PilotDeck 2.0 is an open source AI agent workbench that separates projects into WorkSpaces and pairs visible memory with model routing. The October 15, 2026 release announcement highlights one click setup, office document Skills and long tasks spanning up to 2,000 interaction messages.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer

Create a landscape editorial hero image for this Studio Global article: What is PilotDeck 2.0, the open-source agent operating system released by Tsinghua University’s THUNLP lab, ModelBest, OpenBMB and AI9Stars,. Article summary: PilotDeck 2.0 is an open-source AI-agent workbench jointly developed by Tsinghua University’s THUNLP lab, ModelBest, OpenBMB and AI9Stars; “agent operating system” describes its role in coordinating planning, tools, file. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
PilotDeck 2.0 is an open-source AI-agent workbench developed by Tsinghua University’s THUNLP lab, ModelBest, OpenBMB and AI9Stars. The project calls it an “agent operating system,” but its role is to coordinate an agent’s tasks, tools, files and memory—not to replace a computer’s operating system. Its October 15, 2026 announcement emphasizes project-separated work, editable memory and support for long-running tasks. 9
15
PilotDeck uses WorkSpaces to keep each project’s files, memory and skills separate. That design is meant to limit irrelevant project context from carrying over into other work; it describes context organization, not a security sandbox. 44
45
Its white-box memory is designed to capture useful information from conversations and retrieve it for later tasks. Project documentation describes records with session and timestamp information, while the published feature summaries say users can inspect, edit, delete or pin memories. 17
40
1
Dream Mode is described as consolidating and deduplicating accumulated memory during idle periods, with the option to roll back changes. These are documented capabilities, not evidence that memory will always be complete or error-free. 1
4
PilotDeck’s October 15 release announcement highlights one-click installation, visual configuration, long-running tasks spanning up to 2,000 interaction messages, and Skills for Word, Excel and PowerPoint. 9 Its Always-on mode is described as supporting background monitoring and execution after a chat window closes, with results saved as local files and progress or summary reports.
20
The system also supports Smart Routing: stronger models can handle planning and decisions while lighter models take on routine subtasks such as polishing or formatting. The aim is to reduce model costs while preserving useful results. PilotDeck materials also describe native Model Context Protocol (MCP) support and use across web, command-line and messaging interfaces. 20
25
The project is listed on GitHub under the AGPL-3.0 license. A source roundup lists macOS, Linux, Windows and Docker deployment support. 14
1
The reported savings are promising, but they should be read as results from particular tests—not as a general guarantee. In a social-media workflow, QbitAI reports a cost of $12.58 for an all-Claude Opus 4.5 setup and $2.83 when Opus 4.5 handled planning and Sonnet 4.5 handled text polishing. 20
For a separate seven-task evaluation, QbitAI reports that a Sonnet 4.6 and MiniMax-M2.7 routed setup scored 70.6 at $3.15, compared with 69.1 at $18.36 for Sonnet 4.6 alone. Those reported costs imply a reduction of about 83%, but the available source excerpts do not establish independent replication of the test. 20
A separate WorkBuddy Bench report gives PilotDeck with Qwen3.6-35B-A3B a score of 79.1, compared with 77.3 for Claude Code. This, too, is a reported benchmark result and does not demonstrate that PilotDeck will outperform other tools across different tasks or model configurations. 4
5
PilotDeck 2.0’s main idea is to make agent work more persistent and organized: keep project context in separate WorkSpaces, make stored memory inspectable, and route different subtasks to different models. Its release announcement also emphasizes office-document workflows and longer task runs. The reported cost figures are notable, but they remain tied to specific workloads and should not be treated as a universal savings estimate. 9
20
Studio Global AI
This page includes a source-backed answer you can continue inside Studio Global.
PilotDeck 2.0 is an open source AI agent workbench that separates projects into WorkSpaces and pairs visible memory with model routing.
PilotDeck 2.0 is an open source AI agent workbench that separates projects into WorkSpaces and pairs visible memory with model routing. The October 15, 2026 release announcement highlights one click setup, office document Skills and long tasks spanning up to 2,000 interaction messages.
It is an “agent operating system” in the sense that it coordinates tasks, tools and files—not a replacement for macOS, Linux or Windows.
