Genspark AI Workspace vs Viktor (2026)

Genspark's agents write into systems it does not own. Its Google Workspace add-on, listed on the Google Workspace Marketplace by Mainfunc, works inside Google Docs, Sheets and Slides, reading what the user is working on, including any selected text, cells or slide, and turning plain language requests into edits, formulas, analyses, slides and on brand visuals inside the open file.

It inserts generated images into documents, writes VLOOKUP, XLOOKUP, ARRAYFORMULA, QUERY and Apps Script formulas into cells, creates native Sheets charts placed without overlapping existing ones, applies conditional formatting and number formats, deduplicates and reshapes ranges, and generates summary tabs. The Genspark AI Workspace app lists Super Agent, AI Slides, AI Sheets, AI Docs, AI Developer with full GitHub integration for code editing and pull requests, and AI Meeting Notes, so it can also open pull requests in GitHub. There is no full integrations directory.

Sourceworkspace.google.com/marketplace/app/genspark/592480783716read 2026-09-12

Viktor connects to 3,200+ tools, mostly by narrowly scoped OAuth, and does real work in them. Customers connect tools from the CRM to the ad accounts, such as Google Ads, Google Drive, Stripe, HubSpot, Notion, Linear, Asana, ClickUp, Cursor and GitHub, and Viktor works in them and posts finished work back, with per-integration pages for HubSpot, Salesforce, Jira, QuickBooks, Meta Ads and others.

Every major tool connects via OAuth with the narrowest scopes that get the job done, and API keys sit in an AES-256 vault, injected at execution time by a backend tool gateway, access-logged and rotatable. Admins decide which integrations are connected and who can use them. Actions are real, such as syncing MQLs, drafting invoices and building apps, and credentials are narrowly scoped, rotatable and revocable in one click. The docs home says 9,000+ integrations while the product page says 3,200+.

Sourceviktor.com/product, /security; read; viktor.com/securityread 2026-09-16

Super Agent runs multi step work with several agents. It breaks a request into steps, chooses the right tools and carries the work through to a finished result, coordinating the right agents and tools automatically. Several tasks can run at once in a single project, each building on the others' results.

Super Agent is built for long running tasks, reasoning through every step without losing the thread, and a user can start a task, close the laptop and come back to finished work, so runs continue on Genspark's servers. The earlier generation handled multi step tasks mostly in sequence. No BPMN or named runtime is published.

Sourcegenspark.ai/helpcenter/super-agentread 2026-09-12

Multi step jobs across several tools run to completion, mixing fixed tool operations such as pull, post and draft with the agent's own judgment. Viktor sorts overnight support tickets, answering where he can while the ugly ones are escalated by name with the customer's history attached, so each ticket branches to an answer or a named escalation.

He pulls every open deal from the CRM and works out what moved and which deals deserve a call, and builds an internal order tracker on Shopify and Stripe that he keeps running and uses to flag late orders into a channel. Another job pulls overdue invoices in QuickBooks, drafts reminders for the oldest, posts a spreadsheet and holds the send. Retry and fallback configuration and versioning are not described, and there is no workflow reference.

Sourceviktor.com/product, /no-api; readread 2026-09-16

Work reaches Genspark Claw without a person starting it, on a schedule, on a monitoring interval or through an inbound queue. The Schedules tab creates recurring tasks that Claw runs automatically, set as a plain time or a cron expression with a timezone, and Claw runs them at the set time even when the user is offline. Each task has a name, schedule and enabled or disabled state. When Heartbeat is on, Claw monitors the inbox, calendar and other connected services at a regular interval and acts on anything that needs attention, even without a message.

Genspark positions Schedules for clearly defined tasks and Heartbeat for ongoing awareness without a set task. Claw also has a dedicated, always connected email address at the genspark.email domain with an Allowed Senders list, so mail sent to it becomes work, and messages arrive over WhatsApp, Telegram, LINE, Slack, Microsoft Teams, Discord, Signal, Google Chat and Feishu, with conversations kept separate per channel. SecondBrain's background sync only brings data in and does not start a run.

