Platform Guide

A comprehensive reference for Jarble's runtimes, LLM providers, messaging integrations, deployment lifecycle, and billing.

Runtimes

A runtime is the containerized environment that powers your bot. Each deployment runs inside its own Kubernetes pod with dedicated CPU, memory, and persistent storage.

OpenClaw (Recommended)

OpenClaw is a full-featured Node.js runtime built for production AI bots. It includes an MCP (Model Context Protocol) server that gives your bot access to tools like render_ui, define_component, and component_reference.

Key capabilities:

• Rich UI rendering -- 37 component types (charts, tables, forms, maps, 3D, code editors)
• Multi-platform messaging -- Telegram, Discord, Slack, WhatsApp
• Skills -- web search, weather, calculator, and custom skills
• Custom components -- define reusable components that persist across conversations
• Sandbox mode -- run arbitrary HTML/CSS/JS in a secure iframe with Three.js, D3, and more

ZeroClaw

ZeroClaw is a lightweight Python runtime designed for simple text-based chat bots. It has lower resource requirements and faster startup times than OpenClaw but does not support MCP tools or rich UI rendering.

Best for:

• Quick prototyping and experimentation
• Bots that only need conversational text responses
• Low-resource deployments

Runtime Comparison

Feature	OpenClaw	ZeroClaw
Language	Node.js 22	Python
Rich UI components	Yes (37 types)	No
MCP tools	Yes	No
Messaging platforms	Telegram, Discord, Slack, WhatsApp	Telegram, Discord, Slack, WhatsApp
Skills	Yes	No
Custom sandbox	Yes	No
Startup time	30--90s	10--30s

LLM Providers

Jarble supports four LLM providers. You can use Included Credits (OpenRouter, billed through Jarble) or bring your own API key from any provider. The wizard auto-detects the provider from the key prefix.

OpenRouter (Recommended)

A unified API gateway providing access to 200+ models from OpenAI, Anthropic, Google, Meta, Mistral, and others. Best for flexibility and cost optimization.

• Key prefix: sk-or-
• Default model: openrouter/auto (routes to the best model per request)
• Available models: GPT-4o, Claude Sonnet 4, Claude Haiku 3.5, Gemini 2.0 Flash, and 200+ more
• Get a key: openrouter.ai/keys

Anthropic

Direct access to the Claude model family. Use this if you have an Anthropic API key or a Claude Max subscription.

• Key prefix: sk-ant-
• Models: Claude Opus 4 (most capable), Claude Sonnet 4 (balanced), Claude Haiku 3.5 (fast)
• Get a key: console.anthropic.com

OpenAI

Direct access to GPT-4o, o1, and other OpenAI models.

• Key prefix: sk- (e.g., sk-proj-...)
• Models: GPT-4o (default), GPT-4o Mini, o1
• Get a key: platform.openai.com

Google AI

Direct access to Google's Gemini model family via AI Studio.

• Key prefix: AIza
• Models: Gemini 2.0 Flash (default), Gemini 2.0 Pro
• Get a key: aistudio.google.com

Messaging Platforms

Connect your bot to one or more messaging platforms from the Platforms tab in your deployment configuration. Each platform requires specific credentials and has its own setup flow.

Telegram

Telegram uses a bot token for authentication and a pairing flow for user authorization.

1. Open Telegram and message @BotFather.
2. Send /newbot and follow the prompts to create a bot.
3. Copy the bot token (a long string of numbers and letters).
4. Paste it into the Telegram field on the Platforms tab and save.
5. Your bot restarts with Telegram enabled. Message your bot on Telegram.
6. The bot sends a pairing code. Jarble automatically polls for pending pairings and approves them.
7. The bot confirms pairing is complete. You can now chat freely.

Discord

1. Go to the Discord Developer Portal and create a new application.
2. Navigate to Bot in the sidebar and click Reset Token to generate a bot token.
3. Enable the Message Content Intent under Privileged Gateway Intents.
4. Use the OAuth2 URL generator to invite the bot to your server with the bot scope and Send Messages permission.
5. Paste the bot token into the Discord field on the Platforms tab.

Slack

Slack requires two tokens: a Bot Token and an App-Level Token with Socket Mode enabled.

1. Create a new app at api.slack.com/apps using the "From scratch" option.
2. Enable Socket Mode in the app settings and generate an App-Level Token with the connections:write scope.
3. Under OAuth & Permissions, add the chat:write, app_mentions:read, and im:history scopes.
4. Install the app to your workspace and copy the Bot Token.
5. Paste both tokens into the Slack fields on the Platforms tab: xoxb- (Bot Token) and xapp- (App Token).

