ │
│ nous research │
╰─────────────────╯Self-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
Your agent's complete kit.
Mirrors the wiki's tool-hierarchy.mdc: every tool the agent picks from, ranked. Built-in tools fire in one turn. MCP servers get auto-spawned at bootstrap. Service entries cover 27 platforms (Vercel, Stripe, Supabase, Sentry, ...). Task entries rank which skill or tool to use for code review, design review, QA, research, and more.
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
Mirrors tool-hierarchy.mdc
129 trusted add-ons available
MCPs, CLIs, skills, sources, and providers composable into custom presets.
RUNTIME -- LIVE
What you watch on the dashboard.
6 panels . live every 5s
every panel above is rendered from the same /api endpoints the dashboard polls
SCROLL RUNTIME
Watch the agent machine assemble.
This section behaves like a locked product diagram: the copy stays stable while each workflow panel slides into place as you scroll. No fake dashboard screenshots, just the actual account → provider → agent → storage → registry model.
Account settings become a machine recipe.
Provider, agent, gateway, env, tools, and custom loadout are not onboarding-only values. They are durable account objects that every new machine can inherit.
control plane routeThe runtime router chooses the host shape.
E2B, Sprites.dev, Dedalus Machines, and Vercel Sandbox are interchangeable substrate routes — OpenRouter-style dual routing. Dedalus benchmarks best on boot and sleep/wake; the UI only shows lifecycle actions each lane supports.
four substrate lanesFour agents install into the same durable boundary.
Autonomous agents (Hermes, OpenClaw) have built-in drivers that wake on schedule. Task-driven CLIs (Claude Code, Codex) run per-task. All share the same gateway surface and persist state under /home/machine.
The dashboard reads the same system the agent writes.
Chats, artifacts, logs, terminal, sessions, and settings all converge on the same storage and provider execution model.
Browse and install from six live registries.
skills.sh, the MCP server registry, npm, Cursor plugins, GitHub repos, and URL manifests -- all searchable from one page. Click Add to write config and install on the machine.
FAQ
Common questions about Agent Machines.
12 answers
- 01
Can I run multiple agents for different jobs?
- Yes. Provision a fleet of specialist machines from opinionated presets — e.g. Hermes for research/cron, OpenClaw for browser work, Claude Code or Codex for coding. Each preset bundles runtime, skills, MCPs, and system prompts (the same stack vendors ship as single-purpose products like design or research modes). One dashboard supervises every machine: activity, chat, cron, logs, and cost.
- 02
What is Agent Machines?
- Agent Machines is the product layer above sandboxes: a control plane that provisions a full persistent agent in one unit — runtime, skills, MCP, integrations, cron, observation, and fleet management — on any substrate. Think OpenRouter for agents and containers: pick Hermes, OpenClaw, Claude Code, or Codex and pick E2B, Sprites.dev, Dedalus Machines, or Vercel Sandbox in one account. Spin up design, news, code, or ops agents from opinionated presets (Hermes, OpenClaw, Claude Code, Codex). Vendor SKUs (e.g. Anthropic design modes) are UI + skills + MCPs + system prompts — we ship that stack composable, one click per specialist, one pane of glass for the fleet. The dashboard supervises the fleet; MCP/CLI (roadmap) lets other agents orchestrate workers programmatically.
- 03
How is this different from a regular chatbot?
- A regular chatbot usually stores memory in browser state or a vendor-owned memory layer. Agent Machines persists operational state to a real machine filesystem: chat records, artifacts, USER.md, MEMORY.md, agent sessions, cron schedules, skills, and the runtime venv.
- 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 exposes provision, exec, public URL, and bootstrap through the same MachineProvider abstraction. Dedalus currently benchmarks best on boot latency (~250ms) and sleep/wake in our harness; E2B, Sprites, and Vercel Sandbox are fully supported alternatives.
- 06
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.
- 07
What tools and skills come pre-installed?
- The harness ships 161 SKILL.md files, 27 ranked service routes (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.
- 08
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.
- 09
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.
- 10
What inference providers are supported?
- Models route through 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.
- 11
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.
- 12
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.