See how we build products on Awaken.

01 · First-party Reference App

Awaken Pilot

Create a thin Agent application

Local reference build · public source pending

Give Pilot a Mission. Keep the work in one Session.

The situation: You have the product idea, but the first prototype still needs task state, progress, approval, schedules, files, and recovery.

Start with: Run one bounded Mission in Plan or Execute mode, then review its progress and result in the same Session.

Best for: Agent product teams that want the shortest path from product UX to durable execution.

See the Pilot reference build →

02 · Independent Reference Reimplementation

DeepSeek workbench on Awaken Agents

Build an advanced Agent workbench

Local deterministic demo · live run not published

Keep the workbench experience. Reuse the Agent execution underneath.

The situation: An interaction-rich workbench needs Presets, Plans, approvals, Trajectory, Threads, and Files. Rebuilding the Agent loop first delays the parts users can actually see and judge.

Start with: Start one Preset-driven Session, review its Plan, approve a tool, and reopen the same Trajectory after an interruption.

Best for: Teams building Agent IDEs, coding workbenches, or other interaction-rich Agent products.

See the reference reimplementation →

03 · First-party Reference App

Awaken Design

Build a domain-rich AI product

Local source and test evidence · live and field gates pending

Turn one AI-generated design into a reviewable, revisable, explicitly accepted result.

The situation: Chat history is not a product model when a team must compare versions, attach feedback to the exact result, reject unsafe output, and deliver one selected design.

Start with: Start from one brief, review a ready immutable Revision, request one change, and accept the exact Artifact to deliver.

Best for: AI-native product teams that need a durable domain model, human review, safe publication, and business facts beyond chat.

See the Awaken Design reference build →