VaulticWorkspace, access & tokens
Workspace membership and roles, per-project access grants, and service tokens.
Workspace: vaultic workspace
Workspace-wide membership and roles.
workspace invite
workspace invite <email>| Flag | Description |
|---|---|
-r, --role <role> | admin | developer | read-only (default developer) |
Prints an accept-invite link built from the server's own configured web app URL (the same
webAppUrl the browser-login flow uses, from /auth/providers — so it already includes any path
prefix the server is served under, e.g. the hosted instance's /app), or VAULTIC_WEB_URL if
set, or http://localhost:5173 as a last resort.
workspace invites
Lists pending invites. No options.
workspace revoke-invite
workspace revoke-invite <inviteId>No options.
workspace members
Lists workspace members. No options.
workspace set-role
workspace set-role <email> <role>Positional role: owner | admin | developer | read-only. No flags.
workspace remove-member
workspace remove-member <email>No options.
workspace accept-invite
workspace accept-invite <token>Redeems an invite link as the currently logged-in user. No options.
Access grants: vaultic access
Per-project/per-environment access, layered on top of workspace-wide roles — for giving someone access to one project (or even one environment within it) without making them a full workspace member-level role everywhere.
access grant
access grant <email>| Flag | Description |
|---|---|
-r, --role <role> | developer | read-only (default developer) |
-e, --env <environment> | Scope the grant to a single environment (default: whole project) |
access list
Lists access grants for the current project. No options.
access revoke
access revoke <grantId>No options.
Service tokens: vaultic tokens
Scoped tokens for CI/deploys — used by export/run without a human login. Service-token reads
never consult personal overrides.
tokens create
tokens create <name>| Flag | Description |
|---|---|
-e, --env <environment> | Scope to a single environment |
--read-only | Read-only token (default is read/write) |
--expires <days> | e.g. 90d |
Prints the token once — store it immediately, it isn't retrievable again.
tokens list
Shows id/name/access/status (revoked / expired / active). No options.
tokens revoke
tokens revoke <tokenId>No options.
Agents: vaultic agent
Registered AI/MCP agent identities — a standing scope ceiling plus short-lived, task-scoped sessions that attribute every action to both the agent and the human who authorised it. Every agent connected through the MCP Server's HTTP connector gets one of these automatically on approval; these commands are for managing them directly, or for registering one by hand with a specific scope. See Access, Team & Tokens for the concepts. Everything here requires ADMIN/OWNER.
agent create
agent create <name> --owner <email>| Flag | Description |
|---|---|
--owner <email> | Required — the human accountable for this agent |
-p, --project <slug> | Scope to a single project |
-e, --env <environment> | Scope to a single environment (requires --project) |
--write | Write access (default is read-only) |
--keys <patterns...> | Key allowlist globs, e.g. DEPLOY_* |
--max-session <duration> | Ceiling on a session's duration, e.g. 30m (default 1h) |
--max-actions <n> | Ceiling on a session's action count (default 500) |
agent list
Shows id/name/owner/scope/access/active session count/status. No options.
agent disable
agent disable <name>Immediately invalidates the agent's active sessions. No options.
agent sessions
agent sessions <name>Lists the agent's sessions — active and recent history. No options.
agent session start
agent session start <agent>| Flag | Description |
|---|---|
--purpose <text> | Shown in the sessions panel |
--duration <duration> | e.g. 15m — clamped to the agent's configured ceiling |
-p, --project <slug> | Narrow the agent's standing scope for this session |
-e, --env <environment> | Narrow the agent's standing scope for this session |
--keys <patterns...> | Narrow the agent's key allowlist for this session |
Prints the session token once — store it immediately, it isn't retrievable again.
agent session end
agent session end <agent> <sessionId>No options.
Access reviews: vaultic access-review
Periodic, evidence-producing reviews of every member, access grant, and service token in the
workspace — see Access reviews for the concepts and web UI. Workspace-scoped, same as tokens/access above.
access-review list
Lists campaigns with id, name, status, and reviewed/total progress. No options.
access-review create
access-review create| Flag | Description |
|---|---|
--name <name> | Required, e.g. "Q3 2026" |
--project <slug> | Narrow to a project — repeatable; omit for the whole workspace |
--due <date> | ISO date, e.g. 2026-09-30 |
--recurrence <recurrence> | quarterly | annually — recreates the campaign automatically once it completes |
Creates a DRAFT campaign — nothing is enumerated for review until start.
access-review start
access-review start <campaignId>Notifies every assigned reviewer (in-app + email) and moves the campaign to ACTIVE.
access-review items
access-review items <campaignId>| Flag | Description |
|---|---|
--mine | Only items assigned to you |
The --mine view is what makes CLI-based review practical for engineering-led teams who won't
open the dashboard.
access-review certify
access-review certify <campaignId> <itemId>| Flag | Description |
|---|---|
--note <note> | Optional |
Records the item as reviewed and still needed. No side effect on the underlying access.
access-review revoke
access-review revoke <campaignId> <itemId>| Flag | Description |
|---|---|
--note <note> | Optional |
Immediately removes the underlying access (deletes the grant, revokes the token, or removes the member) through the same code paths — and the same protections, including that a workspace can never be left without an owner — as managing access directly. If execution fails (e.g. that protection triggers), the decision is still recorded with the failure reason attached rather than silently succeeding.
access-review complete
access-review complete <campaignId>Force-completes an ACTIVE campaign. Items still pending stay pending in the evidence record
rather than being silently certified.
access-review export
access-review export <campaignId>| Flag | Description |
|---|---|
--format <format> | csv | json (default csv) |
Prints the campaign's evidence record — every item's snapshot, decision, reviewer, and execution result — to stdout.