Skip to content
Browse documentation

Reference

FAQ and glossary

Quick answers and common Vopit terminology.

Frequently asked questions

QuestionAnswer
Does Vopit provision servers?No. The built-in connectors access existing POSIX SSH/SFTP hosts, PostgreSQL and MySQL/MariaDB targets. Cloud or cluster tooling needs an approved runner; native Windows SSH is not supported.
Can I connect directly to claude.ai?No. Vopit has no direct claude.ai account connection; use the Claude API connection or the Claude Code CLI.
Does a provider catalogue prove that agents can run?No. Availability, entitlement, billing, limits, and a real response are separate checks.
Are desktop installers always available?No. Only files listed on Downloads are available.
Does Vopit update itself?On Windows x64 installations from vopit.fr and the Linux x64 package, Settings → Updates checks for a new version automatically (on by default, at startup and every six hours) and offers Download update. You run the downloaded installer, or install the downloaded package, when you are ready; Vopit never installs it on its own. Installations from the Microsoft Store are updated by the Store; Vopit tells you when a new version is published there and opens its Store page. See App updates and corporate networks.
Can work run on a schedule?Yes, with Cranks: a Clock trigger repeats at an interval and a Daily schedule trigger starts at a local time on selected weekdays. They run only while Vopit is open on that project with an active license and no global pause. Create a Crank Flow, on the Cranks canvas, proposes a whole set from a description: a trigger and the actions it runs, already wired, created disabled for your review.
Can I import a request session?No. Export session writes a JSON record for review and archiving; there is no import or restore interface for it.
Does rollback restore database data?Only if a separately reviewed operation explicitly restores it. Vopit runs the rollback action configured for your project; reverting application code does not automatically revert data.

Glossary

TermDefinition
AgentAn editable worker definition with a mission, capabilities, model, behavior, and output contract.
Stage / nodeA unit of workflow work. An edge determines when another node can run.
HandoffRecorded transfer of a ticket or result between stages.
Provider / modelThe configured inference connection and its selected model.
PermissionA category policy: READ, WRITE, EXECUTE, NETWORK, GIT, GIT_WRITE, GIT_REMOTE, or GIT_HISTORY.
Infrastructure grantAn independent assignment of a saved server connection to an agent.
Recipe / operationA saved SSH command/upload or database SQL definition with policy and timeout.
CheckpointSaved progress that can allow Resume to avoid repeating completed work.
BoardThe project's card board, organized in columns. Agents with board actions can read, add, move and rewrite cards; only you or auto-run decide which cards run.
CrankAn application connection, trigger or action on the Cranks canvas. Connected Cranks form chains that start missions or run actions.
MeetingA conversation with selected agents on the Meetings canvas. Its outcomes, such as facts, instruction changes or actions, change nothing until you apply them.
Speed modeThe team-wide choice between Ultra-fast, Balanced and Big brain (default), trading planning and review depth for speed.
Fast modelThe model a connection uses for the quick stages of the Ultra-fast and Balanced speed modes.
Open pointsBlockers handed off by a stage that no later stage resolved. Fix open points sends only those to a new request.
Linked folderAnother folder linked to the workspace, such as a backend next to a frontend. Agents see it under its name; unlinking does not touch its files.
WorktreeA separate Git checkout in .vopit/worktrees, on its own branch, where a parallel request works without sharing the project folder.

All documentation