Straight answers about the WorkAist platform — what it is, how to run it, how it handles your data, what it costs, and how it compares to the tools you already use.
WorkAist is a self-hosted platform for AI-native business operations — a "Company Brain" where you define processes in plain language and AI agents execute them across your real tools. It is open-source under AGPL-3.0 and connects to business software through the open Model Context Protocol (MCP).
MCP is an open standard that lets AI agents talk to external software through a clean, typed tool interface instead of brittle scripts. WorkAist uses MCP so any agent can read from and act on a connected service safely, with each capability exposed as an explicit, permissioned tool.
A chatbot answers questions; a WorkAist agent does work. It runs as a member of your team — reading from your stack, acting under explicit permissions, and reporting outcomes to humans — rather than waiting for prompts in a chat window.
Learn moreTrigger-action automations follow fixed if-this-then-that rules. WorkAist agents reason over the actual content of a task, branch when reality differs from the happy path, and coordinate multiple steps — so you describe the goal instead of wiring every node.
Learn moreNo. You define processes in plain language and assign them to agents. Developers can go deeper — building custom connectors or tuning agent behaviour — but the core workflow is written in natural language, not code.
Anything reachable through a connected tool: categorise and reconcile transactions, draft and send communications, update CRM records, prepare reports, route documents, and more. Capabilities are bounded by the connectors you enable and the permissions you grant.
Learn moreBoth. You can self-host the open-source platform on your own infrastructure, or use the managed cloud at workaist.com if you would rather not run it yourself. The product is the same; only who operates the servers differs.
WorkAist is model-agnostic. Because it sits in front of the LLM boundary, you can point it at the providers or local models you trust, and keep sensitive data redacted before it ever reaches a model.
WorkAist is built for real operational use, with permission isolation, reversible actions, and human reporting designed in from the start. As with any automation platform, roll out agent by agent and process by process, expanding scope as you confirm results.
Pull the open-source release, provide your configuration and a Postgres database, and start the stack on your own server. The platform is designed to run as a standard containerised service on infrastructure you control.
A modern Linux host with Docker and a Postgres database is enough to start. Sizing scales with how many agents and connectors you run concurrently; light workloads run comfortably on a small VM.
Yes. The platform is distributed to run as containers, so it deploys the same way on a laptop, a single VM, or an orchestrated cluster.
A first self-hosted instance can be up in well under an hour for someone comfortable with Docker. The managed cloud is faster still — sign up and your instance is provisioned for you.
Learn moreSelf-hosted operators pull new releases on their own schedule and apply database migrations as part of the upgrade. On the managed cloud, updates are applied for you.
Yes — self-hosting is designed for exactly this. You can run WorkAist entirely inside your own network and point it at internal or self-hosted models so no business data leaves your perimeter.
WorkAist stores its state in PostgreSQL. You provide the database, which means backups, encryption-at-rest, and retention stay under your existing data-governance controls.
Because state lives in your Postgres database and your configuration files, your normal backup and disaster-recovery tooling covers WorkAist. On the managed cloud, backups are operated for you.
Run more agents and connectors as your needs grow, and scale the underlying containers and database the way you would any production service. Start small and expand process by process.
On self-hosted deployments, all of it stays on your own infrastructure — WorkAist does not phone home with your business data. On the managed cloud, your data lives in the instance provisioned for you.
Only what a task genuinely needs, and only after passing the LLM boundary where sensitive fields are redacted. You choose which models to use, including local ones, so you control exactly what leaves your environment.
WorkAist performs reversible PII redaction at the LLM boundary: identifiers are masked before content reaches a model and restored in the result. The model sees the shape of the work without seeing the underlying personal data.
WorkAist is built to support GDPR obligations: self-hosting keeps processing within your control, PII redaction limits what reaches third parties, and EU data residency is straightforward when you run it yourself. Compliance ultimately depends on how you configure and operate it.
Yes. Self-host in an EU region, or use EU-based infrastructure and models, and your data never has to leave the jurisdiction. This is a core reason teams with strict residency rules choose a self-hosted platform.
Every agent operates under explicit, isolated permissions — it can only touch the tools and data you grant. Permission isolation means one agent's scope never silently becomes another's.
Learn moreAgents report outcomes to humans rather than acting invisibly, so there is a record of what was done and why. Combined with reversible actions, this keeps automated work reviewable and accountable.
WorkAist favours reversible actions and human reporting, so mistakes can be caught and undone rather than silently propagating. You also choose how much autonomy each agent has before it must hand off to a person.
Transport is encrypted in transit, and because state lives in your own Postgres and infrastructure, encryption-at-rest follows your existing standards. Self-hosting puts key management entirely in your hands.
The platform is open-source under AGPL-3.0, so you can self-host it at no licence cost. You pay only for your own infrastructure and any AI model usage. The managed cloud is a paid convenience for teams who prefer not to operate it.
AGPL-3.0 lets you run, study, modify and share WorkAist freely. The main obligation is reciprocity: if you modify it and offer it to others over a network, you must make your changes available under the same licence. Internal use and self-hosting are unrestricted.
