AigenzeyDocs
Home/Docs/Overview

Aigenzey Cowork Documentation

Aigenzey Cowork is an autonomous, skill-aware AI agent engineered for enterprise teams and individual professionals. It connects directly into your workplace tools—mail, calendars, documents, CRMs, ticket trackers, and databases—executing multi-step workflows autonomously and transforming proven playbooks into reusable organizational skills.

Looking for Connector Setup?

Jump straight to our Connectors Setup Reference for detailed configuration instructions on all 49+ OAuth providers, Custom MCP Servers, OpenAPI endpoints, and SQL/NoSQL databases.

What is Aigenzey Cowork?

Unlike conventional chatbots that only provide conversational answers, Aigenzey Cowork is an action-oriented autonomous agent. It connects to your existing software stack using secure user-scoped OAuth tokens, investigates live data across multiple systems, plans and executes workflows, and delivers concrete, finished artifacts (documents, spreadsheets, code diffs, slides, or CRM updates).

Single-Agent Triage
Autonomous Execution
No rigid handoffs or fragmented mini-agents. One general-purpose agent handles simple lookups immediately and executes deep multi-step workflows with continuous context.
🔌
50+ Connectors & MCP
Zero Data Silos
Native integrations for Google Workspace, Microsoft 365, Atlassian, Salesforce, HubSpot, Zendesk, NetSuite, Workday, plus arbitrary Custom MCP servers and SQL databases.
🧠
Progressive Skill Discovery
Adaptive Intelligence
The agent automatically discovers and attaches relevant organizational skills from the Skills Library based on task intent, minimizing token bloat while maximizing precision.
🛡️
Enterprise Governance
Fernet Encrypted & Scoped
All OAuth tokens are Fernet-encrypted at rest and strictly scoped to (user_id + org_id). Complete run observability, token budget quotas, and audit logs.

Core Architecture & Principles

Aigenzey Cowork is built on modern agentic infrastructure engineered for reliability, safety, and transparency:

  • Google ADK + LiteLLM Core: Flexible multi-model support (Gemini 3.7 Flash, Claude 3.7 Sonnet, GPT-4o, o3-mini) with native prompt caching and deterministic fallback strategies.
  • Progressive Disclosure: Rather than dumping dozens of tool schemas into the model prompt, Cowork indexes skills and tools in a vector + full-text search database, dynamically injecting only what is needed for the user's specific goal.
  • Live Server-Sent Events (SSE): Every reasoning step, tool call, token counter, and intermediate status is streamed in real time to the web interface.
  • Artifact Generation: Large structured outputs (formatted reports, HTML dashboards, spreadsheets, slides, code diffs) are automatically segmented into interactive, collapsible artifacts.
  • Strict User-Level Isolation: OAuth tokens and database credentials are keyed strictly by user_id + organization_id. User A can never access or leverage tokens authorized by User B.

Quick Start Guide

Follow these simple steps to get started with Aigenzey Cowork:

Step 1: Sign in to your workspace

Navigate to Aigenzey Cowork. You can sign in using your Google account for personal productivity or use your organization's business login if your IT team has provisioned workspace access.

Step 2: Connect your tools

Click on Connections in the left sidebar (or let the agent prompt you dynamically when you request an action). Authenticate your essential tools—such as Gmail, Google Drive, Slack, GitHub, or Microsoft Outlook. Tokens are encrypted at rest with Fernet cryptography.

Just-in-Time Authorization

You do not need to connect all tools upfront! When you ask Cowork to perform a task (e.g. "Find customer tickets in Zendesk and post a summary to Slack"), the agent will automatically detect missing connectors and provide a one-click Connect button right in the chat stream.

Step 3: Ask the agent to execute a workflow

Type your request in natural language. You can attach documents (PDFs, Word documents, Excel sheets, PowerPoint decks, CSVs) or paste raw data directly. Examples:

  • "Audit all customer churn tickets in Zendesk from last week, cross-reference invoice values in Stripe, and generate an executive summary table."
  • "Review this contract PDF, identify non-standard indemnity clauses, and compare them against our standard procurement guidelines."
  • "Find all Jira tickets assigned to the Mobile Sprint, check pull request statuses on GitHub, and post a standup brief to #eng-team in Slack."

Step 4: Save playbooks as reusable Skills

When the agent completes a multi-step workflow successfully, you can click Save as Skill. Give it a name, define parameter inputs (e.g. date ranges, client names), and share it with your department or organization.

Key Capabilities

FeatureDescriptionWhere to Learn More
Natural Language ChatStreaming, real-time reasoning with tool call previews and interactive artifacts.Usage Guide →
Connector Integrations49+ OAuth providers, custom MCP servers, OpenAPI specs, and SQL databases.Connectors Reference →
Document ParsingNative text & structure extraction for Word, PPTX, Excel, PDF, CSV, and image attachments.File Attachments →
Guided Skill BuilderConvert successful interactions into versioned, department-scoped automated skills.Skills Guide →
Scheduled Tasks & WebhooksAutomate recurring daily/weekly reports or trigger runs from external webhooks.Automations →
Deep Research ModeEscalate complex inquiries to run multi-source corroboration with parallel subagent branches.Effort Modes →
Enterprise SSOSAML/OIDC authentication with Entra ID, Google Workspace, Okta, and JIT provisioning.Single Sign-On →
Admin & QuotasManage organizational members, departments, token quotas, and audit logs.Admin & Security →

Documentation Map

Explore detailed guides across the platform: