Agents & collaboration AI coworkers Supported: Specialized coworkers with their own responsibilities and tools.
Supported: Shared Team Bots built around a role, each with context, plugins, credentials, and memories; public beta on Teams and Enterprise plans.
Supported: Reusable agents with skills and tool access.
Supported: An AI coworker that works across business tools.
Team collaboration Supported: Shared conversations and Slack threads with people and AI.
Supported: Everyone works with the same Bot and its expertise, while each person's conversations and memories stay private.
Supported: Shared agents with owner, editor, viewer, and use-only access.
Supported: Workspace members can work with Viktor in channels and DMs.
Communication channels Conditional: In-app chat, Slack, and email; other channels require setup.
Supported: Slack with a dedicated handle per Team Bot, alongside the Grok Bot app.
Supported: Web chat, Slack, and API access to shared agents.
Supported: Slack and Microsoft Teams.
Agent handoffs Supported: Delegate work to other AI coworkers.
Supported: Bots pass context between each other, hand off tasks, and can pass ownership so you are not the router between tools.
Supported: Agents can hand tasks to other agents.
Conditional: External agents can delegate work to Viktor through its MCP server and public API; internal multi-agent handoffs not documented.
Persistent memory & skills Supported: Persistent memory, company knowledge, and reusable skills.
Supported: Team Bots keep separate context and memories per person while drawing on skills shared across the team.
Supported: Reusable skills and playbooks that evolve with use.
Supported: Workspace-scoped skills and memory.
Automation & tools Schedules & triggers Supported: Schedules, events, and follow-ups with run history.
Supported: Routines run on a schedule or, where supported, after an event; background routines run while the laptop is closed.
Supported: Recurring or one-time scheduled runs plus event triggers from Gmail, Slack, Teams, and more.
Supported: Recurring tasks and proactive automations.
Browser automation Conditional: Hosted browser with persistent sessions; browser service required.
Supported: Every Bot runs on a persistent cloud computer with a browser, filesystem, and terminal, and can log in to services in that browser.
Supported: Interactive browser automation via the Website Scraper's Web Agent mode, an AI browsing node, and the Chrome extension.
Supported: Works the web directly: logs in, clicks through pages, fills forms, and pulls data from real sites.
Meeting intelligence Conditional: Supported meeting recording, transcripts, and follow-up; meeting service required.
Limited: Not offered: connectors cover Gmail and Outlook calendars, but no meeting recording or transcription is documented.
Conditional: No native recording or transcription; meeting work centers on Gumball pre-meeting briefs from a connected calendar.
Conditional: Meeting follow-through runs through connected meeting tools like Symbl.ai, Fireflies, or Gong; native recording and transcription not documented.
Integrations Supported: Prebuilt connectors and custom integrations.
Supported: Built-in connectors, a preconfigured connector catalog, and custom MCP servers.
Supported: Connectors, shared credentials, and MCP support.
Supported: 3,200+ business-tool integrations advertised.
Apps & infrastructure Custom app building Supported: Build apps with AI, inspect code, and iterate in a live preview.
Limited: Bots steer Cursor Projects and launch Cloud Agents for code work; no interactive app-building surface documented.
Limited: Not offered as a business-app builder: the platform ships agents, workflows, and hosted agent pages.
Supported: Build and deploy internal apps connected to business tools.
Internal publishing Supported: Private, individual, and organization-wide app sharing.
Limited: Not offered: Bot share links are public previews on x.ai; no internal publishing surface documented.
Supported: Hosted agent pages on gumloopagents.com with sign-in and role-based access.
Supported: Viktor Pages share finished documents and dashboards at links; Viktor Spaces are internal tools your team signs into.
Row-level permissions Conditional: Role-based access for owners, admins, editors, and members, with apps kept private, shared, or organization-wide.
Limited: Not offered: access follows connector permissions such as permission-aware Google Drive; no per-row data permissions documented.
Limited: Not offered: access control uses organization and team roles plus custom feature roles, not per-row data permissions.
Limited: Not offered: enterprise access is role-based (RBAC); no per-row data permissions documented.
Private resources & SSH Conditional: SSH tunnels and custom connections for private resources.