WhatsApp

WhatsApp uses QR code pairing via WhatsApp Web. No Business API or developer account is needed.

1. Open the Platforms tab and click Connect WhatsApp.
2. A QR code appears. Scan it with your phone's WhatsApp app (Settings > Linked Devices > Link a Device).
3. Once linked, your bot can send and receive messages through your WhatsApp number.

Deployment Lifecycle

Every deployment runs as a Kubernetes pod with its own persistent volume. The lifecycle is managed entirely through the Jarble dashboard -- no infrastructure knowledge required.

States

State	Description
Creating	Pod is being provisioned. Dependencies are installing. Typically 30--90 seconds for a first-time deployment.
Running	Bot is live and accepting messages via web chat and any connected messaging platforms.
Stopped	Pod has been scaled to zero. No resources are consumed. Data on the persistent volume is preserved.
Failed	The pod failed to start or crashed. Check the Logs tab for error details. Common causes: invalid API key, npm cache corruption, resource limits exceeded.

Controls

The deployment configuration panel provides three lifecycle controls:

• Start -- Scales the pod from zero to one replica. Resumes from the existing persistent volume (fast restart, no reinstall).
• Stop -- Scales the pod to zero. Frees compute resources while preserving all data.
• Restart -- Scales down then back up. Useful after configuration changes or to recover from errors.

Real-Time Status

Status updates are delivered via Server-Sent Events (SSE). The dashboard badge updates automatically -- no manual refresh needed. When you trigger a start or restart, you see the state transition in real time.

Configuration

After deployment, configure your bot from the sidebar panel. Available tabs depend on your runtime. OpenClaw deployments have the most options: General, Model, Platforms, Skills, Components, Logs, and Advanced.

System Prompt (soul.md)

The system prompt defines your bot's personality, behavior guidelines, and domain knowledge. It is written as Markdown and stored as soul.md on the pod. Edit it from the General tab. Changes sync automatically when you save.

Skills

Skills extend your bot with additional capabilities beyond conversation. Available skills include:

• Web Search -- Search the web and return summarized results.
• Weather -- Look up current weather and forecasts by location.
• Calculator -- Evaluate mathematical expressions.

Enable or disable skills from the Skills tab. Each skill is rendered as a configuration file on the pod and loaded by the MCP server. Additional skills can be added through marketplace services.

Canvas Components

Your bot can render 37 built-in UI component types in the web chat, including charts, data tables, forms, maps, and code editors. Additionally, you can install marketplace components or define custom components that persist across conversations. The Components tab shows installed components and provides access to the marketplace.

Model Configuration

The Model tab lets you change the LLM provider, API key, or model selection without redeploying. Changes trigger a config sync and automatic pod restart.

Logs

The Logs tab streams pod logs in real time. Use this to debug startup failures, LLM errors, or messaging platform issues. Logs are streamed via SSE and update as new entries appear.

Credit System

Free Trial

New accounts receive a free trial with enough credits to create a deployment and test basic functionality. No credit card is required to start.

Billing

Jarble uses Stripe for billing. When using Included Credits, you select a monthly spending cap. Available tiers:

Plan	Best For
$5/mo	Light usage -- testing and small bots
$10/mo	Moderate usage -- a few hundred messages per month
$25/mo	Active usage -- busy bots with frequent conversations
$50/mo	Heavy usage -- high-volume bots and power users
$100/mo	Enterprise -- maximum capacity for production workloads

The spending cap controls how much LLM usage your bot can consume per month. You are only charged for actual usage up to the cap.

Linked Deployments

Deployments can be linked into credit pools. Linked deployments share a single credit balance, which is useful when you have multiple bots serving different channels but want unified billing. The dashboard graph view visualizes credit pool relationships -- nodes represent deployments and edges show pool linkages.

Dashboard

The dashboard is your central control panel for all deployments.

Deployment List

Each deployment shows its name, runtime, current status (with a color-coded badge), and the connected messaging platforms. Click any deployment to open its web chat or configuration panel.

Graph View

Toggle the graph view to see an interactive node diagram of your deployments. Nodes are arranged using the dagre layout algorithm. Circle icons indicate the deployment type (owner, linked, standalone). Edges show credit pool relationships. Use the filter bar to toggle edge visibility or filter by runtime.

Real-Time Updates

The dashboard subscribes to SSE status streams. When a deployment changes state (starting, crashing, completing a restart), the status badge updates immediately without polling or manual refresh.