Intake
brief.parseYoursRisk: lowReads a task brief and proposes the breakdown: subtasks with estimates, plus the open questions.
Needs a brief of at least 40 characters in the description.AI that does the work
Agents do the work. You keep the keys.
Ask in plain words, hand tasks to agents at the autonomy level you choose, and approve the risky changes in your AI Inbox.
- Ask
- Agents
- Skill library
- AI Inbox
- AI memory
Ask box
Plain words, any project you can open. Research, plan changes, or run a skill.
AI project generator
A brief in, a full plan out.
Agents and team packs
Role agents for 15 teams.
Workspace Guide
In every new workspace. @mention it on a task.
Bring your own AI
Claude, ChatGPT or Claude Code over MCP. It works as you and sees only what you can open.
Trust and control
AI Inbox, audit log, undo and AI budgets.
Connect your AI
AlianHub ships an MCP server, so the AI app you already use can work inside your workspace.
Open AI → Connect your AI, copy the address, add it as a custom connector, then ask in everyday words.
- YouAsk in everyday words
- Your own AIClaude · ChatGPT · Claude Code
- AlianHubActs as you, with your role and permissions
- One task — made at once, with UndoWider change — waits in the AI Inbox for approvalEvery change — written in the audit log
Read the Connect your AI guide
It cannot delete tasks, docs, comments or projects, or change who may do what.
Ask
“What should we get done?”
Ask about anything you can open, or hand an agent work. Proposals come back to you first.
What should we get done?
AskAgents
Ask, plan or build. Type / for a skill, @ to add a task, doc or project.
Project status updateWhat moved, what is stuck and what comes next. Create a taskDescribe the work and get a well-formed task to create. Find my urgent workYour open tasks that are overdue, due soon or high priority. Draft a messageAn email or chat message, grounded in your work.
- Scope it. Choose “Everything you can open” or a single project. Ask only reads projects you can already open — it never widens what you can see.
- Go deeper when you need to. Research for a thorough answer, Plan changes to turn the answer into proposed edits, Skills to run a playbook with
/, and@to pull in a task, doc or project. - Pick the model, or speak. Choose the model per question, or Dictate instead of typing.
- Keep your threads. Your conversations stay under “Your threads”.
- See the work AI can take. Ask groups your open tasks by the way the router would classify them — a scoped query and a set of rules, with no model called to draw it. Tasks with no obvious shape yet are not offered to an agent; give one a brief and it joins the work.
- Hand it to an agent. The Agents tab lists each agent with its autonomy level, its skill and what it writes. “Every write an agent proposes waits for a person.”
AI Agents
Give each agent exactly as much freedom as you choose.
Skills, allowed actions, one autonomy level. Only the model is external.
The autonomy ladder
- L0Answers and suggests
Reads and answers. Changes wait in the AI Inbox.
- L1Suggests changes
Proposes every change. Nothing is written until approved.
- L2Acts, you approve the rest
Low-risk, reversible changes run. The rest waits.
- L3Acts, also on a schedule
Like L2, and may also run on a schedule.
Low risk: a comment, a nudge, a status change within a list, or a report. A project can be stricter than the agent.
Never available to any agent
These are not switches. They do not exist as actions, so even a stolen token cannot reach them.
Pick an action to try it.
Skill library
Every action an agent can take — and the ones it never can.
Versioned playbooks. Dry-run, edit or retire each one.
Reporter
digest.ceoYoursRisk: lowReads the project’s open work and posts a short digest: what moved, what is stuck, what is at risk.
Needs any task in the project it should report on.Project Guide
project.guideYoursRisk: lowAnswers “what next” from the project’s guide and plan, then proposes the next tasks.
Needs any task in the project it should report on.Check a page for defects
qa-reviewBuilt inRisk: lowChecks the page a task points to for defects.
Needs a public URL in the task title or description.Summarise a pull request
pr.summaryBuilt inRisk: lowReads the pull request a task links to, summarises the change and flags risk.
Needs a pull-request or branch link. It reads outside your workspace, so it stays off until the person who runs the server turns external reads on.Fill a task’s AI fields
fields.fillBuilt inRisk: mediumPicks the task’s AI fields that are empty or out of date and fills each with its own prompt.
Needs a brief of at least 40 characters in the description.Draft a PRD from a brief
prd.draftBuilt inRisk: mediumReads a task brief and drafts a product requirements document as a page that stays a draft until approved.
Needs a brief of at least 40 characters in the description.Find pages that have gone stale
wiki.upkeepBuilt inRisk: lowLists the project’s pages that are past their review date or months old, and suggests what each needs.
Needs any task in the project it should report on.Your own skills
Create, version and retire your own skills.
Every skill has a Dry run.Available actions Every action a skill may emit, and what the guard does with it
0Runs on its own0Waits in the AI InboxAlso runs on a schedule
Shown from the autonomy rules above. Reads always run. A project can be stricter than the agent.
Your AI memory
Tell it once how you like to work.
What Ask and your agents should know about you. Private to you.
WHAT TO CALL YOUMaya · Operations lead
PREFERENCES
Short answersUK spellingBullet points
THINGS TO REMEMBER · 3 OF 50
- Our fiscal year starts in April
- Client updates go out on Fridays
- Use the Operations project for suppliers
Use my memoryForget everything
- Private by designOnly you can see it. Admins included.
- Up to 50 things to rememberShort facts about your work.
- Bring memories from another AIFrom ChatGPT, Gemini or Claude. Passwords, keys and card numbers are removed first.
- Switch it off or forget“Use my memory” and “Forget everything”.
AI project generator
Describe the goal. Review a full plan.
Give it a brief. Review the plan before anything is built.
Write with AI
Plan a two-week sprint to launch our new pricing page — design, build, QA and a launch checklist.
1 · Brief2 · Questions3 · Plan
- Tasks, subtasks, checklists, descriptions and sprint plans from plain language
- Starter automations start switched off, and a workspace can set a daily AI budget
- Runs on your own LLM API key — multiple providers supported
AI project generator 1 list · 4 tasks · 12 sub-tasks
ListPricing page launchDue Nov 12
Design the pricing layoutHighMC
- Audit the current plan comparison table
- Sketch three layout directions
- Review the chosen direction with the team
Build the pageHighDP
- Build the plan card component
- Add the plan comparison table
- Add FAQ accordion
QA across devicesMediumPS
- Test mobile and tablet breakpoints
- Verify checkout and tracking events
Launch checklistChecklistSL
- Update sitemap and redirects
- Announce to the team
- Monitor the first 48 hours
You review it before it is built
Some features are switched on by whoever runs the server: team packs, workflows, app connections, connecting your AI by address, the extra MCP tool tiers and skills that read outside your workspace. If the demo shows one as switched off, that is why.
GET STARTED
Put AI agents to work. You approve.
$ curl -O https://raw.githubusercontent.com/aliansoftwareteam/AlianHub-Project-Management-System/main/docker-compose.yml$ echo "JWT_SECRET=$(openssl rand -hex 32)" > .env$ docker compose up -d# AlianHub running at http://localhost:4000$ npm run setup✓ Opens localhost:4000/#/setup→ demo.alianhub.com✓ Demo workspace, nothing to installOpen source. Self-hosted. No per-seat pricing. Or open the live demo first.