Viktor
Viktor is an AI employee that works inside Slack and Microsoft Teams, connecting to business tools to run tasks end to end and deliver finished reports, dashboards, web apps and code for entire teams.
What is Viktor?
Viktor is an AI agent that its publisher positions as an AI employee rather than a conversational assistant. Its only conversation surface is a team channel: you mention @Viktor in Slack or Microsoft Teams, or send it a direct message, and it takes the work from there. The publisher puts it plainly: "Viktor is an AI employee that lives in Slack and Microsoft Teams. He has his own computer in the cloud where he writes and runs code to complete tasks. He's not a chatbot. He's a colleague that does real work." That cloud machine is what separates it from text generators. Viktor writes and runs code in an isolated per-workspace environment, then returns finished deliverables: decks, PDF and XLSX files, dashboards, and internal web applications complete with authentication and a database, deployed on a live link. The marketing pages claim 3,200+ connectable tools, while the developer documentation mentions 9,000+; 48 are named and verifiable in the structured data of the integrations page, among them Slack, HubSpot, Stripe, GitHub, Linear, Jira, Notion, Google Sheets, Shopify, QuickBooks, Xero, Meta Ads, Google Ads and Databricks. Where no integration exists, Viktor can write its own from documentation you supply. Work does not have to be requested every time. Tasks can be scheduled, from recurring reports to overnight checks, and the agent keeps a persistent company memory of decisions and preferences, stored as internal notes it calls skills; a single screen recording is presented as enough to teach it a task. Governance is explicit: Viktor sees a channel only once invited into it, irreversible actions raise an approval button, and permissions are set per person. It can pull roughly 90 days of channel history when it arrives. Inference runs on Anthropic Claude, OpenAI and Google Gemini, with Together AI, Cloudflare Workers AI, fal.ai, xAI, Fireworks AI, ElevenLabs and Groq used for specific functions; each workspace picks its default model and can switch in one click. Developers get three surfaces: a Public API, a streamable HTTP MCP server, and an OpenAI- and Anthropic-compatible API. The publisher is Zeta AI, Inc., a Delaware company whose parent, Zeta Labs, was previously known for the Jace AI agent. The site displays a G2 rating of 4.9.
What it does
- Execute a task described in plain language from a Slack or Microsoft Teams channel
- Produce finished deliverables and post them in the channel: reports, dashboards, decks, PDF and XLSX files
- Build and deploy an internal web application on a live link, with authentication and a database
- Connect to and act inside existing business tools, from CRM and Drive to calendars, analytics and ad accounts
- Run scheduled automations such as Monday morning reports, overnight checks and fixes delivered before the office opens
- Reconcile data across several tools, audit advertising spend and shift budget where the numbers justify it
- Write code, review pull requests and request human approval before any sensitive or irreversible action
When to use Viktor / When not to
A quick filter to help you decide if Viktor is the right fit.
When to use Viktor
- Teams that already live in Slack or Microsoft Teams every day, since these are the only conversation surfaces the product offers
- SMBs and scale-ups that want to absorb more work without hiring, with a single shared agent serving ops, finance, marketing, sales, product and engineering
- Operations and finance teams juggling many SaaS tools that need reconciling, from CRM and ad accounts to billing, spreadsheets and analytics
- Agencies looking to lower the cost of servicing each client account, which the publisher addresses on a dedicated agencies page
- Registered nonprofits, who receive a permanent 15% discount on any monthly plan after a status check completed within one business day
When not to use Viktor
- Teams that work neither in Slack nor in Microsoft Teams: there is no other way to talk to Viktor
- Individuals looking for a personal assistant, since billing is per workspace rather than per seat and the product is designed as a shared team resource
- Anyone needing individual confidentiality: integrations are shared across the whole team and a private mode is announced as roadmap, not delivered
- Organisations that require EU data residency, configurable retention or fine-grained RBAC today: the first is contradicted between two pages of the site, the second is stated as unavailable, the third is described as still being built
- Buyers who need a completed SOC 2 Type 2 or ISO 27001 certificate, and anyone under 18, for whom the service is not intended
How to use Viktor
A typical end-to-end flow, from setup to results.