Sourcegenspark.ai/helpcenter/genspark-claw, Schedules and Channels sectionsread 2026-09-12

Standing work runs on schedules with nobody asking, such as a Monday 9:00 ad spend recap in #marketing, a Friday 16:00 pipeline update in #sales and a 7:00 support inbox triage every morning, so the Monday report is in the channel before anyone opens a laptop. Viktor proposes scheduled tasks from recorded skills, which run once approved, users otherwise message him in Slack or Microsoft Teams, and the Public API starts tasks from other applications. Event triggers, such as a new ticket or record starting a run, are not described.

Sourceviktor.com/product; readread 2026-09-16

SecondBrain is a personal knowledge and memory system, a hub that learns from the user's emails, meetings, files and connected apps. Once a source is connected, data syncs automatically in the background and SecondBrain builds a personal context layer, so the store is kept up to date, not assembled for each run.

Sources include Gmail, Google Calendar, Outlook, Slack, Notion, HubSpot, Salesforce, Google Workspace and Microsoft 365, plus manual uploads up to 500 MB and saved past Genspark activity. Some integrations sync on a schedule, not instantly. SecondBrain searches across all connected sources to find any file, email or note, and Super Agent retrieves SecondBrain data at run time, so the store persists between sessions, is searched on demand and reaches past the context window. Genspark does not name the embedding, graph or index technology behind it.

Sourcegenspark.ai/helpcenter/secondbrainread 2026-09-12

Answers come from the tools and team threads a workspace connects. Viktor can answer what a team decided on a supplier switch from an earlier thread, with a summary attached, and works inside the tools a team connects, such as Google Drive, Notion, HubSpot, GitHub and 3,200+ others, pulling data from them for reports. There is no index of connected content, source refresh or citation surface, and context is assembled per run. Skills, Viktor's persistent memory, are a separate store.

Sourceviktor.com/product; readread 2026-09-16

Claw has a persistent memory that works across sessions and channels. When told to remember something, Claw writes it to long term memory and applies it in future conversations. Long term memory is shared across all channels while each thread stays separate per channel, and memory is what survives when long threads are compressed. Each user gets an isolated cloud computer and each Claw belongs to one person, so a Team plan buys seats, not a shared instance. Memory lasts as long as the cloud computer.

Delete VM permanently erases conversation history, memory, files and credentials, and on cancellation the machine is recycled and its data deleted after the billing cycle. What is stored is preferences, rules and context the user dictates, not a record the agent writes from its own runs. There is no way to list, edit or expire memories one at a time, or to delete a single memory without destroying the whole machine. SecondBrain, the separate knowledge hub, is not part of this memory.

Sourcegenspark.ai/helpcenter/genspark-claw, Memory section and FAQread 2026-09-12

Viktor keeps a persistent, workspace-scoped memory of written skills that a team can read, correct, export and delete. Skills are the persistent memory, scoped to the workspace, encrypted at rest, never used to train models and fully exportable or deletable on request, and they are walled off per workspace.

Viktor turns a screen recording into a skill, a written procedure the team can open and correct, so what he learned sits on paper where it can be read, and he keeps what he learns so week three is better than week one. Scope is the workspace, lifetime is until export or deletion, and skills are separate from the business record. Per user scoping and handling of stale memory are not described.

Sourceviktor.com/security, /product; readread 2026-09-16

Access controls govern what Claw can reach, but no review and approve step sits over what it does. Genspark says Claw is designed with human oversight in mind and that external actions like sending emails or posting messages require permission. Direct messages default to pairing mode, where only contacts the user has approved can message Claw, with a warning that an open policy lets anyone who knows Claw's address send it messages and prompt it to act, and the Claw email address has an Allowed Senders list.

Each service is authorized through an explicit login under a minimum access principle. Permission is granted when a service is connected, not for each action. After that the session persists, and Claw reads and sends emails, manages calendar events and reviews pull requests without the user opening the app, while scheduled tasks and Heartbeat run unattended. There is no approval queue, confirmation step or policy layer deciding which actions need sign off, and nowhere for an administrator to set one.

For actions that are hard to reverse, such as sending bulk emails, posting to social media, deleting files or executing financial transactions, Genspark advises asking Claw to show its plan before it starts, and reviewing planned actions where prompt injection risk is higher. The FAQ itself asks what happens if Claw does something wrong while the user is not watching, and asking the agent to pause is advice, not a control.

Sourcegenspark.ai/helpcenter/genspark-claw, Channels, Security and FAQ sectionsread 2026-09-12

Irreversible actions and new scheduled tasks wait for approval, and admins choose which tools need a yes. Before doing anything that can't be undone, such as emailing a customer list or touching what people pay, Viktor waits in the thread until someone presses Approve, with Approve and Reject shown on a held send. A proposed scheduled task stays paused until someone approves it, a thin brief gets questions and a wrong-looking number gets flagged.

Money moves, code pushes and customer emails wait for explicit approval in Slack, admins put high-risk tools behind human approval, and customers choose which actions need a yes in Slack or Teams before they happen in the tool. Approvals sit at tool and action level and vary by risk. Why a checkpoint fired is shown in examples, and no feature that explains it is named.

Sourceviktor.com/product, /security, /no-api; readread 2026-09-16

SOC 2 Type II and ISO 27001 are certified, ISO 42001 and GDPR are marked in progress, and a trust center sits at app.vanta.com/mainfunc.ai/trust. Genspark separates certified from in progress, and the SecondBrain page repeats the two certifications. The Team Plan adds centralized admin and billing controls covering member roles, usage analytics, SSO/SAML, invoices and connector management, so administrators control which connectors a team may authorize.

One of Genspark's own pages contradicts another. The Google Workspace listing says the add-on never accesses Drive, Gmail or other files, while the permission list on the same page requests a Gmail label scope, a Drive file scope and personal info. That is a broader permission scope than the function needs, a question of scope discipline, not of the certifications.

Sourcegenspark.ai/business Security & Compliance section, corroborated at genspark.ai/helpcenter/secondbrainread 2026-09-12

A SOC 2 Type 1 report, an independent attestation that Viktor's security controls operate as designed, is available under NDA, with Type II and ISO 27001 in progress and CASA Tier 3 shown. SAML SSO works across Okta (inside Slack), Entra ID, Google Workspace, OneLogin and any SAML 2.0 IdP. Admins decide which integrations are connected, who can use them and at what level, and can disconnect any integration, pause any user or kill a running task in one click. Viktor and its model providers are bound by no-training commitments. SCIM provisioning is not named.

Sourceviktor.com/security; readread 2026-09-16

Claw shows whether tasks are running and why they failed, but keeps no audit trail of what it did, when and on whose authority. The Schedules tab shows each task's name, schedule and enabled or disabled status, and if a task fails a Last Run Failed button appears, with Claw diagnosing the error and suggesting a fix.

A Diagnose tool inspects system health, checks running services and can restart components, and Terminal and Files allow direct inspection of the machine. The Team Plan adds usage analytics for administrators.

Nothing records which action Claw took in which connected service, at what time and under which authorization, nothing can be exported, no retention period is stated, and there is no administrator view across a team's agents, since each Claw belongs to one person.

Genspark says there is no built in way to share or export Claw chat history, because the conversation lives on the user's own cloud computer, and suggests asking the agent to write a summary of itself. The Team Plan MSA lets Genspark audit customers' records of their use of the services each year, which is Genspark auditing the customer, not a record of what the agent did.

Sourcegenspark.ai/helpcenter/genspark-claw, Schedules, Settings and Sharing sectionsread 2026-09-12

Enterprise workspaces get an audit log of every integration tool call Viktor makes, with the data that went in and out. The log is exportable by API, checked by daily integrity digests and kept for 400 days. A person can also watch Viktor's browser live through Viktor's Screen.

Sourceviktor.com/docs/audit-logs and viktor.com/changelogread 2026-10-25

Genspark Claw can run on a cloud computer that stays on 24/7 or locally through the desktop app for macOS or Windows, which needs no subscription and uses the customer's own machine, and so has access to its file system. The cloud mode is single tenant. It provisions a dedicated machine for each user with its own CPU, memory, storage and fixed IP address, and nothing is shared between users. That choice covers Claw only.

The rest of the workspace, Super Agent, AI Slides, AI Sheets, SecondBrain and AI Drive, stays cloud only on Microsoft Azure in the United States, and local mode gets a subset of features (chat, channels, skills, memory) that Genspark says will grow over time. There is no region list, EU option, region picker, VPC or on premises server deployment, and the privacy policy's clause on transfers to the United States still applies. A European customer with a localization obligation is served only if Claw on a laptop is all it needs.

Sourcegenspark.ai/helpcenter/genspark-claw, Quick Start Option B and Plans sections, against genspark.ai/privacyread 2026-09-12

Viktor runs from AWS us-east-1. The security page says EU data residency is available on Enterprise contracts, while the enterprise page says it is on the roadmap. Neither names a region list or a way to choose one, and no self hosted or customer VPC option is offered.

Sourceviktor.com/security and viktor.com/enterpriseread 2026-10-25

Customers get prebuilt work in three forms. The Help Center lists separate prebuilt agents, Super Agent, AI Slides, AI Sheets, AI Docs, Genspark Code, AI Meeting Notes, AI Pods, Call For Me, Custom Agent, GenMail, Realtime Voice, Genspark Design and Genspark Claw, and each does its own job, so AI Sheets works without AI Slides and AI Meeting Notes works without Call For Me.

The Tools index at genspark.ai/tools has about 150 tool pages, each usable on its own, under Writing & Content, Design & Visuals, AI Models, Audio & Video, and Business & Productivity, such as AI CRM Builder, AI Pitch Deck Generator, AI Phone Call and AI Invoice Generator, and the Business page counts more than 150 in house toolsets for complex business operations. Skills let a user teach Super Agent a workflow once, save it, run it any time and share it with the whole team, alongside an official skills library.

Sourcegenspark.ai/helpcenter and genspark.ai/businessread 2026-09-12

A Marketplace offers installable skills, bundles and automation templates that a team adopts workspace wide with one click.

Sourceviktor.com/docs/skills-marketplace and docs/automation-templatesread 2026-10-25

Customers can choose models, and Super Agent also routes subtasks automatically across providers including OpenAI, Anthropic, Google, xAI and ElevenLabs when the user does not choose. Genspark's credit guidance tells users to choose a lower model tier when they do not need maximum power. The Team Plan lists the choices for each kind of agent.

Chat offers Claude Opus 5, GPT-5.6, Gemini 3.8 Flash, Grok 4.6, DeepSeek V4 Pro, Kimi K3, MiniMax M3 and GLM-5.3. Image offers Recraft V4.1, Ideogram V4, Qwen Image 3, Seedream 5.0 Pro, GPT Image 2, Nano Banana Pro, Nano Banana 2, Seedream v5 Lite and Flux 2 Pro. Video offers Gemini Veo 3.1, Seedance v2, Kling V3, MiniMax H3 and PixVerse V6. Audio offers ElevenLabs, MiniMax, Mureka, Google, Microsoft and CassetteAI.

Genspark advertises more than 70 AI models including ChatGPT, Claude and Gemini. The Tools index adds landing pages for individual models under an AI Models heading, such as genspark.ai/tools/claude-sonnet-5, /grok-4-5, /gpt-image-2, /seedream-5 and /nano-banana. Azure, AWS and GCP on the subprocessor list are infrastructure, not model providers. There is no admin policy for which models a team may use.

Sourcegenspark.ai/business Team Plan section and genspark.ai/helpcenter/super-agentread 2026-09-12

Admins pick the model across OpenAI, Anthropic, Google and xAI, down to a single scheduled task. A workspace can also bring its own OpenRouter key for more than 300 models.

Sourceviktor.com/enterprise and viktor.com/changelog; viktor.com/productread 2026-10-25

There is no published way to call the platform. Genspark has no REST or GraphQL API, SDK, developer portal, MCP server, A2A agent card, endpoint or authentication scheme. The Help Center's 28 product articles and 6 account articles include no developer, API or integration building section, and the Business page footer has no developer entry. Every surface points outward.