Limited: Custom MCP servers must be reachable over the public internet; local servers require a tunneling service.
Limited: Not offered: tools connect through cloud connectors, MCP, and webhooks; no private-network or SSH access documented.
Limited: Not offered: tools connect through cloud integrations and per-user OAuth; no private-network or SSH access documented.
Models & cost control Model choice Supported: 29 models across 11 providers.
Conditional: Runs Grok models; no cross-provider model catalog documented for Bots.
Supported: 35+ models advertised.
Supported: Choice across OpenAI, Anthropic, and Google.
Automatic model routing Supported: Auto routes tasks to Cheap, Balanced, or Smart tiers.
Limited: Not offered: no automatic task-to-model routing documented.
Supported: Auto picks the model for each agent message, with mid-run escalation and fallbacks.
Conditional: Workspace model and task-level cost presets; automatic category routing not verified.
Custom routing rules Supported: Choose tier models, map task categories, or pin a model.
Limited: Not offered: no task-category-to-model routing rules documented.
Conditional: Admins restrict models, pin per-agent presets, and set fallbacks; task-category mapping not offered.
Conditional: Workspace defaults and individual scheduled-task settings.
Separate coding policies Supported: Separate conversation and coding routing policies.
Limited: Not offered: Cursor team rules and privacy settings apply to Bots; no separate coding model policy documented.
Limited: Not offered: AI model governance restricts and pins models platform-wide; no separate coding policy documented.
Limited: Not offered: model governance is workspace-wide with per-task presets; no separate coding policy documented.
Bring your own keys Conditional: Bring your own provider keys as part of an enterprise setup.
Limited: Credentials connect third-party tool APIs; no customer AI provider keys documented.
Conditional: BYOK on Pro and above; organization-wide keys on Enterprise.
Limited: Not offered: model providers run under Viktor-managed agreements; tool credentials are per-user OAuth handled by the Viktor gateway.
Bring your own gateway Supported: OpenRouter and Vercel AI Gateway.
Limited: Not offered: no bring-your-own AI gateway documented.
Conditional: Provider proxy URL, custom headers, and model name mapping per provider on Enterprise.
Limited: Not offered: models run through the Viktor-managed stack; no bring-your-own gateway documented.
Security & governance Single sign-on Conditional: SAML single sign-on for enterprise.
Conditional: Members sign in with their Cursor account, so existing Cursor SSO and team membership apply.
Conditional: SAML SSO and SCIM for enterprise identity.
Supported: SAML SSO via Okta, Entra ID, Google Workspace, OneLogin, or any SAML 2.0 IdP.
Audit & activity visibility Conditional: Access, agent activity, and tool usage for enterprise review.
Conditional: Advanced audit controls ship on Enterprise plans; Grok Bot-specific audit coverage not documented.
Supported: Agent activity, tool-call tracking, and analytics.
Conditional: Enterprise audit logging; confirm required event coverage.
Human approvals Conditional: Action-specific permissions and contact controls on every channel; explicit approval gates not documented.
Supported: Require Approval rules and Auto Review evaluate tool calls and computer actions; passwords and 2FA use human takeover.
Supported: App guardrails, spending approvals, and scoped policies.
Supported: Sensitive actions can wait for explicit approval in chat.
Usage visibility & budgets Supported: Usage visibility and per-agent credit budgets.
Conditional: Pooled billing with weekly usage allowances on premium seats; per-Bot budget controls not documented.
Supported: Spend caps by agent, team, or organization.
Conditional: Auto-topups with monthly spend caps; per-user spend caps on Enterprise.
Pricing & rollout Billing model Supported: Free plan with $10 in credits every month; paid plans from $20/month. $1 = 100 credits; top-ups available.
Conditional: Individual access via SuperGrok Heavy or Cursor Ultra; self-serve teams use premium seats with usage allowances; Enterprise is rolling out.
Conditional: Pro starts at $37/month; credits and an orchestration fee apply.
Supported: Credit-based plans from $50/month, with trial credits.
Seat pricing Supported: Unlimited builders and viewers in your organization.
Conditional: Team seat management and consolidated billing on paid plans.
Supported: Unlimited seats listed on Pro and Enterprise.
Supported: No per-seat licenses; credits reflect model usage.