- Install Viktor from the Slack App Directory or the Microsoft Teams catalogue
- Sign in with your existing Slack or Teams identity, as there is no separate Viktor account or password
- Let the agent run its own onboarding: it introduces itself and asks what the team needs
- Connect your tools through OAuth, keeping in mind that each connection grants the right to act, not only to read
- Invite Viktor into the channels it should work in, remembering that it reads everything posted there
- Mention @Viktor in a channel or send it a direct message to hand over a first task
- Review the approval prompts it raises before sensitive actions such as emails, code pushes, deployments or payments
- Set the administration rules: which integrations are available, who may use them, which action types require approval
- Teach recurring work by recording your screen once so Viktor can reproduce the task
- Put a proven task on a schedule so it runs on its own, and track credit consumption in the web app
Pros & Cons
Pros
- Delivers finished artefacts, from files to deployed applications, instead of text that still has to be turned into work
- No new interface to learn: everything happens in the team channel staff already use, with installation claimed in minutes and no IT project
- Workspace-based pricing rather than per seat, with all features and integrations included in every plan so only the credit volume changes
- $100 of credits to start without a credit card, unused monthly credits roll over for one month, and trial and bonus credits never expire
- Serious guardrails: human approval required on irreversible actions with conservative defaults, and secrets kept out of the model by an execution gateway
- Free and reversible model choice per workspace, which avoids lock-in to a single AI provider
- Documented data and security posture: contractual no-training commitment at the publisher and its model providers, a public itemised sub-processor list, SOC 2 Type 1, CASA Tier 3, a Slack security review and Microsoft validation
Cons
- The detailed pricing grid is rendered in JavaScript and absent from the static HTML, so only the $50 per month entry point is publicly verifiable
- No permanent free plan: the $100 of credits is a trial, and cost is variable by design, from 100-300 credits for a quick task to 2,000-5,000 for a full project
- Integrations are shared with the whole team, so connecting a personal mailbox exposes it to colleagues, and there is no per-user isolation today
- Private mode and fine-grained RBAC are presented as roadmap rather than delivered features
- Configurable retention policies are explicitly stated as unavailable today
- EU data residency is uncertain: the security page presents it as available on Enterprise contracts while the enterprise page places it on the roadmap
- Integration counts are inconsistent between marketing (3,200+) and documentation (9,000+), SOC 2 Type 2 and ISO 27001 are still in progress, and the publisher acknowledges Viktor can get things wrong
Pricing & Plans
There is no permanent free plan. A new workspace receives $100 of credits with no credit card required, and once those credits are spent a paid plan is needed. The lowest published price point is 50.00 USD per month, billed for the entire workspace rather than per seat. Every plan includes all features and all integrations; only the credit allowance differs. Credits reflect real model usage, and caching is said to reduce the cost of repeated workflows: the publisher indicates 100-300 credits for a quick task, 500-1,500 for a complex workflow and 2,000-5,000 for a full project, and describes a 20,000-credit tier as roughly 40 to 200 tasks. Unused monthly credits roll over for one month, while trial and bonus credits carry no expiry, and plans can be changed upwards or downwards at any time. Registered nonprofits receive a permanent 15% discount on any monthly plan. The complete tier grid is rendered in JavaScript and does not appear in the static HTML, so the intermediate price points could not be verified.
- $100 of credits
- no credit card required
- with all features and integrations included
- from $50 per month for the entire workspace
- with no per-seat billing
- SAML SSO
- RBAC
- audit logs
- a ready-to-sign DPA and a dedicated channel with the Viktor team
- priced on request through the contact sales page
- The commercial names of the intermediate credit tiers are not exposed in the site's static HTML and could not be verified
Data, GDPR & hosting
A consolidated view of how Viktor handles your data.
GDPR overview
GDPR implementation is documented rather than merely claimed. The security page says "GDPR aligned" while the sales page says "GDPR & CCPA compliant". Legal bases cited for EEA and UK users are contract performance, consent (OAuth installation, non-essential cookies) and legitimate interest (security, fraud prevention, reliability). Transfers outside the EU rely on Standard Contractual Clauses embedded in a publicly available DPA. The Article 27 EU representative is named in Annex I: Paweł Siuciak, Zeta AI, Inc., ul. Rondo Daszyńskiego 1, 00-843 Warsaw, Poland, with the Irish Data Protection Commission as the designated supervisory authority. A UK representative will be appointed if the law requires it, which has not yet been done. Access, rectification, erasure, portability, objection, restriction and complaint rights are exercised through support@viktor.com, with a typical 45-day response and a "Privacy Appeal" escalation. No DPO is named. Policy last updated 31 July 2026.
Who owns the data?
Zeta AI, Inc., a Delaware corporation trading under the commercial name Viktor, operates the service. Customer Data is defined as everything submitted to the service on the customer's behalf: OAuth tokens, workspace and user identifiers, files, conversations and outputs, scheduled tasks, approval decisions and logs. Under the published Data Processing Agreement, Zeta AI acts as processor or sub-processor while the customer remains controller or processor and data exporter, so contractual control over the content stays with the customer. The publisher states it does not sell personal data for money. Customers may export data before deletion where the law allows, request a "Clean Workspace" erasure, or have production data deleted within roughly 30 days of a verified request.
Reuse rights
Customer Data is used to run the service: authenticating users, maintaining integrations, executing tasks and generating outputs. Relevant portions are sent to model providers to produce answers, and the publisher states that neither it nor those providers train foundation models on that data: "We do not train our own or third-party foundation models on Customer Data, and we do not permit AI providers to do so." Model providers are said to operate under no-training, zero-retention agreements for Viktor traffic, Slack and Shopify API data is explicitly excluded from any generalised model training, and Slack message content is never used for advertising. One exception is documented: narrow, service-specific internal models such as the model-selection router may be built from Customer Data, but they generate no content and are never supplied to third parties. Aggregated or de-identified data is used to understand usage and improve reliability. Separately, identifiers and usage events are shared with analytics, ad-measurement and attribution partners (PostHog, Google, Meta, TikTok, Reddit, X, LinkedIn, Dub, Rewardful), which the policy notes may amount to a "sale" or a "share" in some jurisdictions. Conversations and outputs are part of Customer Data and can be exported before deletion; the site sets out no separate restriction on the customer's own reuse of what Viktor produces.