The Chrome Extension, Microsoft Office Plugin, Salesforce Integration and the SecondBrain connectors (Gmail, Calendar, Outlook, Slack, Notion, HubSpot, Google Workspace, Microsoft 365) all let Genspark reach into other systems, not let other systems reach Genspark. AgentBase, in a product family with Custom Agent, Skills and Hub, most likely builds agents inside Genspark. GenTerminal is a separate product at genterminal.ai, a terminal, not a way to drive this platform.

Sourcegenspark.ai/helpcenter and genspark.ai/businessread 2026-09-12

The Viktor Public API lets developers create Viktor threads, run tasks, continue conversations and download result files from their own applications. The Viktor MCP Server lets an MCP-compatible agent connect with streamable HTTP and a scoped Viktor API key, and an OpenAI- and Anthropic-compatible API drives Viktor agent runs with the OpenAI or Anthropic SDKs. These SDK-compatible endpoints and the served MCP server let Viktor fit into other agent systems, with reference pages at /docs/public-api, /docs/mcp-server and /docs/chat-completions.

Sourceviktor.com/docs; readread 2026-09-16

Customers have no way to evaluate a change. There is no evaluation harness, scored test case, quality gate in a release path or controlled experiment after deployment, and nothing in Super Agent, SecondBrain, the Business plan or the Help Center lets a customer test an agent before or after relying on it. The new Super Agent checks its own output and builds on earlier progress, but that is the agent checking itself inside a run, with no result the customer reads.

Genspark's engineering blog describes collecting real world examples of successful tool selection and model routing and building them in through reinforcement learning and prompt playbooks, a loop that improves Genspark's engine for all users, which the customer neither runs nor sees the results of. The advice to refine a prompt in AI Chat before handing it to Super Agent is drafting guidance, not a test. No bias or fairness audit is published, and ISO 42001 is still in progress.

Sourcegenspark.ai/helpcenter/super-agent and genspark.ai/businessread 2026-09-12

No way to test a change to skills, instructions or scheduled tasks, before or after it goes live, is described. New scheduled tasks are held paused for human approval and irreversible actions wait for Approve or Reject, Viktor questions thin briefs and flags numbers that look wrong before reporting them, and usage and cost are reported per run. None of these tests a change, and there is no simulation, dry run, test case or quality score.

Sourceviktor.com/product, /security, /no-api, /docs; readread 2026-09-16

Genspark Claw uses a browser, controls a remote computer and can control the local computer. Connecting a service opens a remote desktop window, the browser on the user's cloud computer already on that service's login page, and the session persists so Claw can act in that service afterwards. A Remote Desktop button opens a browser based view of the cloud computer, with a resettable password, a Terminal tab gives command line access in the browser, and a Files tab is a browser based file manager.

The desktop app runs Claw on the user's own computer for macOS or Windows, with access to the file system, and a workspace folder that Genspark calls soft guidance, not a hard boundary. The user picks a tier (Lite, Standard, Powerful), provisions a dedicated machine with its own CPU, memory, storage and fixed IP, watches it through Remote Desktop, drives it through Terminal, and manages it through Refresh Status, Update Configuration, Diagnose, Force Restart and Delete VM.

Genspark warns that when Claw browses websites, reads emails or processes documents, malicious content may attempt prompt injection, and advises reviewing planned actions in less controlled environments. Terminal and shell access is command execution, not interface operation, and Call For Me works through a telephone keypad, not a computer.

Sourcegenspark.ai/helpcenter/genspark-claw, Services, Home, Terminal and Security sectionsread 2026-09-12

Viktor signs into tools that have no API with its own seat and works on their screens from its own cloud computer, so any software a person can use, Viktor can use.

Customers create a seat for Viktor in a tool with its own email and permissions, and Viktor signs in from his own computer in the cloud and works on the same screens the team uses, for example logging into Greenhouse to move candidates between stages, or working in QuickBooks to pull overdue invoices and draft reminders, without an API or developer key.

Because the agent works third party web applications through their user interface, changes to those screens are a real risk, and resilience to UI changes and the list of supported applications are not described.

Sourceviktor.com/no-api; readread 2026-09-16