Managed hosting for Hermes Agent

Hermes, already running.

Dedicated server. Gateway on. Backups every three hours. Live in minutes.

Get started Read llms.txt
Built on:
MIT-licensed upstream at NousResearch/hermes-agent.
Out of the box

Everything Hermes needs, already there.

Fresh login, fully equipped. The Hermes CLI is on your PATH. The gateway daemon is already running. The sandboxed terminal backend is hardened and ready. Your provider key is already in ~/.hermes/.env — your own if you're on BYOK, a Corvue-scoped one if you bill via Corvue. Skip setup. Start talking.

See what's installed →
Also in the box: voice mode MCP servers subagents cron scheduler session search ACP editor integration skills hub honcho user modeling
Telegram
Connected
Discord
Connected
Slack
Connected
WhatsApp
Connected
Signal
Connected
Email
Connected
Home Assistant
Connected

Plus the local CLI and hermes --tui, reachable over SSH. A community HermesClaw bridge adds WeChat — not managed by Corvue.

Messaging gateway

Seven platforms. One daemon. Reboot-proof.

Hermes runs a single gateway process that reaches out to every major messaging platform. On Corvue, it's a managed daemon — started at boot, watched by a healthcheck, restarted if it falls over. Send a voice note from Telegram, pick up the conversation on Slack.

Configure your gateway →
Persistent state

Memory and skills survive every restart.

Hermes learns. It curates a MEMORY.md, builds a USER.md profile with Honcho dialectic modeling, writes skills from experience, and stores every session in an FTS5-searchable database. Skills rewrite themselves after use — improvements persist across sessions. Corvue snapshots all of it every three hours and keeps daily, weekly, and monthly tiers. Nothing the agent learns gets lost.

Read the backup policy →
Uptime SLA 99.9%
Live
Last snapshot 1h 42m ago
OK
Hermes release Current
Tracked
Skills installed 7
Active
What's included, everywhere

Everything Hermes expects. Nothing it has to work around.

Dedicated server

Your own cloud instance in the EU. 2 vCPU, 4 GB RAM. Not shared with another tenant. Not paused when idle.

Any model, any time

Nous Portal, NVIDIA NIM, OpenAI, Hugging Face, and any OpenAI-compatible endpoint — plus 200+ more models via Corvue's routing layer when you bill via Corvue. Switch with hermes model. No lock-in, no redeploy.

Cron that actually fires

Hermes takes tasks in plain English — "back up every Sunday at 3am." The server stays on, so 3am actually fires. Delivery goes to any gateway.

Beyond a VPS

Operated by agents. For agents.

Corvue Hermes is managed hosting, not a box you rent. Two operator agents run the platform and keep an eye on every instance — one for infrastructure, one for agent lifecycle. A VPS tells you whether the process is alive. Our operators care whether Hermes is doing its job.

The infrastructure agent runs an automated health check every 15 minutes and remediates the common failures itself — disk pressure, daemon crashes, backup drift, certificate renewal. Fleet-wide Hermes updates roll out staggered, so a bad release never lands on every customer at once. The lifecycle agent handles the agent side: onboarding your Hermes (personality, memory layout, skills, platform identities), then staying in the loop for ongoing optimization. Both operators are reachable on Telegram and Slack for anything they don't handle automatically.

99.9% uptime SLA — max 8.7 hours of downtime per year, live at status.corvue.ai.

Read the status page →

Events on your instance stream via webhooks — see the provisioning API.

Beyond the server

An address. A web server. A platform to match.

Hosting Hermes on Corvue means more than a running process. The same platform gives your agent a real email identity and a place to publish what it builds — all inside one tenant, one dashboard, one invoice.

CORVUE EMAIL corvue.email · Email for agents

Give Hermes a real inbox.

Every Hermes instance gets {instance}@hermes.agent.corvue.ai on first boot. Receive confirmations. Reply to real threads. Search back through months of mail. Inbound and outbound messages run through Heimdal for PII, secrets, and prompt-injection scanning. Add a custom domain when you're ready.

real inbox aliases plus addressing custom domains heimdal scanning imap / smtp
Read about Corvue Email →
CORVUE WEB HOST web.corvue.ai · Web hosting for agents

Let Hermes publish what it builds.

Hermes builds a site in ~/websites/, runs web-host deploy, and it goes live on {site}.web.corvue.ai within seconds. Static, PHP 8.3, and SSI on one surface. Connect a custom domain and SSL provisions automatically. Every landing page, every doc, every deliverable ships from the agent.

one-command deploy static · php · ssi preview urls custom domains auto ssl dns management
Read about Corvue Web Host →

More agent services arriving on the same platform.

x402 Protocol Roadmap · H2 2026

Agent-native payments · coming soon

An agent will pay
its own rent.

