FOR SOLO FOUNDERS BUILDING WITH AI

Project Memory for AI Coding Workflows

Keep tasks, decisions, and context across Claude Code, Codex, Cursor, Gemini, and new sessions without re-explaining your product every time.

Start Free — No Credit CardSee Pricing
14-day Pro trial included MCP-native Works across multiple models
What survives every session

Workunit keeps context attached to the work itself

The brief, the current state, and the why stay with the work.

Problem and success criteria stay visible

Task state stays current

Decisions and failed attempts live inside the workunit

AI can generate the work. It still forgets why you're building it.

Workunit keeps tasks, decisions, attempts, and source material where they belong — inside the workunit.

Without Workunit

Claude Code

recap again

"Paste the repo rules and current decisions first."

Cursor Notes

copied forward

The same architecture summary keeps getting pasted into every fresh session.

Tasks

out of sync

Done in the repo. Todo in the tracker. Unknown in the next AI session.

Slack Handoff

summarize first

"Can you explain what changed before I touch billing?" becomes a recurring ritual.

context-handoff.md

manual copy

You're maintaining a ritual, not a system.

Add Stripe Subscription Billing

Workunit Active · High · billing stripe

Problem to Solve

The app is free-only with no way to collect revenue. Users keep asking for a Pro tier but there is no billing system in place. Integrate Stripe so the product can charge for subscriptions and grow sustainably.

Success Criteria

Users can upgrade to Pro via Stripe Checkout, subscriptions renew automatically, and the app correctly gates Pro features based on billing status. Webhooks handle failed payments gracefully.

Progress
2/6 tasks · 2 assets

Tasks (6)

In Progress (2)
Handle Stripe webhook events
Current step for the next AI session
High In Progress
Gate Pro features based on subscription status
Normal In Progress
To Do (1)
Add billing portal link for plan management
Normal To Do
Done (2)
Create Stripe Checkout session endpoint
Updated Mar 5, 14:22
High Done
Build pricing page with plan comparison
Updated Mar 5, 16:10
Normal Done
Won't Do (1)
Support annual billing discount
Low Won't Do

Context (3)

Decision: use Stripe Checkout instead of custom form

Stripe-hosted checkout handles PCI compliance, 3D Secure, and tax calculation out of the box. Less code to maintain.

Attempt: embedded pricing table broke on mobile

The Stripe embedded pricing table rendered inside an iframe that didn't resize. Switched to a custom pricing page calling the Checkout API directly.

Insight: 73% of Pro interest comes from solo founders

Survey data shows solo founders value unlimited workunits most. The free tier limit should cap workunits, not seats.

Linked Assets (2)

Payment Service

System · Handles Stripe API calls and webhook processing

Stripe API Docs

Knowledge · Official integration guide and webhook reference

Works with your entire stack

Claude, GPT, Gemini, Cursor, VS Code, OpenRouter, and the tools around them can all pick up the same workspace memory.

Models
OpenAI
Anthropic
Google Gemini
DeepSeek
Mistral
Moonshot
Perplexity
Meta
Cohere
Groq
Editors, agents, and runtimes
Cursor
GitHub
Zed
Windsurf
OpenRouter
Sprites.dev
Cline
Z.ai
Ollama
Cerebras
VS Code
One source of truth across models and tools

One workspace. Every AI model on the same page.

Use any model or coding tool you want. Workunit keeps the brief, task state, decisions, and surrounding context in one place so the next session starts with context instead of catch-up.

STEP 1

Frame the work once

Every workunit starts with a clear problem and a definition of done. No model has to guess your intent from a chat log.

STEP 2

Track state across sessions

Tasks, progress, and blockers live inside the workunit. When a new session starts, the model sees what's finished and what needs attention next.

STEP 3

Keep the reasoning, not just the code

Context atoms capture decisions, failed attempts, and insights so the next model inherits the why — not just the what.

STEP 4

Link the surrounding context

Assets connect the systems, people, and docs that matter. The workunit gives a model real project context, not just a task list.

Organization context before the prompt

AI That Knows Your Business

Most AI tools only see your current conversation. Workunit lets you describe your entire organization, so AI understands the full context before you even ask.

People

AI knows who's on your team

Skills, roles, availability

Products

AI knows what you're building

Features, versions, roadmap

Systems

AI knows your tech stack

Infrastructure, tools, dependencies

Knowledge

AI knows your standards

Docs, conventions, templates

Stop re-explaining your tech stack, your team structure, or your coding conventions. Document it once, and every AI conversation starts with full context.

Cloud execution for the work you already scoped

Run AI Agents in the Cloud — No Setup Required

No terminal, no git clone, no local environment. Launch an AI agent from the browser — it reads your codebase, proposes tasks, and ships pull requests.

