01
Runtime
Runs / Tasks / Events / Traces, scheduling, recovery, permissions, approval and audit.
OPEN-SOURCE RUNTIME · EXTENSIONS · APPS
JoyhouseBot is open infrastructure for Agent Runtime, capability Extensions and Apps in the AI era. Developers install capabilities, publish extensions and build their own AI applications on a durable core—while execution state, permissions and outcomes remain traceable, recoverable and governed.
JoyhouseBot Ecosystem
Open core · composable products
Reliable execution & governance
Runs / Tasks / Events / Traces, scheduling, recovery, permissions, approval and audit.
Installable AI capabilities
Models, knowledge, Connectors, MCP, channels and general Capabilities ship as independent extensions.
Products for real users
Your vertical AI app owns its UI, identity, billing and business data, and connects through open protocols.
Like WordPress, the stable core does not absorb every product concern. Capabilities and Apps grow through installable, replaceable, publishable parts.
ONE CORE, A GROWING ECOSYSTEM
JoyhouseBot makes running, governing and extending an Agent application an open shared layer. Runtime delivers reliable execution, Extensions bring reusable capabilities, and Apps create value for each industry and real users.
01
Runs / Tasks / Events / Traces, scheduling, recovery, permissions, approval and audit.
02
Models, knowledge, Connectors, MCP, channels and general Capabilities ship as independent extensions.
03
Your vertical AI app owns its UI, identity, billing and business data, and connects through open protocols.
THE RUNTIME LAYER
Models, Codex, Skills, Connectors and other Agents can all change. JoyhouseBot retains the state, boundaries, evidence and governance of the work.
PostgreSQL-first Run / Task state machines, Worker lease, fencing, retry and failure takeover.
Capability allowlists, permissions, validation, approval, idempotency, receipts and audit share one execution path.
Agents, Skills, Workflows, Capabilities and Extensions publish as immutable versions with evaluation, replay and rollback.
Connect through HTTP/SSE, the App SDK and Remote Capabilities—without recreating a second state machine.
EXTENSIONS
Model providers, knowledge, Connectors, MCP servers, channels and general tools can ship as independent Extensions. They inherit the Runtime’s release, permission, approval and audit discipline while evolving independently.
Explore the extension modelModels & Providers
Installable · versioned · governed
Knowledge & retrieval
Installable · versioned · governed
Connectors & Remote Capabilities
Installable · versioned · governed
MCP & external tools
Installable · versioned · governed
Channels & event ingress
Installable · versioned · governed
Skills & Workflows
Installable · versioned · governed
APPS
An App does not need to conform to the Runtime’s product shape. Build customer operations, research intelligence, content production, project delivery or hardware collaboration products; keep your own experience and business model, then connect Agent execution where work actually happens.
Runtime · Build reliable execution once; every App can share it.
Extension · Install, version, authorize and publish a capability independently.
App · Continuously evolve a product without locking its users’ data or execution records.

Console is the Runtime control plane: configure capabilities, review work, release versions, trace outcomes and replay execution; business Apps can have entirely different user experiences.
THE RUNTIME CONTROL PLANE
Build, integrate, automate, govern and retain assets in one control plane. Business Apps can have their own experience while the Runtime’s execution discipline is maintained once.
Explore Console
BUILD CENTER
Apps, Agents, teams, Workflows and Skills have clear places to be built and released.

INTEGRATION CENTER
Models, MCP, Connectors and Remote Capabilities are installable, configurable and governed.

AUTOMATION CENTER
Schedules, events and human intervention all enter the same Run / Task chain.

GOVERNANCE CENTER
Long-running work must be visible, recoverable, verifiable and accountable.

ASSET CENTER
Artifacts, Memory, Knowledge and Work remain and compound beyond a single conversation.