Resume the same Session
Preserve committed Session events across reconnects and interruptions, then recover under an explicit policy.
Durable execution →AwakenWorks · Open infrastructure for production AI Agent applications
Build Agent applications faster.
AwakenWorks builds open infrastructure designed for production AI Agent applications.
Awaken Agents is open-source, self-hostable infrastructure designed for building and operating production AI Agent applications. For AI product teams: start with a real job, a compatible Agent, or an Agent you already own. Your application keeps its UX, domain model, and acceptance rules.
- Tested Managed Agents API baseline
- Official Anthropic SDK
- One durable Session
Open source · stable release coming soon
First useful result One real application creates, runs, reconnects, and reopens the same Session
Product workflow
Set up Awaken and connect an application.
Use Awaken Console to configure and inspect the same Agent and Session that your application uses.
- 01 Connect model
- 02 Publish Agent
- 03 Run Session
- 04 Inspect evidence
- 05 Connect application
Permission decisions and recovery remain attached to the same durable Session.
What is an AI Agent Runtime? →Current Awaken Console · verified release build
Captured from Awaken v1.0.0-dev release UI at revision 50d5035c68456c9106626f748cf4c169c2057beb. Controlled local data is shown; these are product screens, not customer activity.
What Awaken provides
Use one system for Session continuity, execution control, and recovery.
Your application owns its UX, domain model, and acceptance rules. Awaken manages Agent publication, Sessions, permissions, Workers, Sandboxes, and recovery.
Choose where it runs
Choose where control, Session data, credentials, Workers, tools, and Sandboxes run.
Deployment control →Reuse the same Agent
Connect multiple application surfaces and execution paths to the same published Agent and Session authority.
Application reuse →Three concrete adoption paths
Choose the platform work you want Awaken to own.
Awaken Agents is the available product. The high-value entry points are an Agent product that must ship, an internal Agent that must run inside enterprise boundaries, or repeatable custom delivery across customers. Objects and Workforce remain early product directions, not prerequisites.
AI-native product team
Best first pathShip an Agent workbench, copilot, or vertical application
Keep the product experience, domain model, workflow, and result acceptance in the application. Reuse Awaken for the Agent platform lifecycle underneath.
First useful result
One real application creates, runs, reconnects, and reopens the same Session
Open this path →Enterprise internal Agent
Run coding, knowledge, or operations Agents inside an approved boundary
Use Awaken where source code, internal documents, credentials, model routes, or tool execution cannot depend on an external control plane.
First useful result
An authenticated user runs one Agent path without moving control or execution outside the approved boundary
Open this path →Solution provider or delivery team
Reuse one Agent foundation across custom customer solutions
Change the vertical workflow, interface, integrations, and rules for each customer while keeping one platform model for Agents, Sessions, resources, permissions, and execution.
First useful result
Deliver a second application without copying the first project's platform code or creating another Session authority
Open this path →Built on Awaken
Reference products built with Awaken.
Pilot, our DeepSeek workbench reimplementation, and Awaken Design begin with three different product needs. Each keeps the experience and domain model in the application, reuses Awaken Agents execution, and states what is available today.
These pages document our own builds. They are not customer stories and do not represent customer adoption or endorsement.
First-party Reference App
Awaken Pilot
Create a thin Agent application
Give Pilot a Mission. Keep the work in one Session.
Local reference build
See the Pilot reference build →Independent Reference Reimplementation
DeepSeek workbench on Awaken Agents
Build an advanced Agent workbench
Keep the workbench experience. Reuse the Agent execution underneath.
Local deterministic reference demo
See the reference reimplementation →First-party Reference App
Awaken Design
Build a domain-rich AI product
Turn one AI-generated design into a reviewable, revisable, explicitly accepted result.
Local source-tested reference build
See the Awaken Design reference build →Available now and product direction
Agents is available. Objects and Workforce are early product directions.
Awaken Agents Open source · stable release coming soon Execution and recovery Awaken Agents is open-source, self-hostable infrastructure designed for building and operating production AI Agent applications. → Awaken Objects Early preview Business context Give people, applications, and Agents the same business facts and actions. → Awaken Workforce Early preview Ownership and acceptance Keep ownership, exceptions, and acceptance connected until the work is done. →
Get started
Run Awaken or discuss a private deployment.
Use the quickstart for an independent evaluation. Contact us if control, data, credentials, Workers, or Sandboxes must run in dedicated infrastructure.