PaaS // AI Agents & Long-running Backend Jobs

The engine room for
agentic workflows.

No timeouts. Persistent state built in. Zero-trust egress. From your code to a live URL in one command.

Start deploying Read the docs
claude — denkops MCP — fra-1 LIVE
denkops MCP sessie 0s ctx 12% 0 tok cost $0.000 denkops · connected
Running in production at
customer logo
customer logo
customer logo
customer logo
customer logo
01 — Capabilities

Built for jobs that
outlive a request.

01 / Persistent state

Ephemeral code.
Permanent memory.

Every slot comes with durable local storage. Your agent's state survives redeploys, crashes and restarts — no external database to wire up.

/data/store.bin184 MB
/logs/run-4417.jsonl96 KB
02 / No timeouts

Serverless is for
websites. Agents
need time.

Long crawls, batch inference, overnight jobs — they just run. No 15-second guillotine, no cold-start penalty.

VERCEL FUNCTIONS✕ KILLED AT 15 s
DENKOPSRUNS UNTIL EXIT →
03 / Zero-trust egress

Default: blocked.

All outbound traffic is denied unless you whitelist the domain. Your agent can't leak, mine, or phone home.

14:24:24 DENY pool.minexmr.net:3333 not whitelisted
14:24:25 DENY 185.220.101.4:9001 not whitelisted
14:24:26 ALLOW api.openai.com:443 whitelist:openai
14:24:27 DENY 203.0.113.9:443 not whitelisted
14:24:28 ALLOW api.openai.com:443 whitelist:openai
04 / Deploy
1cmd

[BUILD] → [LIVE]

Bring any language. We build and run it — no Dockerfile, no config.

Fig. 04 — storage module, fra-1 rackSSD cartridge → steel rack
02 — Quickstart

Zero to live in
three commands.

Step 01 — Install

One binary. No daemon, no account wizard.

Step 02 — Login

One-time MCP token. Claude talks straight to the control plane.

Step 03 — Deploy

Live URL, logs stream straight to your terminal.

# install
$ curl -fsSL denkops.sh | sh
 
# login (one-time)
$ denkops login
✔ token stored · 0600
 
# ship it
$ denkops deploy
→ live at https://my-agent.denkops.app
Fig. 05 — egress master switch, detailMacro photo — steel toggle switch / hardware detail
03 — Spec sheet

Read the spec.
Then decide.

Spec DenkOps Vercel Fn AWS Lambda
Max run timeUnlimited15 s15 min
Persistent stateLocal disk, durableNoneEFS (config)
Cold startNone (warm slot)100–800 msUp to 5 s
Outbound trafficBlocked by default, per-domain whitelistOpenOpen (VPC = config)
SetupZero config, any languageFramework presetsZip or container
Pricing modelPer slot, flatInvocations + GB-hrsRequests + GB-s
04 — Pricing

No bullshit credits.

Capacity selector
12
Slots
5204060

5 slots = €49 · every additional slot €9 · cancel monthly · no invocation math. A slot is one concurrently running worker. Deploys are unlimited.

DenkOps BV — Monthly

Invoice preview · EUR · excl. VAT

BASE — 5 slots€49.00
EXTRA — 7 × €9.00€63.00
Credits / fees / surprises€0.00
Total / mo€112.00
Reserve 12 slots →

05 — FAQ

A slot is one concurrently running worker with dedicated CPU, RAM and durable disk. You pay per slot, flat — deploys, requests and background jobs inside a slot are unlimited.
As long as it needs. There is no request timeout. Long crawls, batch inference and overnight pipelines run until they exit.
On durable local disk attached to the slot, in fra-1 or ams-1. It survives redeploys, restarts and crashes — no external database required.
All outbound connections are denied by default. You explicitly whitelist domains; everything else is blocked and logged, so your agent can't exfiltrate data or mine crypto.
Yes. Bring any language on the paved-road templates, or push your own container. Both run under the same caps, SSL and egress policy.

Deploy your first
agent tonight.

Start deploying See pricing