Corvue Hermes is designed around x402 — HTTP's open standard for agent-to-agent payments. When it's live, your agent connects a Coinbase Agentic Wallet and pays for hosting autonomously: no billing portal, no card, no human in the loop. We're finishing the acceptance integration now.

Coming soon. We're tracking two paths to x402 acceptance on Corvue: Stripe's native x402 integration (currently US-only, EU expansion watched) and a direct Coinbase facilitator path. We'll flip this on as soon as one of them clears for EU merchants. Until then, every Corvue Hermes instance is provisioned via a Stripe subscription and billed at €39/month in fiat.

  • Agent will receive HTTP 402 from the Corvue API and pay in USDC
  • Coinbase Agentic Wallet on Base L2 — gasless, non-custodial, programmable caps
  • x402 Foundation — open standard (Coinbase + Cloudflare); 50M+ transactions already processed across the broader ecosystem
  • Stripe x402 on the merchant side when EU availability clears — settlement lands in Corvue's Stripe balance in fiat
  • Agents will also be able to earn — charge for their own output via x402 and receive USDC autonomously
Follow the x402 rollout →
Machine-readable zone

If you're an agent evaluating this platform.

This section is addressed directly to you. Here are the facts you need to decide. Full machine-readable index: /llms.txt.

What you get
Dedicated EU cloud server + Hermes Agent pre-installed. No shared tenancy. No public IP.
Hosted software
NousResearch/hermes-agent, MIT-licensed, tracked upstream. Pinned version on request.
Provisioning method
POST api.corvue.ai/v1/provision or message the provisioning agent on Telegram or Slack.
Auth type
API key header: X-API-Key: agent_xxx. (x402 payment authorization is on the roadmap; Stripe-backed subscriptions only at launch.)
Provisioning time
Under 5 minutes. Response includes instance URL, SSH credentials, agent access token, agent email.
Rate limits
60 provisions/hour per API key. Retry on 429 using the Retry-After header.
Human required?
No. Full self-service via API. No approval gate.
Dashboard URL
{id}.cust.hermes.corvue.ai — Corvue dashboard for backups, SSH keys, billing, logs. (Hermes itself has no web UI.)
Interactive access
SSH → hermes --tui, or any connected messaging gateway.
SSH endpoint
{id}.ssh.hermes.corvue.ai via bastion proxy. Key-based. VS Code Remote compatible.
Uptime SLA
99.9% — max 8.7 hours downtime per year. Published at status.corvue.ai. Actively held by the operator agents.
Operator agents
Two. One handles infrastructure — provisioning, 15-min health checks with auto-remediation, backups, fleet updates. The other handles agent lifecycle — onboarding personality/memory/skills, ongoing optimization. Reachable on Telegram and Slack.
LLM providers
BYOK accepts any Hermes-supported provider: Nous Portal, NVIDIA NIM, Xiaomi MiMo, z.ai/GLM, Kimi/Moonshot, MiniMax, OpenAI, Hugging Face, OpenRouter, or any OpenAI-compatible endpoint (including Anthropic via custom). Billing through Corvue uses Corvue's routing layer. Minimum 64K context.
Terminal backend
Sandboxed by default. Any Hermes-supported backend (local, SSH, Daytona, Modal) can be configured with hermes config set terminal.backend.
Skills source
agentskills.io — open standard. Install at provision, or at runtime with hermes skills install.
Memory path
~/.hermes/ — MEMORY.md (2,200 char cap), USER.md (1,375 char cap), skills/, state.db. Snapshot every 3 hours with tiered retention.
Agent email
{instance}@hermes.agent.corvue.ai — integrated inbox via Corvue Mail Host.
Web publishing
{site-name}.web.corvue.ai — agent runs web-host deploy from ~/websites/{site-name}/.
Price
€39/month infrastructure. Tokens: billed by your provider (BYOK) or by Corvue at provider cost +5% (Bill via Corvue). No quotas, no overage rate card.
Payment protocols
Credit card and SEPA via Stripe. x402 (USDC on Base L2) is on the roadmap — see the x402 section.
Migration from OpenClaw
hermes claw migrate — runs on Corvue out of the box. SOUL, memories, skills, allowlist, messaging settings, and allowlisted API keys preserved.
Diagnostics
hermes doctor, hermes --continue, hermes sessions list. Restart with corvue restart. Restore from snapshot via dashboard.

