Plans drafted by AI
Paste a brief. Get a full draft — epics, tasks, owners, a timeline. Review, edit, then materialize. Nothing is committed to your plan until you say so.
AI-drafted · Chat-controlled · Local or cloud
Project plans,
drafted by AI.
Project plans,
drafted by AI.
Steered by chat.
Drop in a brief — shadow-planner builds the full project structure. Then steer it by conversation: re-schedule, re-baseline, query workload, rebalance, all in plain language. Run it on a local model, or bring your own Claude or OpenAI key.
- Brief → plan in one prompt
- Talk to your project
- Local or cloud — your choice
shadow-planner — Accelerated Plan
What makes shadow-planner different
Steered by chat
Ask, query, change. Push a launch by two weeks. Find who's overbooked. Re-baseline an epic. The agent reads your plan live and asks for approval before changing anything.
Your AI, your choice
Run the assistant on a local model via Ollama — fully offline — or connect Anthropic Claude or OpenAI in settings. Switch providers anytime; your plan and SQLite database stay on disk either way.
Drafted by AI
Drop in a brief. Get a project plan.
Tell shadow-planner what you want to build. It drafts the full structure — epics, tasks, owners, a timeline — in seconds. Review, edit, then materialize. Nothing touches your real plan until you say so.
brief.md
markdown
# Mobile App Launch — Q3 ## Goal Ship iOS & Android v1 by mid–August. Public TestFlight by July. ## Constraints - 3 backend devs, 2 mobile devs, 1 designer - Auth must reuse the existing platform service - No new vendor dependencies ## Streams - Mobile foundation - Auth integration - Offline sync - Store readiness
draft plan
awaiting approval
-
Mobile App Launch
- iOS App scaffold 20 PD
- iOS auth integration 16 PD
- Android scaffold 20 PD
- Offline sync engine 21 PD
- Store readiness 8 PD
Steered by chat
Talk to your project.
Once the plan exists, you don’t have to leave the chat to manage it. Ask questions. Make changes. Run analysis. Anything destructive needs your explicit approval — you stay in control.
Chat — Accelerated Plan
qwen2.5:7b · scoped to scenario
connected
Who is overbooked in May?
2 employees over 110% in May.
- 132% Lena Schultz — Mobile App, Customer Intelligence
- 118% Tom Becker — Platform Modernisation
Push the iOS launch by 2 weeks. What breaks?
Moving iOS Store readiness from Aug 15 to Aug 29.
3 dependent tasks shift forward. Customer Intelligence release unaffected.
change · needs approval
How are we doing on Mobile App Launch?
54.6% complete · 198 PD remaining of 419 planned.
Trending 8 days behind plan, pace dropped 38% since Feb.
Ask anything — “rebalance Q3 across the team”, “show critical path”...
- Query: status, workload, critical path, capacity
- Change: reschedule, re-baseline, rebalance, split tasks
- Approve: every destructive op shows a diff first
Local or cloud
One agent. Three engines.
The chat and planning agents are provider-agnostic. Default to a local model that never phones home, or connect a cloud key when you want more horsepower — chosen in settings, scoped per scenario.
Default · Local
Ollama
Runs a model on your own hardware — qwen2.5:7b
out of the box. No API key, no account, nothing leaves the machine.
- Free — uses your hardware
- Works fully airgapped
- Rescue parser keeps small models on-task
Cloud · your key
Anthropic Claude
Paste a Claude API key for stronger reasoning on large plans. Prompt caching keeps repeat calls fast and inexpensive.
- Best-in-class tool use
- Prompt caching built in
- Provider-aware history budget
Cloud · your key
OpenAI ChatGPT
Prefer GPT? Drop in an OpenAI key and the same agent, with the same tools, runs on the Chat Completions API.
- Same agent, same tool surface
- Automatic prompt caching
- Drop-in — no config sprawl
Local stays local. Choose a cloud provider and only the requests you send — scoped to the active scenario — are transmitted to it.
Parallel realities
Ask the AI to plan it three different ways.
Every plan lives in a scenario. Run a baseline, an accelerated, and a re-org plan side by side — ask the agent to draft each one, then compare. Snapshots and templates keep your favorites for next time.
MarAprMayJunJulAug
- Mobile App Launch
- Platform Modernisation
- Customer Intelligence
- Payments Refresh
The plan, materialized
Real Gantt. Real utilization. Real numbers.
The AI drafts and edits the plan, but the plan itself is a proper PM tool. Drag-to-reschedule timelines, capacity-aware utilization tables, and portfolio rollups — all editable by hand whenever you want.
Gantt
A timeline that thinks in dependencies.
Drag tasks to reschedule. Collapse epics. Toggle dependency arrows on or off. Day or month grain. PD vs. PD comparisons inline. Color-coded task types so you can read the chart at a glance.
- Drag-to-reschedule with snap-to-day
- Plan vs. actual person-days
- Collapse and expand by epic
- Dependency arrows you can toggle
Gantt — Accelerated Plan
Resource utilization
See who's overbooked, before they tell you.
A weekly allocation table that shows planned vs. actual person-days per project, per employee. Capacity ceilings highlight overload immediately. No more pivot-table archaeology before each steerco.
- Plan / actual PD by project × employee
- Capacity ceilings highlight overloads
- Roll up by team, label, or manager
- Export to Excel in one click
Utilization
Everything around the plan
Portfolio rollups
RAG status, cost, and person-day totals across every project — at a glance.
Jira-style hierarchy
Project → Epic → Task → Sub-task, with task, story, bug and milestone types.
Audit log
A readable trail of every change — whether you, the chat agent, or the planner made it.
Snapshots & templates
Freeze any plan as a point-in-time record; save reusable templates to bootstrap new scenarios.
Labels & capacity
Group projects and employees with free-form labels; track daily rate and a manager hierarchy.
Import & export
Import employees from Excel and projects from CSV; export any scenario as JSON, Excel, or CSV.
Yours to keep
Your roadmap is your roadmap.
shadow-planner runs on your laptop, your workstation, or a private server you
control. The database is a single SQLite file you back up with cp. Run
the assistant on a local model via Ollama
and nothing leaves the machine at all — or opt into a cloud provider, where only
your scoped requests go out. No telemetry, no analytics, no account needed to use the app.
-
Local-first. Default to Ollama and the app runs fully offline — your network can be airgapped.
-
One file to back up. SQLite means a single file. Snapshot it, version it, encrypt it.
-
Open & audited. Read the code, run the tests — and every change lands in a readable audit trail.
Built with
Boring on purpose, fast by design.
- Python · FastAPI REST API
- SQLite · SQLAlchemy Local-first storage
- Vanilla JS · Chart.js No framework tax
- Ollama · Claude · OpenAI Pluggable LLM providers
- Docker · uvicorn Single-command deploy
- MSIX · Snap Store packaging
Pricing
Free to start. €3 to go unlimited.
Install free from the Microsoft Store or Snap Store and use the full tool right away. Subscribe only once you outgrow the starter limits.
Free
€0
Microsoft Store & Snap Store download
- 1 scenario
- 3 projects
- 3 employees
- Gantt, utilization & portfolio — unrestricted
- Chat & planning agents — unrestricted
- Audit log, snapshots & export
Pro
€3 / month
Stripe subscription — cancel anytime
- Unlimited scenarios
- Unlimited projects
- Unlimited employees
- Every feature in Free, no caps
- Upgrade in-app via Stripe Checkout
- Manage or cancel in the Stripe portal
Get started
Install free in seconds.
Grab it from the Microsoft Store or Snap Store — the app opens right away, no purchase required.
Free download · €3/month unlocks unlimited use