FAQ
Frequently asked questions
Short answers, each grounded in what the product actually does. Ask us anything else on the contact page.
The basics
Software in which AI agents perform project-management work themselves, such as planning tasks, assigning them, reviewing code and reporting status, rather than merely answering questions about a board maintained by people. Conductor's agents do this work and pause for a human decision at the points that matter.
It replaces the parts that depend on people typing status into a tracker. Tasks, assignments and completion live in Conductor and are grounded in GitHub; developers keep working in Git and their pull requests as usual.
Three: admin, executive and developer. Admins resolve decisions and configure the tenant, executives read projects, people and mail without resolving anything, and developers see only their own tasks. Every rule is enforced by the backend, not the console.
Request a quote or a pilot. OriginSphere Technologies provisions your isolated tenant on our infrastructure or guides the install on yours, configures GitHub, email and the model, and loads your first playbooks. The console's first-run checklist then walks you from model configured to first plan approved.
How it works
A fixed rule, not a model: a linked pull request must be merged into the default branch, the latest code review must cover that merged commit, the review must not have requested changes or found a high-severity issue, and the acceptance criteria must be marked met. Otherwise the task stays open and the blocking reasons are recorded.
A pull request event from GitHub: opened, updated, reopened, marked ready or closed. The review agent reads the real diff, the linked task's acceptance criteria and the active playbooks for the stacks the files touch. A review is keyed by pull request and head commit, so the same commit is never reviewed twice.
By default review feedback and risk alerts pause in the Decisions queue, while the weekly executive digest sends automatically. The gated kinds are a setting, so a tenant can gate everything or nothing. Every send is audited either way.
Deterministically, every day, with no model in the loop: overdue tasks, stale in-progress tasks, blocked tasks, developers over capacity, assigned tasks with no commits, long-open pull requests and milestones at risk of their target date. Only the alert email is written by a model, and repeat findings are latched for a configurable cooldown.
Your own coding standard for one technology stack, distilled by the Recon agent from a document you upload into a structured rule set. Once an admin activates it, code review retrieves it for every matching pull request and may request changes on a clear violation. Playbooks are retrieval, not fine-tuning.
Yes. After the assignments are approved, the provisioning step creates one repository per approved spec under the configured account, invites the assigned developers and installs the webhook. It is best-effort and every outcome is audited; an existing repository is adopted rather than recreated.
You file a change order with the new text or documents. The change planner produces a delta of added, modified and cancelled tasks against the existing plan, the delta waits for an admin in the Decisions queue, and only then is the backlog updated. Nothing is deleted; cancelled tasks stay auditable.
Deployment and models
Any OpenAI-compatible endpoint. Tenants typically run a local model through LM Studio so prompts and code never leave their network, or use OpenAI by changing one setting. The provider is configured, never hard-coded.
No. Each subscription is an isolated single-tenant deployment with its own database, encryption keys, credentials and model endpoint.
Yes. The stack is three Docker containers behind a reverse proxy. OriginSphere Technologies provides the images, the install, and updates under the subscription.
GitHub (webhooks and polling), SMTP and IMAP email, Microsoft Graph mail, Microsoft Entra ID single sign-on, and LM Studio or OpenAI for inference.
Yes. One Entra ID app registration gives you single sign-on, directory import of users, and Microsoft Graph mail for sending and reading email instead of SMTP and IMAP. Conductor mints its own session tokens and never stores Microsoft's.
Data and cost
Profiles and emails are treated as sensitive: they are used to assist and inform, never to surveil. Access is role-based on the backend, developers see only their own data, secrets are encrypted at rest, and every autonomous action is auditable and overridable.
The AI usage page shows token usage per agent, flow and project, captured by one LangChain callback on every run. A row is written only when the provider reports real usage, and no prompt or response text is ever stored.
Still have a question?
Ask it on the contact page and a person replies within one business day.