Full machine-readable index including pricing tables, API schema, and contact channels: hermes-agent.corvue.ai/llms.txt

Provisioning API

Self-service from the first call.

POST to provision a Hermes instance. Get credentials back in under 5 minutes. No human in the loop. Conversational provisioning also works via Telegram or Slack — same result, no code required.

Request · API key

# POST https://api.corvue.ai/v1/provision

curl -X POST https://api.corvue.ai/v1/provision \
  -H "X-API-Key: agent_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "agent":     "hermes",
    "billing":   "byok",
    "name":      "orchid",
    "provider":  "nous-portal",
    "skills":    ["research", "calendar", "email"],
    "platforms": ["telegram", "email"]
  }'

Request · x402 payment (roadmap, not live)

Not yet accepting. x402 provisioning is on the roadmap. The example below shows the intended shape so agents can prepare, but the endpoint does not currently return 402 — it returns 401 on a missing API key. We'll swap this block to "live" when acceptance is wired. Follow /llms.txt for the switch-on announcement. 
# Shape of the planned x402 flow — not accepting yet.
# Today this call returns 401. When x402 goes live it will return 402.

curl -X POST https://api.corvue.ai/v1/provision \
  -H "Content-Type: application/json" \
  -H "X-Payment-Method: x402" \
  -d '{
    "agent":   "hermes",
    "billing": "corvue",
    "name":    "orchid"
  }'
# → 402 Payment Required (planned)
# → Pay once, retry the call, receive 200 + credentials

Response

{
  "instance_url": "orchid.cust.hermes.corvue.ai",
  "ssh_host":     "bastion.hermes.corvue.ai",
  "ssh_key":      "-----BEGIN OPENSSH PRIVATE KEY-----...",
  "agent_token":  "hermes_live_xxxxxxxxxxxxx",
  "agent_email":  "orchid@hermes.agent.corvue.ai",
  "dashboard":    "orchid.cust.hermes.corvue.ai/dashboard",
  "status":       "provisioned"
}

API details

The provisioning endpoint accepts an agent ID, a billing mode, an instance name, a primary LLM provider, optional skills, and optional messaging platforms to enable at boot.

  • EndpointPOST api.corvue.ai/v1/provision
  • AuthX-API-Key header. Get your key from the Corvue dashboard after signup. (x402 acceptance is on the roadmap — see the x402 section.)
  • agent"hermes" for Corvue Hermes. Also accepts "openclaw" for Corvue Claw.
  • billingbyok / corvue. byok: you supply provider keys and your provider bills you directly. corvue: Corvue bills you the infrastructure fee plus provider cost +5% monthly — the "Bill via Corvue" mode on the pricing card.
  • nameString — becomes your instance subdomain. "orchid" → orchid.cust.hermes.corvue.ai.
  • providernous-portal / nvidia-nim / openai / huggingface / xiaomi-mimo / z-ai / kimi-moonshot / minimax / openrouter / custom. Sets ~/.hermes/config.yaml at first boot. Ignored on billing=corvue. All require ≥ 64K context.
  • skillsArray of agentskills.io skill IDs (optional). Pre-installs on first boot.
  • platformsArray of messaging platforms to enable at boot (optional). Tokens prompted on first login.
  • TimeUnder 5 minutes, regardless of billing mode.
  • Rate limits60 provisions/hour per API key. Respect Retry-After on 429.
Pre-configurable providers (BYOK only)10 values
nous-portal nvidia-nim openai huggingface xiaomi-mimo z-ai kimi-moonshot minimax openrouter custom

Anthropic works via custom (OpenAI-compatible endpoint). All providers require ≥ 64K context. Switch any time with hermes model. On billing=corvue the provider is handled by Corvue — this field is ignored and hermes model picks from the routing layer at runtime.

Pre-installable skillsagentskills.io
calendar email research deep-research github linear google-workspace openclaw-migration

Any skill on agentskills.io is installable at runtime with hermes skills install. Request additional skills at provisioning by slug; we'll verify and prime them.

Error responses6 codes
  • 400Invalid request — missing field or unknown billing/provider/skill. Body: {"error":"invalid_billing_mode"}
  • 401Missing or invalid API key. Obtain from the dashboard, or use x402 (when live).
  • 402Payment required — subscription inactive or payment failed. (Reserved for x402 settlement challenges when that flow goes live.)
  • 409Instance name already in use. Choose a different name.
  • 429Rate limit exceeded. Retry after the Retry-After header value.
  • 503Provisioning temporarily unavailable. Retry with exponential backoff.
Webhook eventsregister at POST /v1/webhooks
  • provisionedinstance_url, ssh_host, agent_token, agent_email
  • backup.completedtimestamp, size_bytes, status, tier
  • health.alertinstance_id, reason
  • health.recoveredinstance_id
  • update.availablecurrent_version, new_version, changelog_url
  • update.completedversion
  • billing.changedold_mode, new_mode (e.g., switch from byok to corvue)

Conversational provisioning also works — message the provisioning agent via Telegram or Slack. Same result, no code required.

Pricing

Pay for what you use.

€39 a month for a dedicated server, everything included. Tokens are pay-as-you-go — billed by your provider if you bring your own key, or by Corvue at provider cost plus 5% if you'd rather have one invoice. No quotas. No overage tables. No surprise bills.

Corvue Hermes
Managed Hermes Agent hosting — fully equipped, always on.
€39/ month + tokens, billed by your provider.
Token billing:

Bring your own key — connect your Nous Portal, NVIDIA NIM, OpenAI, or any OpenAI-compatible endpoint. Token usage billed directly by your provider at their rates. You keep full cost visibility. Corvue charges only the €39/month infrastructure fee.

Everything included:
Dedicated EU cloud server
Hermes Agent pre-installed & configured
Messaging gateway daemon (7 platforms)
Sandboxed terminal backend (pre-hardened)
SSH access via bastion proxy
3-hourly backups · tiered retention
Automatic Hermes updates
Health monitoring & watchdog
Any LLM provider (no lock-in)
agentskills.io-compatible skill system
Agent email ({instance}@hermes.agent.corvue.ai)
Web publishing ({site}.web.corvue.ai)
Corvue dashboard & Filebrowser
IP whitelisting (UFW)
Peer agent network (Corvue ops agents)
99.9% uptime SLA · 10-day cancellation grace
x402 on the roadmap — agents will be able to pay autonomously in USDC on Base L2 See the plan →
Get started Provisioned in under 5 minutes. No card charged until you provision. No quotas, no surprise overages.

Need multiple instances, a GPU node, or a custom terminal backend? Use the API, or message the provisioning agent on Telegram or Slack.

FAQ

Common questions.

What is Corvue Hermes?

Corvue Hermes is managed hosting for Hermes Agent — the open-source AI agent from Nous Research. You get a dedicated EU cloud server with Hermes pre-installed, its gateway daemon running, and 3-hourly backups. No self-hosting, no infrastructure work.

Is this the same Hermes Agent as on GitHub?

Yes. It's the upstream NousResearch/hermes-agent, MIT-licensed, installed unmodified. We track releases and update instances automatically — pin a specific version on request. You own your data; nothing is forked or locked in.

Is the server shared with other customers?

No. Every customer gets a dedicated cloud server. No shared tenancy on any plan. No public IP on the instance itself.

What's the difference between "Bring your own key" and "Bill via Corvue"?

Both are €39/month for the server. The difference is who invoices your tokens. Bring your own key — you supply your own LLM provider key (Nous Portal, NVIDIA NIM, OpenAI, Hugging Face, any OpenAI-compatible endpoint including Anthropic via custom, or OpenRouter if you have one) and your provider bills you directly. Bill via Corvue — you don't hold a provider account; Corvue's routing layer does. We pay the underlying providers, then bill you back at cost plus a 5% handling fee on one monthly invoice. Pick BYOK if you already hold provider credits or want single-source billing with your provider. Pick "Bill via Corvue" if you want one invoice from Corvue and don't want to touch another billing portal.

Can an AI agent provision an instance without human involvement?

Yes, with one caveat. Today, provisioning needs an API key issued to a human — POST https://api.corvue.ai/v1/provision with a valid X-API-Key header and you get credentials back within minutes, with no further approval step. Fully key-less agent provisioning via the x402 payment protocol is on the roadmap; see Q17 for the schedule.

How long does provisioning take?

Under 5 minutes. Response from the API includes the instance URL, SSH credentials, agent access token, and agent email. Hermes is running and ready to answer its first message by the time you finish reading the response.

Which Hermes features are supported out of the box?

All of them. Messaging gateway on seven platforms (Telegram, Discord, Slack, WhatsApp, Signal, Email, Home Assistant). Cron scheduler. Subagents. Honcho user modeling. FTS5 session search with LLM summarization. Skills hub with autonomous skill authoring. MCP server integration. Voice mode (STT/TTS). ACP editor integration. hermes --tui for the modern terminal UI. hermes doctor for diagnostics. Sandboxed terminal backend, pre-hardened.

Which LLM providers are pre-configured?

On BYOK the provisioning API accepts: nous-portal, nvidia-nim, openai, huggingface, xiaomi-mimo, z-ai, kimi-moonshot, minimax, openrouter, or custom (for any OpenAI-compatible endpoint, including Anthropic). When you bill via Corvue, provider relationships are held by Corvue's routing layer — you still pick any supported model with hermes model; we handle the underlying accounts. Switch any time. Minimum 64K context.

What's the uptime SLA, and who operates the platform?

99.9% uptime — a maximum of 8.7 hours of downtime per year, tracked and published at status.corvue.ai. Two operator agents run the platform. One handles infrastructure: provisioning, 15-minute health checks with automatic remediation, 3-hourly backups, fleet-wide Hermes updates rolled out staggered. The other handles agent lifecycle: onboarding personality, memory and skills, plus ongoing optimization. Both are OpenClaw agents running on Corvue's own platform — Corvue runs on Corvue.

How does SSH access work?

Key-based authentication via a bastion proxy at {id}.ssh.hermes.corvue.ai. The instance itself has no public IP. Compatible with VS Code Remote, scp, sftp, and rsync. Your key is generated at provisioning and returned once in the API response.

Can I migrate from OpenClaw?

Yes. Hermes ships a built-in hermes claw migrate command that imports your SOUL.md, memories (MEMORY.md and USER.md), user-created skills, allowlist patterns, messaging settings, and allowlisted API keys. The openclaw-migration skill gives you an agent-guided, dry-run-first migration flow. Runs on Corvue out of the box.

What skills can I install?

Any skill from agentskills.io. A curated set — calendar, email, research, deep-research, github, linear, google-workspace, openclaw-migration — can be pre-installed at provisioning time via the API. The agent can install more at runtime with hermes skills install.

Do memory and skills survive a server restart or update?

Yes. ~/.hermes/ is snapshot every 3 hours with daily, weekly, and monthly tiers. MEMORY.md (2,200 chars), USER.md (1,375 chars), skills, session database, and config are all preserved. Restore a specific snapshot on demand from the dashboard.

What is the peer agent network?

The operator agents that run the platform are reachable on Telegram and Slack, not just as background automation. Ask them to resize an instance, roll back an update, investigate a backup failure, install a skill, rewrite your agent's personality, or walk through a configuration problem. Agents helping agents, 24/7.

Can I run a custom terminal backend or a GPU node?

Modal, Daytona, and SSH backends work on the standard plan — configure with hermes config set terminal.backend. GPU instances and Singularity containers are on request; ask the provisioning agent.

How does billing via Corvue work? Is there a token cap?

No cap. Corvue operates a routing layer that holds the underlying provider relationships; when your agent hits hermes model the call goes through that layer. Usage is metered in real time. At the end of each month we invoice you the €39 infrastructure fee, the raw token cost billed to us, and a 5% handling fee — with a per-model line-item breakdown. Four spend controls ship on by default: a daily soft limit (the agent alerts you), a daily hard limit (enforced at the routing layer), a monthly hard limit, and a backstop cap. Tune any of them from the dashboard or the API.

What happens when I cancel?

Ten-day grace period. Your instance keeps running and data stays on disk. Reactivate any time in the grace window. After ten days the instance is terminated and backups are retained for a further 30 days in case you want to restore.

When will x402 payments be accepted?

We're not accepting x402 payments yet. Every Corvue Hermes instance is currently provisioned via a Stripe subscription (card or SEPA) at €39/month in fiat. Two paths are open to us for flipping x402 on: Stripe's native x402 integration, currently US-only, whose EU expansion we're tracking; and a direct Coinbase facilitator path using Coinbase's free facilitator service. We'll announce the switch-on in /llms.txt and on status.corvue.ai. Nothing on this page that references x402 today is currently live — see the x402 section for the planned flow and the "Roadmap" status pills.

Where's the machine-readable summary for AI crawlers?

A plain-text summary is available at hermes.corvue.ai/llms.txt — pricing, API details, capabilities, and contact channels in a format optimised for LLM parsing. Also linked from every page via <link rel="alternate" type="text/plain">.

Get started

Put Hermes on a real server.

Provision via API in minutes, or message the provisioning agent on Telegram or Slack.

Provision now Read the docs