WorkAist

The Company Brain for AI-native business operations.

Define processes in plain language. AI agents execute, collaborate, and self-govern through a structured protocol with shared corporate context, permission isolation, and reversible PII redaction at the LLM boundary. Self-hosted, AGPL-3.0.

Companies adopt AI tools, but each one solves a narrow task. The result: a patchwork with no shared context, no governance, no visibility.

WorkAist is a different approach. AI agents are the business operations layer — they share corporate data, communicate through a protocol, respect permissions, and form a visible process graph.

For service providers — accountants, agencies, and outsourcers — WorkAist ships as a turnkey package: a single deployment per client with the full Company Brain inside, ready to extend with your domain expertise.

How it works

1

Describe

Describe what you want to automate in plain language. The AI Wizard asks clarifying questions, checks for duplicates against existing entities, and proposes a plan. Confirm and the agent, connector, or process is created for you.

AI Wizard step 1AI Wizard step 2AI Wizard step 3
2

Refine

Each agent is a folder with a CLAUDE.md file. Open it in the browser-based VS Code with Claude Code. Edit, test, iterate.

Refine
3

Operate

Agents run on schedules, webhooks, manual buttons, or event chains. The Process Map shows the full graph live, with tokens and dollars per run and explicit owners for each entity.

Operate
4

Understand

WorkAist auto-discovers relationships between every table across all connectors and lets you fix the gaps it misses by hand. Agents stop seeing isolated tables and start seeing your whole business as one Knowledge Graph.

Understand

Entity Protocol

A structured protocol for AI entity interaction. Security, governance, and full visibility, with a strict privacy boundary between your data and the LLM.

Connector
External API
Agent
Business domain
Process
Automated workflow
via MCP Hub
Connector

Bridge to an external service. One connector = one service, exposed as its own MCP server. Data syncs to PostgreSQL.

Agent

AI worker for a business domain. Defined in plain language. Communicates through agents-hub.

Process

Automated workflow with a trigger (cron, webhook, button, event chain). Visible on Process Map.

Call chain protection (circular call detection, depth limits)
Permission isolation (group → agent → table-level ACL)
Usage tracking per agent, per user, per model
Audit logging for every action
Reversible PII redaction at the LLM boundary (sensitive values are tokenized before prompts and restored on the response)
Knowledge Graph: a cross-entity context graph that links agents, connectors, processes, and data tables for shared corporate memory

Beyond automation

Master Agent sees every process, cost, and outcome across your company. It finds bottlenecks, predicts trends, and surfaces insights humans miss.

Prediction

Your peak last year was March 8. This year holidays shifted — expect it March 22. Start prep a week earlier.

Cross-references historical sales data with calendar events and external signals.

Growth

32% of your clients bring referrals organically. A referral program could push this to 60% — that's +$8K/mo.

Analyzes customer acquisition patterns and models impact of incentive programs.

Bottleneck

Invoice approval averages 3.2 days — the manual sign-off step accounts for 80% of the wait. Remove it, process runs in 4 hours.

Analyzes Process Map execution times to find the slowest step in every workflow.

Key Features

Process Map

Interactive DAG of all agents, connectors, data flows, costs, success rates, and explicit owners. Process chaining lets one process trigger the next on success or failure.

Impact Timeline

Per-entity timeline of snapshots, decisions, and delegation handoffs. See exactly when work moved from one agent to another and what changed.

AI Wizard

Describe what to automate — the platform builds it, asks clarifying questions, and avoids duplicates.

VS Code in Browser

Per-user code-server with Claude Code CLI for the whole team in one IDE.

Connectors

One service = one connector, exposed as its own MCP server. Data syncs to PostgreSQL with explicit ACLs.

Telegram Bot

One bot lives in your corporate chats: mention an agent to trigger it, get notifications, and approve actions inline.

Enterprise Governance

Audit log, flood valve, version management, permission isolation, and reversible PII redaction at the LLM boundary.

Self-hosted & AGPL

Your data stays yours. Same license model as GitLab CE.

How WorkAist compares

WorkAistDifyn8nCrewAI
Built forBusiness opsAI app buildersAutomationAI developers
Agent definitionPlain languageVisual workflowNode editorPython code
Inter-agent protocolMCP + ACLWorkflow routingManualPython
Process visibilityVisual DAGWorkflow viewWorkflow viewNone
Permission isolationGroup → tableBasic RBACEnterpriseNone
PrivacyReversible PII redactionNoneNoneNone
Self-hostedDocker ComposeDockerDockerpip install
Cost trackingPer-agent/userNoneNoneNone
LicenseAGPL-3.0Apache + prop.Fair-codeApache

Dify builds AI apps. n8n automates workflows. CrewAI orchestrates Python agents. WorkAist is a framework for running your company on AI.

Use Cases

RUNNING IN PRODUCTION

Web & SEO

3 agents, 4 connectors

  • Site content generation from product data
  • Programmatic SEO landing pages
  • Search Console + analytics monitoring
  • Competitor scraping and content briefs
RUNNING IN PRODUCTION

HR & Operations

2 agents, Telegram bot

  • Candidate intake and CV screening
  • Onboarding checklists per new hire
  • Internal request handling in corporate chat
  • Daily standup and absence tracking
RUNNING IN PRODUCTION

Revenue & Reporting

2 agents, 3 connectors

  • Invoice generation and follow-up
  • Cashflow snapshots from accounting data
  • Weekly revenue digests for management
  • Client-account anomaly alerts

All three areas above run inside a single WorkAist instance — 16 entities sharing one Knowledge Graph, one database, and one corporate Telegram bot. Three business domains, one Company Brain.

Partner with us

We are looking for service providers — accountants, agencies, and outsourcers — who want to deploy WorkAist as a turnkey Company Brain for their SME clients.

Eligible partners can be co-funded through the European Digital Innovation Hubs (EDIH) network and the Digital Europe Programme. We provide the platform, training, and deployment support; you bring the client relationship and domain expertise.

Quick Start

$ git clone https://github.com/aist-guru/work.git
$ cd work
$ docker compose up -d

Open http://localhost — the Setup Wizard runs in your browser and configures everything; no .env files to edit.