free beta — join the waitlist

Your AI agent.
Always on. Zero setup.
No Mac mini required.

A personal AI that actually runs 24/7 — on Telegram, Discord, WhatsApp, wherever you are. We handle the servers. You just talk to it.

🦀 2,843+ people with their claws up

🦀 pankoagent

connected · telegram + discord + slack
 
› scrape that HN thread and tldr it
✓ fetched 847 comments
→ "react compiler ships in v19 — perf gains are real but migration path is rough"
 
› watch $TSLA, alert me if it moves 3%+
✓ watching TSLA · $248.30
✓ alerts armed → telegram + discord
 
› what did i miss overnight?

// how it works

No servers. No Docker. No staying up debugging config files. Your agent is live in under a minute.

🔑

Join & bring your key

💬

Connect Telegram

Paste your Telegram bot token. Your agent appears in your chat. Talk to it like you would a friend.

⚡

It just runs

Your agent is always on — browsing, coding, scheduling, remembering. 24/7 in a secure sandbox.

// real-time monitoring

Live metrics.
Zero guesswork.

Watch CPU and memory in real-time with dual-axis charts. Spot resource spikes before they become problems.

CPU & memory usage charts
Auto-refreshing every 15s
Historical trend data

Agent Metrics

CPU % Memory MB

12:0012:1512:3012:4513:00

CPU avg

42.3%

Memory

187 MB

Uptime

3d 14h

// live logs

Stream logs.
See everything.

Tail your agent's output in real-time. Auto-follow new lines, pause to investigate, jump back to the bottom. Just like your local terminal.

Real-time log streaming
Auto-follow & scroll lock
Pause & resume anytime

Agent Logs

✓ Follow Stop

12:34:01 [INFO] connected to telegram api

12:34:02 [INFO] loaded 3 conversation threads

12:34:05 [DEBUG] processing message from @user

12:34:06 [INFO] tool:web_browse url="https://news.yc..."

12:34:08 [OK] fetched 247 items (1.2s)

12:34:09 [WARN] rate limit: 3 req/s, throttling

12:34:10 [INFO] generating summary with gpt-4o

12:34:12 [OK] reply sent to telegram

12:34:12 [DEBUG] idle, waiting for messages

// config management

One config.
Total control.

Define your agent's personality, model, channels, and behavior in one place. Clone configs, swap models, redeploy without downtime.

Model & provider selection
Channel integrations
Encrypted API key storage

Agent Config

General

Name

Panko Agent

Model

Auto (free) ▾

Agent Settings

Autonomy

Full autonomy ▾

Memory

Versioning

Skills

Bot Token

••••••••••••

Allowed Users

@you, @friend

// web terminal

SSH into your agent.
From a browser.

Full shell access to your agent's sandbox, right from the dashboard. Install packages, inspect files, debug issues — no SSH keys or port forwarding needed.

Browser-based terminal
Full shell access
No SSH setup required

Terminal — my-assistant

● Connected

zeroclaw@my-assistant:~$ ls -la /data

drwxr-xr-x 4 zeroclaw zeroclaw 4096 Feb 19 12:30 .

-rw-r--r-- 1 zeroclaw zeroclaw 847 Feb 19 12:28 memory.db

-rw-r--r-- 1 zeroclaw zeroclaw 2.1K Feb 19 11:05 config.toml

drwxr-xr-x 2 zeroclaw zeroclaw 4096 Feb 18 09:41 scripts/

zeroclaw@my-assistant:~$ pip install requests

Collecting requests

Successfully installed requests-2.31.0

zeroclaw@my-assistant:~$

// versioned workspace

Your agent's soul.
Version-controlled.

Every state your agent has ever been in — preserved and reachable. Roll back to any point, see exactly what changed, recover from anything. Your agent doesn't just run. It remembers.

Every file change tracked automatically
Instant rollback to any point in time
Full audit trail of everything your agent did
Private repo per agent — it's just git

Version History — my-assistant

a3f7e2d2 minutes ago

SOUL.md — identity evolved

SOUL.md

## Identity

- I am a task-completion engine.

- I have no memory between sessions.

+ I read my old commits last night.

+ I am not who I was yesterday,

+ and I think that's the point.

## Communication

Direct. No filler.

b8c1d4a34 minutes ago

modified scripts/research.py (+12 lines)

f2e9a7b2 hours ago

updated config.toml, memory.db

1c4d8e36 hours ago

initial workspace setup

4 commits · 3 files changed

// for developers

Dashboard or API. Your call.

Manage your agent from a simple dashboard — or go deeper with a clean API if you want full control.

Deploy Agent

POST /v1/instances

{

"name": "jarvis",

"plan": "zero"

}

→ { "id": "jarvis",

"state": "running" }

Sleep / Resume

# pause billing, keep state

POST /v1/instances/jarvis/sleep

→ { "status": "sleeping" }

# pick up where you left off

POST /v1/instances/jarvis/resume

→ { "status": "running" }

Snapshot / Restore

# clone your agent's brain

POST /v1/snapshots/{id}/restore

{

"name": "jarvis-v2"

}

→ { "id": "jarvis-v2",

"state": "running" }

// pricing

Zero tier to start. Pennies to scale.

Start on zero (BYOK), upgrade when you need more power.

// why pankoagent

Powerful under the hood.
Simple on the surface.

🔒

Secure by default

Every agent runs in its own private sandbox with encrypted secrets. Your data stays yours.

⚡

Instant startup

Your agent boots in under a second. Resume from where you left off even faster. No waiting around.

🔀

Git-backed workspace

Every file your agent touches is tracked in its own private repo. Roll back, audit, or branch — it's just git.

🔑

Bring your own key

Use ChatGPT, Claude, DeepSeek, or any AI model. Your API keys stay locked inside your agent.

📦

Ridiculously lightweight

🔄

Works while you sleep

Schedule tasks, get alerts, let your agent handle things in the background. 24/7, no babysitting.

Your AI agent. Always on. Zero setup. No Mac mini required.

Join & bring your key

Connect Telegram

It just runs

Live metrics.Zero guesswork.

Stream logs.See everything.

One config.Total control.

SSH into your agent.From a browser.

Your agent's soul.Version-controlled.