Cowork

Overview

Cowork is a plugin-based workspace and deployment surface for AI workflows, introduced as a research preview that lets Claude access, read, and edit files in a designated folder to automate multi-step, non-technical tasks. Across product updates, it has evolved from a file-centric assistant environment into a broader execution layer with plugins, MCP connectors, live dashboards, Windows support, and local model support. It is also explicitly referenced as a place to run packaged Claude workflows, including financial-services agent templates.

For AI Product Managers, Cowork matters because it represents a practical interface between general-purpose AI models and operational business work. Rather than treating AI as just a chat surface, Cowork shows how agentic workflows can be packaged, connected to enterprise data sources, and deployed into repeatable task environments. That makes it useful both as a product benchmark for AI-native workspaces and as a case study in how deployment surfaces for agents are converging around files, connectors, plugins, and managed execution.

Key Developments

• 2026-01-13: Claude introduced Cowork as a research preview. The initial product let Claude access, read, and edit files in a designated folder to automate non-technical tasks such as generating spreadsheets from screenshots or drafting documents from scattered notes.
• 2026-01-18: Clement Delangue highlighted local model support in Cowork, positioning it as an option for keeping data on-device rather than sending it to the cloud.
• 2026-02-11: Cowork launched on Windows with full MacOS feature parity, including file access, multi-step task execution, plugins, and MCP connectors. Boris Cherny also covered this launch.
• 2026-03-03: Santiago called out friction in Cowork's multi-step file workflow, noting that granting folder access, copying files, and then generating a plan creates overhead when working with many or frequently changing files. He suggested more native chat-based file support.
• 2026-04-21: Claude added the ability to build live dashboards and trackers in Cowork, connected to apps and files and refreshed with current data whenever reopened.
• 2026-05-06: Claude launched ready-to-run financial services agent templates for pitch building, valuation reviews, and month-end close workflows. These were made available as plugins in Cowork and Claude Code, or deployable via cookbooks as Managed Agents.

Relevance to AI PMs

• Evaluate agent deployment UX: Cowork is a useful reference for how AI workflows move from prototype to operational tool. PMs can study its mix of folders, plugins, connectors, and templates to understand what users need to trust and adopt agentic systems.
• Benchmark enterprise readiness: Features like Windows parity, MCP connectors, local model support, and live dashboards signal the capabilities buyers increasingly expect in AI workspaces. PMs can use Cowork as a checklist for platform gaps in their own products.
• Design repeatable workflow packaging: The financial-services templates show how domain-specific AI tasks can be turned into reusable packaged workflows. PMs building vertical AI products can apply the same pattern to standardize high-value recurring jobs.

Related

• Claude: Cowork is closely tied to Claude as the model and product ecosystem powering many of its workflows and launches.
• Claude Code: Financial-services agent templates were distributed in both Cowork and Claude Code, suggesting parallel deployment surfaces for packaged AI workflows.
• Managed Agents: The same templates available in Cowork could also be deployed as Managed Agents, linking Cowork to a broader execution and orchestration model.
• MCP / desktop-commander-mcp: Cowork's support for MCP connectors places it in the emerging model-context integration ecosystem for tools and data access.
• Anthropic: As the company behind Claude, Anthropic is the broader platform context in which Cowork evolves.
• Claude Desktop: Related as part of the Claude product environment, especially where desktop-based workflows and local execution surfaces overlap.
• Boris Cherny, Santiago, Clement Delangue: These individuals helped shape external understanding of Cowork through launch coverage, critique, and discussion of local model support.