Business software and AI production in one account — messages, tasks, files, signatures, portals, CRM, support, documents, media and automation.
Every tool is exposed to compatible agents over MCP, against the same account data your team already uses.
Raq.com is a business workspace designed for people, teams and AI agents to use together. An agent can list your boards, upload a file, run an image batch or read a project record — and everything it does stays visible in the product, where the team can review, correct and carry on.
Small and medium businesses build their operating system by accumulation — a chat product, a task board, a signing service, a form builder, a monitoring tool and several AI subscriptions. Each has its own permissions and its own history. An agent becomes far less useful when it can only reach one piece of that.
Raq.com exposes its tools to compatible agents through the Model Context Protocol. An agent gets structured operations — list a board, create a card, upload a file, run an image batch, read a project record — instead of imitating clicks in a browser. The browser and the agent work against the same records.
Open the board, the document or the approval queue and continue from where the agent stopped — no export, no re-entry.
Bring Codex, Claude Code or another compatible client and point it at the same account, with the permissions attached to that connection.
Access is account-scoped and tool access can be limited. Sensitive or external actions use explicit approvals — Ask My Agent and Voice Calls are both approval-controlled.
Role-based permissions and audit logging show who — or what — did what, so agent activity does not disappear into a terminal.
One subscription includes the full public catalogue — conventional business software alongside specialist AI production and agent operations. Plans differ in AI usage and service limits, not in which tools you can open.
Supercomputer is Raq.com's own agent. A user describes a task in plain English and it selects suitable platform tools and runs them against the account — no coding environment to configure, no separate subscription to buy.
It works with the same files, records and permissions as connected external agents, so nothing is duplicated.
The Skill Library stores versioned instructions that guide Supercomputer across sessions, so a team's way of working sticks.
Connect Codex, Claude Code or another compatible client and work against the same account instead.
Supercomputer will not message an external party without a decision from someone in the account.
A single tool is useful on its own. The shared records are what let a job carry on to the next step.
Automations run these multi-step processes on an event or a schedule. Orchestrator runs templated workflows across managed Codex and Claude sessions, while Nature Work and Agent Operator HQ keep the terminals under supervision.
These are not chat products. They deal with execution, supervision, approvals, hand-offs and the record left behind — the parts of agent work that get difficult once more than one is running.
Run templated workflows across managed sessions, so a job that worked once can be run again without rebuilding it by hand.
A full-screen command view with keyboard controls and a live message queue, so one operator can keep several streams of work moving without losing the thread.
Connects local terminal sessions to a browser-based viewer with keyboard and supported controller input, so an operator can step away from the desk without stepping away from the work.
Separates work into human, mixed and agent-owned tasks, so an operator can tell at a glance what is running, what is waiting on a decision and what needs doing by hand.
The distinction matters, because the runtime, the commercial responsibility and the support model are different. One is a self-build; the other is managed software.
Prices exclude VAT where applicable. Plans differ in monthly AI usage and service limits — monitoring quantities, included hosted sites, FinUK queries — not in tool access.
For individuals and very small teams getting set up.
Limited monitoring quantities
For teams putting agents to work on real operations.
Higher AI usage and service limits
For companies running several teams and client accounts.
Substantially larger allowances
For organisations with the highest usage and requirements.
Highest AI and service limits
New accounts get a free trial: one use of each AI tool and limited monitoring. No card required to begin.
AI work draws from the account's monthly allowance and is shown as AI usage, not internal credit counts. If it runs out, AI tools pause until renewal or a plan change. Non-AI tools carry on.
Upgrade at once, or schedule a downgrade for the next billing date. The total changes as users join or leave the account.
A short rewrite, a multi-model report and a generated video consume very different amounts. The pricing page gives examples per tool rather than promising one universal request count.
Core text and chat requests use Zero Data Retention routes by default, so the model provider cannot keep prompts or responses for training. Raq.com states that it does not train models on customer content.
Default route for core text and chat, so prompts are not retained for provider training. Live web search has its own setting.
Encryption in transit with TLS 1.3, and AES-256 full-disk encryption at rest on dedicated hardware.
Role-based permissions and audit records show who did what — including what an agent did on the account's behalf.
Dedicated servers in Hetzner's German data centres, with UK GDPR-aligned handling.
Optional two-factor authentication for user accounts, alongside account-scoped agent access.
Continuous monitoring and a defined incident-response process, backed by Raq.com's own billing and reliability records.
A distinction worth keeping. Image, video and voice tools send the prompt and required generation parameters to specialist providers, who operate under their own API retention terms. The Zero Data Retention statement applies to core text AI, not to every media tool. Teams assessing Raq.com for regulated or sensitive data should read the current Privacy Policy, AI Privacy & Security page and Subprocessors list, then map their own legal and sector requirements onto the specific tools they plan to use.
Raq.com does not claim feature-for-feature equivalence with every mature specialist product. The value is the breadth of useful tools, the shared account and the agent layer across them.
In those cases Raq.com can still sit around the existing system through Automations, APIs or a Custom Module — but the connection and compliance work need their own assessment.
Raq.com is built by Vu, an AI and software company based in the United Kingdom. Vu uses Raq.com for its own work and builds the managed Custom Modules behind it — so the platform is tested on real operational jobs, not only on demos.
The public comparisons on Raq.com describe overlapping workflows rather than exact replacements, and the product changes quickly. Prices, limits and the tool catalogue should be checked on raq.com before you rely on them.
Create an account, try one use of every AI tool, and point an agent at it when you are ready. No card required to start.