ChatGPT 桌面端支持导入其他智能体工作数据
Core Highlights
The ChatGPT desktop app has added cross-agent data import, letting the output of other AI agents sync in real time with ChatGPT Work and Codex. Users can import projects, chat histories, skills, and plugins, and can view import history and enable automatic updates. This effectively opens a "workflow pipeline" between different agents that previously could not see each other's work. The move signals OpenAI's ambition to make ChatGPT less a single assistant and more a hub where multiple agents cooperate on shared material. For anyone running a stack of specialized tools, that changes how daily work is organized and reduces the duplicated effort that piles up when each tool hoards its own context and refuses to share it.
It also hints at a future where the desktop app becomes the default control plane for a user's entire scattered AI toolchain and history.
What Happened
Users can now bring an external AI agent's project files, conversation traces, custom skills, and plugin lists into the ChatGPT desktop app in one go. Once imported, the content appears in the workspaces of ChatGPT Work and Codex and can auto-update as the source changes, so a revision made elsewhere shows up without manual re-upload. A settings panel provides an import-history list for easy review and removal of a sync, giving users a clear audit trail over what entered their environment. The feature is now live in the desktop client rather than limited to the web interface, which matters for people who live inside the app all day and treat it as their primary workspace rather than a tab they close between unrelated tasks.
Early users report that syncing a research agent's notes into Codex noticeably reduced the need to re-explain context from scratch each session.
Technical Details
The key to this capability is a unified artifact and context-description format. Artifacts from different agents are converted into an intermediate representation that ChatGPT can parse, then mapped onto the project structures of Work and Codex without losing provenance or breaking references between files. Auto-update relies on incremental sync, pushing only changed fragments to cut bandwidth and token cost, which keeps recurring imports cheap even at scale. On permissions, import requires explicit user authorization, preventing agents from reading each other's data without consent and limiting the blast radius of a compromised or misbehaving integration that might otherwise reach across workspaces it was never meant to touch or trust.
The format is deliberately vendor-neutral in spirit, so a third-party agent can publish compatible artifacts without asking OpenAI for special permission.
Versus Competitors
Compared with Claude's Projects, Google's Gemini, and various RPA tools, ChatGPT takes "multi-agent collaboration" from concept to desktop practice with a concrete import surface. Most competitors still keep a single assistant in a closed loop, and moving workflows across products usually means manual copying or brittle scripts that break on the next update. ChatGPT's standardized import reduces the friction of switching between multiple AIs and makes the desktop app a coordination layer rather than just a chat window. That is a different bet from rivals that optimize a single model's quality inside one walled garden and leave interoperability to the user, who pays the integration tax in lost time and repeated copy-paste.
The risk is fragmentation if every vendor invents its own import schema and users face yet another hidden integration tax across tools.
Industry Impact and Use Cases
For heavy AI users, this means no more shuttling results between assistants; research and development flows can be chained together into one continuous workspace that remembers context across tools instead of restarting from zero. For enterprises, it helps unify knowledge accumulation across multi-agent teams and reduces duplicated effort between departments that each run their own tools in isolation. For tool builders, it creates an incentive to expose clean, importable artifacts rather than opaque chat logs that no other system can read or act on. Simply put, AI agents used to be isolated islands; ChatGPT now gives them a shared transit hub that makes the whole ecosystem more useful and far less redundant, so the value of each agent compounds instead of evaporating at every boundary it cannot cross.
If adoption grows, hiring and onboarding may shift toward importing a candidate's prior agent context as a portable sample of their real work.