AI assist
AI that runs on your account, not ours
Eleven assists across issues, projects and sprints. Every one of them uses your organization's own provider key, under a policy your administrators set, and produces a suggestion a person reviews before anything changes.
The eleven assists
This is the complete list. There is no twelfth.
On an issue
- Draft a descriptionContext, goal and constraints in Markdown, from the title and what is already there.
- Suggest acceptance criteriaA testable checklist under its own heading, appended to the description.
- Break into subtasksThree to eight concrete subtasks, created as child issues when you apply it.
- Summarise the threadWhat it is, where it stands, and what is blocking or next.
- Suggest a priorityA priority and a one-sentence rationale. Informational — it does not set the field.
On a project
- Draft a delivery planPhases, the main work in each, and rough sequencing.
- Propose milestonesThree to six milestones, each with its outcome.
- Identify risksDelivery risks with an impact level and a mitigation for each.
On a sprint
- Plan the sprintWhat looks achievable, what to defer, and any sequencing concerns.
- Summarise progressA stakeholder update: what shipped, what slipped, and the headline.
- Surface blockersLikely blockers and stalled work, with the concern.
AI assist
Your provider. Your key. Your models.
An administrator connects OpenAI, Anthropic or Google Gemini with the organization's own API key. There is no shared Vorkhurry key, and the key never leaves the server — the API returns a masked suffix only.
An organization administrator
AI is off until someone turns it on
Connects a provider
OpenAI, Anthropic or Google Gemini
With your own API key
Encrypted at rest; the API returns a masked suffix only
Selects models and a budget
Allow-list, a default, per-feature overrides, a monthly cap
AI assist across Vorkhurry
Eleven actions on issues, projects and sprints
There is no shared Vorkhurry key
Model usage runs on your organization's own provider account and is billed by that provider. Vorkhurry does not resell, subsidise or proxy anyone's model usage.
Models, allow-lists and budget
Choose which models are enabled, set a default and per-feature overrides, and cap monthly spend. Every call is recorded, so usage is a number you can look at.
Suggestions are reviewed, not applied
Output arrives as a suggestion you accept or reject. Applying it is a second, deliberate step — nothing a model wrote edits your issues on its own.
Supported providers
Connect one. The key is encrypted at rest and never returned by the API — reading it back gives you a masked suffix.
OpenAI
Use your organization's OpenAI account for AI assist.
Anthropic
Use your organization's Anthropic account for AI assist.
Google Gemini
Use your organization's Google AI account for AI assist.
What Vorkhurry does not do
- It does not pay for your model usage. There is no shared Vorkhurry key — your provider bills your organization directly.
- It does not train anything on your data. Vorkhurry has no model of its own; it calls the provider you connected.
- It does not act on its own. Every assist produces a suggestion; applying it is a separate action a person takes.
- It does not turn itself on. A workspace that never opens the AI settings never sends anything to a model.
Connect a provider when you are ready
AI assist is one setting away, and off until an administrator turns it on.