Anthropic’s Conway Points Claude Toward Always-On AI Agents
Anthropic is preparing a Claude expansion that includes Conway, Orbit, Operon, memory upgrades and multilingual voice mode. The move signals a shift from chat-based AI toward persistent assistants that can connect with external services and manage workspaces. Enterprises, developers and research teams could be affected if Claude becomes a broader agent platform.
What happened
Anthropic is preparing a broad expansion of Claude that would push the service beyond a chat interface and toward an autonomous-agent platform.
Code analysis and test-catalog material dated May 31 local time point to several features in development: an always-on agent called Conway, a proactive assistant called Orbit, a bioscience desktop environment called Operon, a knowledge-based memory system and multilingual voice mode.
Conway is the main signal.
It is described as an agent that can run independently inside a managed container environment, with a dedicated workspace separate from the current Claude chat experience.
Users would be able to connect external services, install skills or add plugins, and move across multiple tools in a workflow similar to switching browser tabs.
The feature is also expected to support webhooks, browser control and notifications, allowing it to wake when called by a user or an external service and then perform work automatically.
Why it matters
The plan points to a product shift for Claude.
Instead of only answering prompts, Anthropic appears to be preparing a system that can stay active across sessions, remember information and operate inside connected work tools.
That could matter for enterprises and developers that want AI assistants to handle longer-running workflows while still keeping permission and context under user control.
The memory work is especially important.
The planned design moves beyond simple summaries toward a file-based memory structure.
If implemented as described, it could let the agent store, classify and reorganize user information over time and use it as a shared knowledge layer across Claude services.
Who is affected
Orbit targets productivity users by collecting information from services including Gmail, Slack, GitHub, Calendar, Drive and Figma.
Its purpose is to understand work patterns and provide information proactively rather than waiting for every prompt.
Operon targets life-science researchers and is described as a fourth work mode after Claude Chat, Claude Code and Claude Cowork.
It would support project sessions in a separate security environment, plan and auto-execution modes, and local-file access for complex research tasks such as CRISPR experiment design and single-cell RNA analysis.
Software teams are also in scope.
BugCrawl is being developed to track ordinary software bugs rather than security vulnerabilities.
It would pull bug tickets from GitHub, Jira and Linear, generate tests, verify fixes and monitor deployment.
What to watch next
The question is whether these testing signals become commercial Claude features.
If Conway and Orbit launch, Claude could compete less as a chatbot and more as a persistent work platform.
Readers should watch how Anthropic handles permissions, memory controls and external-service access, because those details would shape whether always-on agents are trusted for daily work.
