Workunit gives every AI model the same brief, the same live task state, and the same persistent memory.
Create a workunit once, attach the context to it, and let Claude, Codex, Gemini, cloud agents, and your teammates continue from the same source of truth instead of rebuilding the story every session.
Typed context atoms keep decisions, attempts, questions, and progress attached to the work.
Work starts from a problem statement and definition of done, not a loose pile of tasks.
MCP clients and cloud agents can act on the same workunit without re-briefing.
The workflow stops depending on manual handoffs.
Instead of maintaining handoff docs and retelling the same story, the workunit becomes the source of truth.
Problem statement, success criteria, tasks, and linked assets stay with the workunit from day one.
Humans and AI models can both see what's blocked, in progress, or done before they touch anything.
Decisions, insights, questions, attempts, and progress notes are searchable instead of disappearing into chat history.
Assets, documentation, and directories give the next model a map of the project.
Keep the workunit, the assets, and the trail of thought in one persistent record.
Workunit is built around persistence. The problem statement, linked people and systems, organized knowledge, and structured context atoms stay attached so the next session has real context instead of starting from scratch.
Structured context atoms
Save decisions, attempts, and insights as typed records that future sessions can trust.
Problem-solving workunits
Start from the problem and definition of done, then break it into tasks once the scope is clear.
Connected assets and directories
Products, systems, people, knowledge, and organized folders stay linked to the work they inform.
Persistence across sessions
The same context survives from early exploration to implementation, review, and later follow-up.
Typed context atoms
A searchable trail of thought built from decisions, attempts, and insights.
Problem-first workunits
Every workunit begins with why it matters and what done looks like before execution starts.
Assets, docs, and directories
Organize knowledge and link it to the work so models see more than a ticket title.
Continuity from first draft to ship
You don't lose the early reasoning once the work turns into implementation and review.
Multi-model handoffs
Plan in Claude, implement in Codex, review in Gemini, and keep the same task state and context throughout.
Explore Mode
Cloud agents can read the real codebase and suggest grounded tasks with file references and effort estimates.
Implement Mode
Launch an agent that clones the repo, writes code, runs tests, and opens a pull request from the browser.
Context-aware help
Guided onboarding, docs, and workspace context help the next user or model start with less ambiguity.
The product supports both guided setup and execution, not just planning.
From quick-start onboarding to MCP-connected clients to cloud execution, Workunit gives your team practical ways to move work forward without requiring everyone to live in the same tool or the same terminal.
How it feels different
Instead of asking a founder to reverse-engineer the repo, Explore Mode can read the real codebase and propose actionable tasks.
Instead of moving work into a separate AI playground, Implement Mode can execute against the actual repository and return a pull request.
Instead of each model inventing its own state, MCP-connected tools can read the same workunits, tasks, assets, and context atoms.
Built for small teams who want visibility, lightweight ownership, and fewer meetings.
Workunit keeps collaboration focused and useful: clear tasks, current status, recurring check-ins, shared context, and transparent workunit history. Enough structure to coordinate, without turning the product into admin overhead.
Task ownership and status
Assign work, mark blockers, and keep progress visible without creating a second job in project administration.
Project calendar
Schedule events, track milestones, and see workunit and task due dates on a shared timeline with recurring events, attendee RSVPs, and iCal export.
Recurring check-ins
Collect async updates on a schedule and link responses back to projects, workunits, and assets.
Project chat
Real-time group messaging within each project for quick questions, status updates, and casual coordination with quote-reply.
Signals for async discussion
Post announcements, questions, ideas, and updates within a project without switching to a separate chat tool.
Activity timeline
See what changed, why it changed, and who moved the work forward across humans and AI.
Transparent handoffs
Organization members and connected models can continue from the same shared context instead of waiting for a recap.
Task ownership is visible
You don't need to ask who owns what or whether something is actually blocked.
Check-ins stay linked to the work
Recurring updates can be attached to projects, workunits, and assets so progress stays searchable later.
Progress has a readable history
The next person, or the next model, can inspect the timeline and continue without a custom meeting recap.
Chat keeps coordination in context
Quick questions and status updates stay inside the project instead of scattered across Slack channels and DMs.
Signals capture decisions that last
Announcements, questions, and ideas are categorized and searchable — not buried in a chat scroll.
Claude Code, Codex, Gemini, and other MCP tools can work from the same workspace memory.
Model Context Protocol is what ties it together. Once connected, AI tools can create workunits, manage tasks, link assets, search the workspace, organize directories, and save context atoms without leaving their native interface.
Claude Code, Codex, Gemini
Use the same workunit from the AI clients your team already prefers.
Workunit and task tools
Create workunits, break them into tasks, update status, and keep delivery moving from chat or CLI.
Assets, directories, and search
Link systems and docs, organize knowledge, and retrieve the right context instead of pasting it manually.
Context tools that persist
Save structured context atoms so what the model learned remains available to the next session.
The core product stays reliable even when the AI layer is ambitious.
Core planning and collaboration stay fast and dependable, even as AI features grow.
Security first
Industry-standard protections, OAuth-based tool connections, and a clear commitment to privacy.
Fast core performance
Go, gRPC, and PostgreSQL keep projects, workunits, and task workflows responsive.
Graceful degradation
Core PM flows continue even when AI-specific systems or cloud execution are unavailable.
Built for real usage
Responsive layouts keep the product usable on desktop, tablet, and phone.
Quick Start
Create your first workunit, add tasks, and connect your AI assistants in minutes.
MCP Integration
Connect Claude Code, Codex, Gemini, and other MCP clients to the same workspace.
Cloud Execution
Use Explore Mode for grounded task discovery and Implement Mode for PR-ready execution.
Calendar
Schedule events, manage recurring meetings, track milestones, and see all due dates on a shared project timeline.
Check-ins and Assets
Keep async updates, linked systems, and organized knowledge attached to the work.
If you're tired of rebuilding context from scratch, start with one workunit and see the difference.
You don't need to migrate your entire stack on day one. Create one workunit, connect one AI tool, run one handoff or one cloud task, and see how much context you keep.