PilotDeck 2.0 is an open source AI agent workbench that separates projects into WorkSpaces and pairs visible memory with model routing. The October 15, 2026 release announcement highlights one click setup, office document Skills and long tasks spanning up to 2,000 interaction messages.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer

Create a landscape editorial hero image for this Studio Global article: What is PilotDeck 2.0, the open-source agent operating system released by Tsinghua University’s THUNLP lab, ModelBest, OpenBMB and AI9Stars,. Article summary: PilotDeck 2.0 is an open-source AI-agent workbench jointly developed by Tsinghua University’s THUNLP lab, ModelBest, OpenBMB and AI9Stars; “agent operating system” describes its role in coordinating planning, tools, file. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
PilotDeck 2.0 is an open-source AI-agent workbench developed by Tsinghua University’s THUNLP lab, ModelBest, OpenBMB and AI9Stars. The project calls it an “agent operating system,” but its role is to coordinate an agent’s tasks, tools, files and memory—not to replace a computer’s operating system. Its October 15, 2026 announcement emphasizes project-separated work, editable memory and support for long-running tasks. 9
15
PilotDeck uses WorkSpaces to keep each project’s files, memory and skills separate. That design is meant to limit irrelevant project context from carrying over into other work; it describes context organization, not a security sandbox. 44
45
Its white-box memory is designed to capture useful information from conversations and retrieve it for later tasks. Project documentation describes records with session and timestamp information, while the published feature summaries say users can inspect, edit, delete or pin memories. 17
40
1
Dream Mode is described as consolidating and deduplicating accumulated memory during idle periods, with the option to roll back changes. These are documented capabilities, not evidence that memory will always be complete or error-free. 1
4
PilotDeck’s October 15 release announcement highlights one-click installation, visual configuration, long-running tasks spanning up to 2,000 interaction messages, and Skills for Word, Excel and PowerPoint. 9 Its Always-on mode is described as supporting background monitoring and execution after a chat window closes, with results saved as local files and progress or summary reports.
20
The system also supports Smart Routing: stronger models can handle planning and decisions while lighter models take on routine subtasks such as polishing or formatting. The aim is to reduce model costs while preserving useful results. PilotDeck materials also describe native Model Context Protocol (MCP) support and use across web, command-line and messaging interfaces. 20
25
The project is listed on GitHub under the AGPL-3.0 license. A source roundup lists macOS, Linux, Windows and Docker deployment support. 14
1
The reported savings are promising, but they should be read as results from particular tests—not as a general guarantee. In a social-media workflow, QbitAI reports a cost of $12.58 for an all-Claude Opus 4.5 setup and $2.83 when Opus 4.5 handled planning and Sonnet 4.5 handled text polishing. 20
For a separate seven-task evaluation, QbitAI reports that a Sonnet 4.6 and MiniMax-M2.7 routed setup scored 70.6 at $3.15, compared with 69.1 at $18.36 for Sonnet 4.6 alone. Those reported costs imply a reduction of about 83%, but the available source excerpts do not establish independent replication of the test. 20
A separate WorkBuddy Bench report gives PilotDeck with Qwen3.6-35B-A3B a score of 79.1, compared with 77.3 for Claude Code. This, too, is a reported benchmark result and does not demonstrate that PilotDeck will outperform other tools across different tasks or model configurations. 4
5
PilotDeck 2.0’s main idea is to make agent work more persistent and organized: keep project context in separate WorkSpaces, make stored memory inspectable, and route different subtasks to different models. Its release announcement also emphasizes office-document workflows and longer task runs. The reported cost figures are notable, but they remain tied to specific workloads and should not be treated as a universal savings estimate. 9
20
Studio Global AI
This page includes a source-backed answer you can continue inside Studio Global.
PilotDeck 2.0 is an open source AI agent workbench that separates projects into WorkSpaces and pairs visible memory with model routing.
PilotDeck 2.0 is an open source AI agent workbench that separates projects into WorkSpaces and pairs visible memory with model routing. The October 15, 2026 release announcement highlights one click setup, office document Skills and long tasks spanning up to 2,000 interaction messages.
It is an “agent operating system” in the sense that it coordinates tasks, tools and files—not a replacement for macOS, Linux or Windows.