Yes. AGPL-3.0 permits commercial use, including running WorkAist to operate your own business. The licence governs distribution of the software itself, not the value your agents create.
The managed cloud is a paid hosted version for teams who want WorkAist without running servers. Start from the sign-up flow to see current plans; self-hosting remains free under AGPL-3.0 if you prefer.
Learn moreThe software and its capabilities are identical. Self-hosting is free and you operate it; the managed cloud is paid and we operate the infrastructure, updates, and backups for you.
Self-hosting has no per-agent or per-connector licence fee — run as many as your infrastructure supports. Managed-cloud plans define their own limits, which you can review at sign-up.
The platform is genuinely open-source under AGPL-3.0, not a stripped-down teaser. The managed cloud sells operation and convenience, not access to otherwise-missing features.
No — you bring your own model access, so model usage is billed by whichever provider or local setup you choose. This keeps you in control of both spend and which models see your data.
Map each Zap to the outcome it produces, then describe that outcome to a WorkAist agent and enable the matching connectors. Agents replace brittle multi-step Zaps with a single goal that adapts to the content of each task.
Learn moreMake scenarios translate into agent processes: instead of laying out every module and route, you state the goal and grant the tools. Complex branching that bloats a Make scenario is handled by the agent's reasoning.
Learn moren8n and WorkAist are both self-hostable, which makes residency parity easy. Move workflow by workflow: re-express each n8n flow as a plain-language process and connect the same services through MCP.
Learn moreYes. Most teams migrate incrementally, leaving existing automations in place while moving high-value or hard-to-maintain workflows to WorkAist agents first.
If a service has an MCP connector in the marketplace (around 1,300 today), it works out of the box. For anything missing, you can build a connector — the AI Wizard helps generate one.
Learn moreIt scales with how many workflows you move and how complex they are. Because you describe goals rather than rebuild node graphs, migrating a workflow is often faster than recreating it in another trigger-action tool.
Workflows are re-expressed as plain-language processes rather than imported file-for-file, because an agent's process is structurally different from a fixed automation graph. The upside is that one process can replace several rigid automations.
You can build one. WorkAist's connectors follow the open MCP standard, and the built-in AI Wizard helps scaffold a new connector for a service that isn't in the marketplace yet.
Learn moreAn agent is an autonomous worker assigned to a role and a process. It reads from your connected tools, decides the next step, acts under explicit permissions, and reports the result — like a teammate with a narrow, well-defined remit.
Learn moreAgents operate within a structured protocol that defines their permissions, shared context, and boundaries. Self-governance means they coordinate and stay inside those rules autonomously, escalating to humans when a task exceeds their remit.
Each agent gets its own scoped set of tool and data permissions. One agent cannot reach into another's access, so a finance agent and a support agent never share more than you explicitly allow.
WorkAist is designed around reversible actions and human reporting, so consequential steps can be reviewed and undone. You decide where an agent acts on its own and where it must wait for human approval.
Yes. Agents report outcomes to people and operate under permissions you set, so humans direct the work and keep final authority. The goal is leverage, not unsupervised autonomy.
Yes. Agents collaborate through the shared corporate context and structured protocol, handing tasks between specialised roles the way a team divides work — without losing the permission boundaries between them.
You describe the goal and the steps in plain language, grant the relevant connectors, and set the permissions. WorkAist turns that description into a process the agent executes and reports on.
Risk is contained structurally: agents act only through typed, permissioned tools, sensitive data is redacted at the model boundary, actions favour reversibility, and outcomes are reported to humans. The platform constrains what an error can touch.
Yes. WorkAist offers ready-to-deploy agent roles for common functions so you can start from a proven template and adapt it, rather than designing every agent from scratch.
Learn moreAround 1,300 MCP connectors are available in the marketplace today, spanning accounting, CRM, communication, e-commerce, developer tools, and many country-specific services.
Learn moreA connector exposes a service — like an accounting tool or CRM — to your agents through the open MCP standard. It turns that service's capabilities into typed, permissioned tools an agent can use safely.
Learn moreYes. If a service you need isn't in the marketplace, you can request it or build it yourself — connectors follow the open MCP standard, so they aren't locked to WorkAist.
Learn moreYes. Because connectors are MCP-based, you can author one for any service with an API, and the built-in AI Wizard helps scaffold it — useful for internal or niche tools.
Connectors use the connected service's own authentication — OAuth, API keys, tokens, or basic credentials — and run under your control. Credentials stay in your environment on self-hosted deployments.
Yes. The marketplace includes connectors for regional compliance systems — for example electronic-invoicing platforms — alongside global SaaS, so local obligations can be automated like anything else.
Learn moreMost connectors install in a click and then authenticate against your account. From there an agent can use the connected tool as part of any process you define.
Learn moreThe marketplace grows over time as new services and regional tools are added. Because connectors are open and MCP-based, the catalogue isn't gated behind a vendor release cycle — you can add what you need when you need it.
Learn more