For Product Owners

You want to create realistic, actionable tasks for your team — but you shouldn't need to know every corner of the codebase to do it. Run Explore Mode and an AI agent reads your code, understands the problem, and suggests tasks — with file references and effort estimates.

  • ✓ Reads your real codebase, not a description of it
  • ✓ Suggests tasks with file references and effort estimates
  • ✓ No code is executed — just intelligent analysis
  • ✓ Powered by Explore Mode

For Software Engineers

You want a first-pass implementation without leaving your project management tool. Point the agent at a workunit, and it clones the repo, writes the code, runs the tests, and opens a pull request — ready for your review.

  • ✓ Clones and works in your actual repository
  • ✓ Writes code and runs the test suite automatically
  • ✓ Opens a pull request when done
  • ✓ Powered by Implement Mode
STEP 1

Point any model at your codebase

Pick a repository, choose an AI model, and hit Start — all the workunit context is already attached. No terminal, no git clone.

STEP 2

Watch the agent work in real time

A cloud VM spins up, clones your repo, and runs the AI agent. Every step streams back live — you see exactly what the agent is reading, thinking, and doing.

STEP 3

Accept tasks grounded in real code

An Explore run reads your codebase and proposes real tasks based on actual files. Accept what makes sense, skip the rest.

STEP 4

Ship a pull request, not a plan

An Implement run writes the code, runs the tests, and prepares a PR. You review the diff and merge.

Bring your own API keys — no Workunit execution fees. You pay Sprites.dev for the cloud VM and OpenRouter for the AI tokens. We just coordinate.

Read the Cloud Execution Guide →
For solo builders and small teams using AI daily

Built for Builders Like You

Whether you're the only one carrying the context or coordinating a small team, Workunit is built for people who already use AI to get real work done.

Solo Founders

You're the only one who knows the full context. Workunit makes sure AI knows it too.

Agency & Freelancers

Juggle multiple client projects without losing context. Switch projects, AI switches with you.

Dev Teams

New team members take weeks to ramp up. With shared context, both people and AI start informed.

AI Power Users

Use Claude for code, GPT for writing, Gemini for research. One workspace, every AI tool.

One MCP connection across your AI tools

Connect Any AI Tool in 60 Seconds

Model Context Protocol (MCP) is the bridge between your AI assistant and Workunit.

# Claude Code
claude mcp add --transport http workunit https://workunit.app/mcp

# OpenCode
opencode mcp add && opencode mcp auth workunit

# OpenAI Codex
codex mcp add workunit --url https://workunit.app/mcp

# Gemini CLI
gemini mcp add --transport http workunit https://workunit.app/mcp

Also works with Cursor, Crush, and any MCP-compatible tool.

After setup

Just Talk Naturally

# Quick commands
"Create a workunit for the authentication refactor"
"What tasks are blocking the API project?"
"Mark the database migration task as done"

# Or paste any Workunit URL
"Create a workunit for https://workunit.app/projects/371594b8..."
"Let's implement https://workunit.app/.../workunits/aa39c53b..."
"Add a task to https://workunit.app/workunits/f7e2b..."
Read the full MCP Integration Guide →
Start on free, upgrade when it sticks

Start Small. Upgrade When the Limits Become Real.

The product should be easy to evaluate on a real project. Free gives you room to test the workflow, Pro keeps growing with you once it becomes part of how you build.

FREE

€0

Best for trying Workunit on a real project — no credit card needed.

No credit card

  • 3 projects

  • 25 workunits per project

  • 2 seats

  • 250 MB file storage

  • MCP integration and full API access

Start Free
PRO

€8/user

Per month, once Workunit becomes part of how the team actually operates.

14-day Pro trial

  • Unlimited projects

  • Unlimited workunits

  • Per-seat pricing with seats added as needed

  • 20 GB file storage with add-ons available

  • Priority support once it's part of the workflow

Review Pricing
All plans include MCP integration and full API access.
Storage can grow with add-ons instead of forcing an early upgrade.
Start free, then move up when the workflow proves itself.
Ready to work across sessions

Start with one workunit. Let the product prove itself.

You don't need a migration project or a full process rewrite. Create one workunit, connect one tool, and see how much easier it is when tasks, decisions, and assets survive the next chat.

No credit card
Start without a credit card.
Real workflow
Evaluate it on actual work, not a demo script.
Model-agnostic
Works with the tools already in your stack.

Start free. Upgrade once it becomes part of the workflow.

Create Your First WorkunitView Pricing
No credit card required
Setup in under 2 minutes
Cancel anytime

Need a hand?

Join the discussion on GitHub Discussions or email us at [email protected]