Data retention & training
Hosting summary
Viktor's data is hosted in the United States. The publisher states that production infrastructure runs on AWS us-east-1, with customer data stored in US regions across named providers: Modal for the main application compute, sandboxed execution and file volumes; PlanetScale for the production Postgres database holding customer data; Cloudflare for CDN, DDoS protection and R2 object storage; and Vercel for frontend hosting. Data is encrypted in transit with TLS 1.2+ and 1.3 and at rest with AES-256 using cloud-provider key management. Access controls include role-based permissions, MFA, least privilege, logging and monitoring, and the public sub-processor list details around thirty entries with the nature of the data each receives. Model providers process data in the United States or in the regions set out in their enterprise terms. One point deserves clarification before signing: the site contradicts itself on EU residency. The security page states "Hosted on AWS us-east-1. EU data residency is available on Enterprise contracts", while the enterprise page states "EU data residency is on the roadmap - ask us for the current timeline".
Things to keep in mind
Risks and trade-offs to weigh before adopting Viktor.
- Broad access by default: Viktor reads every message in the channels it belongs to, not only those that mention it, and can pull roughly 90 days of channel history when it joins
- If a user grants Slack search scopes, Viktor can read the private channels and direct messages that this user can access
- Integrations sit at workspace level, so every member inherits the connections another member has set up; the publisher itself recommends "treating Viktor like a shared team resource and avoiding sharing sensitive personal information in direct messages"
- Insights derived from an integration can survive its disconnection, and a "Clean Workspace" request is needed for a complete erasure
- Identifiers and usage events are shared with analytics and advertising partners, which may qualify as a sale or a share of personal data depending on the jurisdiction
- Prompt injection is a genuinely new attack surface; the publisher documents it and relies on human approval to contain it, but approval fatigue in a busy channel would quietly defeat that safeguard
- Over-reliance and accountability: the publisher acknowledges Viktor can be wrong, a team that stops reviewing its output inherits errors at machine speed, and the contracting entity is a Delaware company at a private mailbox address rather than an operating headquarters
Setup & Integrations
Technical difficulty
Low for end users, light for administrators. Viktor is installed from the Slack App Directory or the Microsoft Teams catalogue, and the publisher claims it is live in two minutes, with no deployment project or IT ticket. There is no separate Viktor account or password, since identity is inherited from Slack or Teams. Tools are connected through OAuth with the narrowest scopes, and the agent runs its own onboarding. Day-to-day use needs no technical skill. Administrators also choose integrations, users and approval rules; SAML SSO applies only to Enterprise, and the API, MCP server and compatible endpoints remain optional.
Deployment
Integrations
Supported languages
Behind Viktor
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Viktor.
Frequently asked questions
Where do you actually use Viktor?
How is it different from a chatbot?
How many tools does it connect to?
How much does Viktor cost?
Is there a permanent free plan?
Is my data used to train AI models?
Where is the data hosted?
Which messages can Viktor read?
Can Viktor act without validation?
Which certifications does the vendor hold?
Should you pick Viktor?
Viktor occupies a distinctive position: an agent that ships finished artefacts rather than a conversational assistant that explains what to do. Its two strongest arguments are practical. There is no new interface to learn, since work is requested in the Slack or Teams channel a team already uses, and pricing is per workspace rather than per seat, with every feature and integration included in each plan so that only the credit volume changes. The guardrails are more serious than the category average: human approval on irreversible actions with conservative defaults, secrets injected at runtime by a gateway so they never reach the model, per-workspace isolation, a contractual no-training commitment covering the publisher and its model providers, and a public, itemised sub-processor list. The reservations are just as concrete. The access perimeter is broad by default, since Viktor reads everything in the channels it belongs to, integrations are shared across the team, and per-user isolation does not exist today. The detailed pricing grid is not public in the static HTML and cost is variable by nature, so a budget can only be estimated from the credit ranges the publisher gives. EU data residency and configurable retention are not settled: the site contradicts itself on the first and states the second is unavailable. Maturity should be weighed too. Zeta Labs was founded in 2023, raised $75M in a Series A in May 2026 and holds SOC 2 Type 1, with Type 2 and ISO 27001 still in progress: a fast-moving young vendor with real funding rather than an established platform. Before buying, get three points in writing: whether EU residency is genuinely available on your contract, what the per-user isolation roadmap actually delivers, and which credit tier matches your intended workload.
- Choosing a selection results in a full page refresh.
- Opens in a new window.